# Launchpad HQ: full data export > Plain-text export of the Launchpad HQ catalog, comparison pages, getting-started guides, workflow templates and model families. As of 2026-10-27. Canonical: https://www.launchpadhq.io/llms-full.txt Index: https://www.launchpadhq.io/llms.txt JSON: https://www.launchpadhq.io/data/v1/index.json Generated: 2026-10-27T00:36:56.292Z Every tool price comes from the tool's own record (the same data as its page and /data/v1/tools/{slug}.json). Prices in guides and workflows are filled in from the same catalog data; none are typed by hand. Model API prices come from the model registry behind /models. Sponsorship never changes inclusion or order. Attribution: Data from Launchpad HQ (https://www.launchpadhq.io). Please credit Launchpad HQ and link to the tool's Launchpad HQ page. License: Free to use with attribution. Facts are dated snapshots; confirm prices on the vendor's site before buying. ## Catalog (221 tools) Canonical: https://www.launchpadhq.io/discover ### Activepieces Canonical: https://www.launchpadhq.io/platform/activepieces Data: https://www.launchpadhq.io/data/v1/tools/activepieces.json Category: Workflow & Automation Company: Activepieces (Open Source) Website: https://activepieces.com Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Primary use: Open-source workflow automation (Zapier alternative) Tasks: workflow-automation Free tier: yes (Free forever, no card: 1,000 credits, unlimited flows, up to 5 users. The official page lists the Free reset as daily in its comparison table but monthly in its FAQ) Starting price: Free, from $20/mo flat Lowest paid plan: Plus, $20/mo flat Plans: - Free: $0 (1,000 credits) - Plus: $20/mo flat (10,000 credits/mo, then $0.007/credit) - Team: $200/mo flat (50,000 credits/mo, 25 users) - Ultimate: Custom - Embed from: $36k/year Alternatives: n8n, Zapier, Make (Integromat), Pabbly Connect Guide: https://www.launchpadhq.io/platform/activepieces/guide Data as of: 2026-10-25 Sources: - https://activepieces.com ### Adobe Acrobat AI Assistant Canonical: https://www.launchpadhq.io/platform/adobe-acrobat-ai-assistant Data: https://www.launchpadhq.io/data/v1/tools/adobe-acrobat-ai-assistant.json Category: Document & PDF Processing Company: Adobe Website: https://acrobat.adobe.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI-powered PDF reading, summarization, and Q&A Free tier: yes (Included with Acrobat paid plans; no standalone free tier; limited preview for free Acrobat users) Starting price: $22.99/mo Lowest paid plan: Included with Acrobat Pro, $22.99/mo Plans: - Included with Acrobat Pro: $22.99/mo Alternatives: ChatPDF, Anara, Gemini Notebook Guide: https://www.launchpadhq.io/platform/adobe-acrobat-ai-assistant/guide Data as of: not dated Sources: - https://acrobat.adobe.com ### Adobe Firefly Canonical: https://www.launchpadhq.io/platform/adobe-firefly Data: https://www.launchpadhq.io/data/v1/tools/adobe-firefly.json Category: Image Generation & Editing Company: Adobe Website: https://firefly.adobe.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Commercially safe AI image generation/editing Tasks: image-generation, graphic-design Free tier: yes (25 generative credits/month free; basic image generation; text effects; generative fill) Starting price: Free, from $199.99/mo Lowest paid plan: Premium, $199.99/mo Plans: - Free: Free (credits) - Included with Adobe CC: See pricing - Premium: $199.99/mo Alternatives: Midjourney, Ideogram, Leonardo AI Guide: https://www.launchpadhq.io/platform/adobe-firefly/guide Data as of: not dated Sources: - https://firefly.adobe.com ### AI21 Labs (Jamba) Canonical: https://www.launchpadhq.io/platform/ai21-labs-jamba Data: https://www.launchpadhq.io/data/v1/tools/ai21-labs-jamba.json Category: International & Regional AI Company: AI21 Labs Website: https://ai21.com Cost tier: FREE Difficulty: BEGINNER Lifecycle: active Primary use: Enterprise AI writing and language models Free tier: yes (Free: API trial with rate limits; playground access; Jamba Mini free tier) Starting price: Free, from Custom Lowest paid plan: Enterprise, Custom Plans: - Free: Free (Free: rate-limited API; pay-per-use for production; Enterprise: custom) - Enterprise: Custom Alternatives: Cohere, OpenAI API, Anthropic API Guide: https://www.launchpadhq.io/platform/ai21-labs-jamba/guide Data as of: not dated Sources: - https://ai21.com ### Aider Canonical: https://www.launchpadhq.io/platform/aider Data: https://www.launchpadhq.io/data/v1/tools/aider.json Category: AI Coding & Development Company: Open Source Website: https://aider.chat Cost tier: FREE Difficulty: ADVANCED Lifecycle: active Primary use: Open-source terminal pair programming Tasks: coding Free tier: yes (Free and open-source (Apache 2.0); uses your own LLM API key) Starting price: Free, from See pricing Lowest paid plan: bring your own API key, See pricing Plans: - Free: Free (Unlimited (API costs apply); supports many LLM providers) - bring your own API key: See pricing Alternatives: Claude Code, OpenAI Codex CLI, Gemini CLI, Cursor Guide: https://www.launchpadhq.io/platform/aider/guide Data as of: not dated Sources: - https://aider.chat ### AIVA Canonical: https://www.launchpadhq.io/platform/aiva Data: https://www.launchpadhq.io/data/v1/tools/aiva.json Category: Music Generation Company: AIVA Technologies Website: https://aiva.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Orchestral/cinematic AI composition Free tier: yes (3 compositions/month; non-commercial; MIDI + MP3 download; limited styles) Starting price: Free, from €15/mo Lowest paid plan: Standard, €15/mo Plans: - Free: Free (Free: 3 tracks/mo, max 3 min each; Standard: 15/mo; Pro: 300/mo) - Standard: €15/mo - Pro: €49/mo Alternatives: Suno, Udio, Soundraw, Beatoven.ai Guide: https://www.launchpadhq.io/platform/aiva/guide Data as of: not dated Sources: - https://aiva.ai ### Aleph Alpha (Luminous) Canonical: https://www.launchpadhq.io/platform/aleph-alpha-luminous Data: https://www.launchpadhq.io/data/v1/tools/aleph-alpha-luminous.json Category: International & Regional AI Company: Aleph Alpha Website: https://aleph-alpha.com Cost tier: ENTERPRISE Difficulty: BEGINNER Lifecycle: active Status: On September 16, 2026, Cohere announced a definitive business combination agreement with Aleph Alpha. Cohere says the deal still needs final regulatory approvals and is expected to close later in 2026. Primary use: European sovereign AI for enterprise Free tier: no (No free tier; enterprise contracts; EU sovereignty focus) Starting price: Custom Lowest paid plan: Enterprise, Custom Plans: - Enterprise: Custom Alternatives: Cohere, Mistral Vibe Guide: https://www.launchpadhq.io/platform/aleph-alpha-luminous/guide Data as of: 2026-10-20 Sources: - https://cohere.com/blog/cohere-and-aleph-alpha-sign-agreement (checked 2026-10-20) ### AlphaSense Canonical: https://www.launchpadhq.io/platform/alphasense Data: https://www.launchpadhq.io/data/v1/tools/alphasense.json Category: Finance & Real Estate AI Company: AlphaSense Website: https://alpha-sense.com Cost tier: ENTERPRISE Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI market intelligence and financial research Free tier: no (Free trial available; no permanent free tier) Starting price: Custom Lowest paid plan: Enterprise, Custom Plans: - Enterprise: Custom Alternatives: Bloomberg GPT / Terminal AI Guide: https://www.launchpadhq.io/platform/alphasense/guide Data as of: not dated Sources: - https://alpha-sense.com ### Amazon Bedrock Canonical: https://www.launchpadhq.io/platform/amazon-bedrock Data: https://www.launchpadhq.io/data/v1/tools/amazon-bedrock.json Category: AI APIs & Developer Services Company: AWS Website: https://aws.amazon.com/bedrock/ Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Managed access to multiple foundation models Free tier: yes (AWS Free Tier: limited model usage for first 2 months; varies by model) Starting price: See pricing Lowest paid plan: Pay-per-use, See pricing Plans: - Pay-per-use: See pricing Alternatives: Google Vertex AI, Azure OpenAI Service, OpenAI API, Anthropic API Guide: https://www.launchpadhq.io/platform/amazon-bedrock/guide Data as of: not dated Sources: - https://aws.amazon.com/bedrock/ ### Amazon Q Developer Canonical: https://www.launchpadhq.io/platform/amazon-q-developer Data: https://www.launchpadhq.io/data/v1/tools/amazon-q-developer.json Category: AI Coding & Development Company: AWS Website: https://aws.amazon.com/q/developer/ Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI coding for AWS and cloud dev Free tier: yes (Free tier: code suggestions in IDE; chat; limited security scanning; CodeWhisperer legacy features) Starting price: Free, from $19/user/mo Lowest paid plan: Pro, $19/user/mo Plans: - Free: Free (Free: code suggestions unlimited; chat limited; Pro: $19/user/mo with higher limits; agents included) - Pro: $19/user/mo Alternatives: GitHub Copilot, Cursor, Devin Desktop, Tabnine Guide: https://www.launchpadhq.io/platform/amazon-q-developer/guide Data as of: not dated Sources: - https://aws.amazon.com/q/developer/ ### Anthropic API Canonical: https://www.launchpadhq.io/platform/anthropic-api Data: https://www.launchpadhq.io/data/v1/tools/anthropic-api.json Category: AI APIs & Developer Services Company: Anthropic PBC Website: https://console.anthropic.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Claude models via API Free tier: no (No ongoing free API tier: you pay per token used. Anthropic's pricing docs say new users receive a small amount of free credits to test the API. The free Claude chat plan is a separate product (see the Claude listing).) Starting price: $0.10 input / $0.50 output per million tokens Lowest paid plan: Claude Haiku 5.5, $0.10 input / $0.50 output per million tokens Plans: - Claude Haiku 5.5: $0.10 input / $0.50 output per million tokens (For prompts up to 100,000 tokens. Prompts over 100,000 tokens: $0.50 input / $2.50 output per million tokens) - Claude Sonnet 5.5: $2 input / $10 output per million tokens - Claude Opus 5.5: $4 input / $20 output per million tokens (Fast mode is 2x standard pricing) - Claude Fable 5.1: $10 input / $50 output per million tokens API pricing: USD per million tokens, input / output. Claude Fable 5.1: $10 / $50. Claude Opus 5.5: $4 / $20. Claude Sonnet 5.5: $2 / $10. Claude Haiku 5.5: $0.10 / $0.50 for prompts up to 100,000 tokens; $0.50 / $2.50 for prompts over 100,000 tokens. Features: - Claude Models on Google Cloud (Anthropic managed) is generally available as of 2026-10-22: Anthropic operates the service and acts as the data processor, billing goes on a single Google Cloud invoice, and usage draws down an existing Google Cloud commitment. Alternatives: OpenAI API, Google Vertex AI, Amazon Bedrock, Azure OpenAI Service Guide: https://www.launchpadhq.io/platform/anthropic-api/guide Data as of: 2026-10-26 Sources: - https://claude.com/pricing (checked 2026-10-14) - https://platform.claude.com/docs/en/about-claude/pricing (checked 2026-10-26) - https://claude.com/resources/articles/claude-models-on-google-cloud-anthropic-managed (checked 2026-10-26) ### Apollo.io Canonical: https://www.launchpadhq.io/platform/apollo-io Data: https://www.launchpadhq.io/data/v1/tools/apollo-io.json Category: Sales, Marketing & SEO AI Company: Apollo.io Website: https://www.apollo.io Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Description: B2B contact database and outreach. It processes third-party personal data., GDPR, CAN-SPAM, and similar outreach-law compliance is the user’s responsibility. Official pricing page (apollo.io/pricing) is JavaScript-rendered; dollar figures below are unverified (third-party pages citing that site in Sep–Oct 2026, not confirmed from the official HTML): Free 900 credits/yr; Basic $49/user/mo annual ($65 monthly); Professional $79 ($99 monthly); Organization $119 annual-only with a 3-seat minimum. Primary use: B2B contact database plus outreach sequences Free tier: yes ($0 Free plan) Starting price: Free, from $49/month Lowest paid plan: Basic, $49/month Plans: - Free: $0/month - Basic: $49/month - Professional: $99/month - Custom: Talk to sales Guide: https://www.launchpadhq.io/platform/apollo-io/guide Data as of: 2026-10-14 Sources: - https://www.apollo.io/pricing (checked 2026-10-14) ### AssemblyAI Canonical: https://www.launchpadhq.io/platform/assemblyai Data: https://www.launchpadhq.io/data/v1/tools/assemblyai.json Category: AI APIs & Developer Services Company: AssemblyAI Website: https://assemblyai.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI transcription and audio intelligence API Tasks: transcription Free tier: yes (Free: limited hours; basic transcription and AI features) Starting price: Universal-2 $0.15/hr Lowest paid plan: Pay-as-you-go pre-recorded, Universal-2 $0.15/hr Plans: - Pay-as-you-go pre-recorded: Universal-2 $0.15/hr - Universal-3.5 Pro: $0.21/hr - $50 free credits on signup: $50 Alternatives: Deepgram Guide: https://www.launchpadhq.io/platform/assemblyai/guide Data as of: 2026-10-14 Sources: - https://assemblyai.com ### Azure OpenAI Service Canonical: https://www.launchpadhq.io/platform/azure-openai-service Data: https://www.launchpadhq.io/data/v1/tools/azure-openai-service.json Category: AI APIs & Developer Services Company: Microsoft Website: https://azure.microsoft.com/en-us/products/ai-services/openai-service Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: OpenAI models on Azure infrastructure Free tier: yes (Azure free account: $200 credits for 30 days; OpenAI models included) Starting price: See pricing Lowest paid plan: Pay-per-token via Azure, See pricing Plans: - Pay-per-token via Azure: See pricing Alternatives: Google Vertex AI, Amazon Bedrock, Anthropic API Guide: https://www.launchpadhq.io/platform/azure-openai-service/guide Data as of: not dated Sources: - https://azure.microsoft.com/en-us/products/ai-services/openai-service ### Beatoven.ai Canonical: https://www.launchpadhq.io/platform/beatoven-ai Data: https://www.launchpadhq.io/data/v1/tools/beatoven-ai.json Category: Music Generation Company: Beatoven.ai Website: https://beatoven.ai Cost tier: FREE Difficulty: BEGINNER Lifecycle: active Primary use: AI background music for videos/podcasts Free tier: yes (5 songs free; up to 15 min per track; royalty-free) Starting price: Free, from See pricing Lowest paid plan: Pro plans, See pricing Plans: - Free: Free (Free: 5 downloads; Pro: unlimited downloads; Business: commercial + custom branding) - Pro plans: See pricing Alternatives: Soundraw, Mubert, AIVA, Suno Guide: https://www.launchpadhq.io/platform/beatoven-ai/guide Data as of: not dated Sources: - https://beatoven.ai ### Bloomberg GPT / Terminal AI Canonical: https://www.launchpadhq.io/platform/bloomberg-gpt-terminal-ai Data: https://www.launchpadhq.io/data/v1/tools/bloomberg-gpt-terminal-ai.json Category: Finance & Real Estate AI Company: Bloomberg LP Website: https://bloomberg.com/professional/ Cost tier: ENTERPRISE Difficulty: ADVANCED Lifecycle: active Primary use: AI-enhanced financial data terminal Free tier: no (No free tier; Bloomberg Terminal subscription required ($24K+/year)) Starting price: ~$24,000/year Lowest paid plan: ~$24,000/year (Terminal), ~$24,000/year Plans: - ~$24,000/year (Terminal): ~$24,000/year Alternatives: AlphaSense Guide: https://www.launchpadhq.io/platform/bloomberg-gpt-terminal-ai/guide Data as of: not dated Sources: - https://bloomberg.com/professional/ ### Bolt.new Canonical: https://www.launchpadhq.io/platform/bolt-new Data: https://www.launchpadhq.io/data/v1/tools/bolt-new.json Category: AI Coding & Development Company: StackBlitz Website: https://bolt.new Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI full-stack web app builder in browser Tasks: coding Free tier: yes (Free: 1M tokens/month with a 300K daily limit, Bolt branding, 10MB upload limit) Starting price: Free, from $25/mo billed monthly Lowest paid plan: Pro, $25/mo billed monthly Plans: - Free: $0 (Free: 1M tokens/mo (300K/day); Pro: from 10M tokens/mo with rollover; Teams: from 10M tokens per member; Enterprise: custom) - Pro: $25/mo billed monthly ($18/mo billed annually) - Teams: $30/member/mo billed monthly ($27 billed annually) - Enterprise: Custom Alternatives: Lovable, v0, Replit Agent Guide: https://www.launchpadhq.io/platform/bolt-new/guide Data as of: 2026-10-25 Sources: - https://bolt.new ### Brisk Teaching Canonical: https://www.launchpadhq.io/platform/brisk-teaching Data: https://www.launchpadhq.io/data/v1/tools/brisk-teaching.json Category: Education & Learning AI Company: BRISK Website: https://www.briskteaching.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Browser extension for teachers. Tools can handle student work and student data —, check your district privacy policy and FERPA/COPPA obligations before using it on student identifiers or student work. Official FAQ/plans : Free Forever includes 20+ tools (new users get a 14-day Premium trial first). Premium and Intelligence are custom, priced per student, sold to schools or districts. Individual Educator Pro is not priced on the official plans page. Primary use: Chrome/Edge extension with AI tools inside Google Docs and Classroom Free tier: yes ($0 for teachers (Chrome & Edge extension).) Starting price: Free, from custom pricing Lowest paid plan: Premium, custom pricing Plans: - Free: $0 (For teachers; Chrome & Edge extension) - Premium: custom pricing - Intelligence: custom pricing Data as of: 2026-10-14 Sources: - https://www.briskteaching.com/ (checked 2026-10-14) - https://www.briskteaching.com/plans (checked 2026-10-14) ### Canopy (Tax) Canonical: https://www.launchpadhq.io/platform/canopy-tax Data: https://www.launchpadhq.io/data/v1/tools/canopy-tax.json Category: Finance & Real Estate AI Company: Canopy Website: https://canopytax.com Cost tier: FREEMIUM Difficulty: ADVANCED Lifecycle: active Primary use: AI practice management for tax professionals Free tier: yes (Free trial; no permanent free tier) Starting price: $50/mo Lowest paid plan: From, $50/mo Plans: - From: $50/mo Guide: https://www.launchpadhq.io/platform/canopy-tax/guide Data as of: not dated Sources: - https://canopytax.com ### Canva AI (Magic Studio) Canonical: https://www.launchpadhq.io/platform/canva-ai-magic-studio Data: https://www.launchpadhq.io/data/v1/tools/canva-ai-magic-studio.json Category: Image Generation & Editing Company: Canva Website: https://canva.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI image gen/edit within Canva design platform Tasks: graphic-design, logo-design, slides, image-generation Free tier: yes (Canva AI is available to everyone: a range of AI tools works on Canva's Free plan, and paid plans add more usage and more advanced features. The page does not state free usage limits.) Starting price: Free, from $12.99/mo Lowest paid plan: Pro, $12.99/mo Plans: - Free: Free - Pro: $12.99/mo Features: - AI Pass is an optional monthly add-on for Canva Pro, Teams, Business and Nonprofits plans, priced at USD $100 per person per month (as of 2026-10-26). - AI Pass gives Canva Pro, Teams and Nonprofits subscribers 40x more AI, and Canva Business subscribers 20x more AI. - AI Pass is not available on Canva Free, Education or Enterprise, on free trials or prepaid plans, or on subscriptions managed through the Apple App Store or Google Play. Alternatives: Microsoft Designer, Adobe Firefly, Figma AI Guide: https://www.launchpadhq.io/platform/canva-ai-magic-studio/guide Data as of: 2026-10-26 Sources: - https://www.canva.com/canva-ai/ (checked 2026-10-26) - https://www.canva.com/help/ai-pass/ (checked 2026-10-26) ### CapCut Canonical: https://www.launchpadhq.io/platform/capcut Data: https://www.launchpadhq.io/data/v1/tools/capcut.json Category: Video Creation & Editing Company: CapCut Website: https://www.capcut.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Video editor with a free starting path. Official pricing (capcut.com/pricing): start for free; Pro shown as USD 19.99; Ultra shown as USD 79; a USD 12.49 figure also appeared — billing period for those figures is unverified. Per-second AI video rates were not confirmed on this pass. US services are operated by TikTok USDS Joint Venture LLC per the CapCut US Privacy Policy (updated 2026-04-15, capcut.com/clause/privacy); review that policy for data handling and jurisdiction. Availability can vary by country. Primary use: Free-to-start video editor with paid AI video features Free tier: yes (Free templates; basic editing tools) Starting price: Free, from $12.49 Lowest paid plan: Teams, $12.49/month per seat Plans: - Free: $0/month (Free templates; basic editing tools) - Pro: $19.99/month (1,200 monthly AI credits, 1 TB storage) - Ultra: $79/month (8,645 credits) - Teams: $12.49/month per seat (Shown for 2 seats) Guide: https://www.launchpadhq.io/platform/capcut/guide Data as of: 2026-10-14 Sources: - https://www.capcut.com/pricing (checked 2026-10-14) - https://www.capcut.com/ (checked 2026-10-14) ### Captions Canonical: https://www.launchpadhq.io/platform/captions Data: https://www.launchpadhq.io/data/v1/tools/captions.json Category: Video Creation & Editing Company: NOCAP, Inc. d/b/a Captions Website: https://www.captions.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: AI video editor (phone, computer, and browser). Official homepage (captions.ai): free to sign up for some features; studio-quality edits from $24.99/mo. Higher-tier dollar amounts were not listed on the fetched homepage — unverified. Primary use: AI video editor for captions, avatars, and auto-edits Free tier: yes (1 caption template, no AI credits. Other limits unpublished.) Starting price: Free, from $24.99/month Lowest paid plan: Max, $24.99/month Plans: - Free: Free (1 caption template, no AI credits) - Max: $24.99/month (500 credits/mo) - Frontier: $69.99/month (1,400 credits) - 2x: $139.99/month (2,800 credits) - 4x: $279.99/month (5,600 credits) - Enterprise: Custom pricing Guide: https://www.launchpadhq.io/platform/captions/guide Data as of: 2026-10-14 Sources: - https://www.captions.ai/pricing (checked 2026-10-14) - https://www.captions.ai/download (checked 2026-10-14) ### CoCounsel Canonical: https://www.launchpadhq.io/platform/casetext-cocounsel Data: https://www.launchpadhq.io/data/v1/tools/casetext-cocounsel.json Category: Legal AI Company: Thomson Reuters Website: https://www.thomsonreuters.com/en/cocounsel Cost tier: ENTERPRISE Difficulty: ADVANCED Lifecycle: acquired Status: CoCounsel is now a Thomson Reuters product, and casetext.com redirects to Thomson Reuters' CoCounsel page. Primary use: AI legal research assistant Free tier: no (Free trial available via Thomson Reuters; no permanent free tier) Starting price: See pricing Lowest paid plan: Enterprise (via Thomson Reuters), See pricing Plans: - Enterprise (via Thomson Reuters): See pricing Alternatives: Harvey AI, Spellbook, Luminance Guide: https://www.launchpadhq.io/platform/casetext-cocounsel/guide Data as of: 2026-10-20 Sources: - https://www.thomsonreuters.com/en/cocounsel (checked 2026-10-20) ### Character.AI Canonical: https://www.launchpadhq.io/platform/character-ai Data: https://www.launchpadhq.io/data/v1/tools/character-ai.json Category: Gaming & Creative AI Company: Character Technologies Inc. dba Character.ai Website: https://character.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Status: On October 12, 2026, Character.AI announced a voluntary Chapter 11 process to run a court-supervised sale of the company. It says the platform remains up and running, and that it will be wound down if no buyer keeps it operating. Description: Companion/role-play chatbot. Official Terms (character.ai/tos): users under 13 (or under 16 in the EEA/UK) are not authorized to use the service. Official help on age assurance: users verified under 18 are placed in a reading-only experience without character chat. Parental Insights lets a teen invite a parent or guardian for weekly activity reports (time spent and top Characters; not full chat content) — see character.ai/safety/parental-insights. This listing is not suitable for children; use the official Safety Center and parental controls. Published subscription prices unverified (official pricing did not render this session). Primary use: Companion and role-play chatbot with original characters Free tier: yes (Ads, Limited daily features, slow mode, basic memory, 15 pins. No numeric limits published.) Starting price: $94.99/yr Lowest paid plan: c.ai+, $94.99/yr Plans: - c.ai+: $94.99/yr Data as of: 2026-10-20 Sources: - https://blog.character.ai/an-update-on-character-ai/ (checked 2026-10-20) - https://character.ai/about (checked 2026-10-14) - https://character.ai/subscribe (checked 2026-10-14) ### ChatGPT Canonical: https://www.launchpadhq.io/platform/chatgpt Data: https://www.launchpadhq.io/data/v1/tools/chatgpt.json Category: Text & Conversational AI Company: OpenAI Website: https://chat.openai.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Conversational AI assistant, content generation, image/video via GPT Image 1.5 and Sora 2 Tasks: writing, cover-letter, resume, email-writing, coding, image-generation, summarization Free tier: yes (Unlimited text chats with GPT-6 (abuse guardrails apply)) Starting price: Free, from $8/month Lowest paid plan: Go, $8/month Plans: - Free: $0/month - Go: $8/month (May include ads) - Plus: $20/month - Pro 100: $100/month (Billed monthly only) - Pro 200: $200/month (Billed monthly only) - Pro 500: $500/month (Billed monthly only; only Pro tier with Ultrafast) - Business: $20/user/month (Billed annually, 2+ users; $25 per user/month billed monthly) - Enterprise / Edu: Contact sales Alternatives: Claude, Gemini, Grok, Microsoft Copilot Guide: https://www.launchpadhq.io/platform/chatgpt/guide Data as of: 2026-10-24 Sources: - https://chatgpt.com/pricing (checked 2026-10-24) - https://learn.chatgpt.com/docs/pricing (checked 2026-10-24) ### ChatPDF Canonical: https://www.launchpadhq.io/platform/chatpdf Data: https://www.launchpadhq.io/data/v1/tools/chatpdf.json Category: Document & PDF Processing Company: ChatPDF Website: https://chatpdf.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Chat with PDF documents using AI Free tier: yes (Free: 2 PDFs/day; 120 pages/PDF; 10 questions/day; basic features) Starting price: Free, from $5/mo Lowest paid plan: Plus, $5/mo Plans: - Free: Free (2 PDFs/day) - Plus: $5/mo Alternatives: Adobe Acrobat AI Assistant, Anara, Gemini Notebook Guide: https://www.launchpadhq.io/platform/chatpdf/guide Data as of: not dated Sources: - https://chatpdf.com ### Clarifai Canonical: https://www.launchpadhq.io/platform/clarifai Data: https://www.launchpadhq.io/data/v1/tools/clarifai.json Category: AI APIs & Developer Services Company: Clarifai Website: https://clarifai.com Cost tier: FREE Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI computer vision and NLP platform Free tier: yes (Free: 1,000 operations/month; community plan; basic models) Starting price: Free, from Custom Lowest paid plan: Enterprise, Custom Plans: - Free: Free (Free: 1K ops/mo; Essential: 10K; Professional: 100K; Enterprise: custom) - Enterprise: Custom Alternatives: Scale AI, Google Vertex AI Guide: https://www.launchpadhq.io/platform/clarifai/guide Data as of: not dated Sources: - https://clarifai.com ### Claude Canonical: https://www.launchpadhq.io/platform/claude Data: https://www.launchpadhq.io/data/v1/tools/claude.json Category: Text & Conversational AI Company: Anthropic Website: https://claude.ai Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Description: Anthropic’s assistant for long-context writing, coding, analysis, and reasoning across web and API. Primary use: Coding, analysis, writing assistant Tasks: writing, cover-letter, resume, coding, summarization Free tier: yes (Chat on web, desktop and mobile; search the web, create files and run code; memory across conversations; connect apps and tools; Artifacts; Claude Design, Slides and Docs; up to 5 Projects.) Starting price: Free, from $17/month Lowest paid plan: Pro, $17/month Plans: - Free: $0/month - Pro: $17/month (With annual subscription ($200 billed up front); $20 if billed monthly) - Max: From $100/month (Choose 5x or 20x more usage than Pro; billed monthly) - Team Standard seat: $20/seat/month (Billed annually; $25 if billed monthly. Teams of 2 to 150) - Team Premium seat: $100/seat/month (Billed annually; $125 if billed monthly. 5x more usage than standard seats) - Enterprise: $20/seat/month (Billed annually, 20 seat minimum, plus usage at API rates) Features: - Auto mode becomes the default for new sessions in most Claude apps on Pro, Max and Team plans: Anthropic's 2026-10-22 update says it rolls out the following week (stated plan, as of 2026-10-26). In auto mode a classifier reviews each action, so safe ones go ahead without an approval prompt and risky ones are blocked. Alternatives: ChatGPT, Gemini, Grok, Microsoft Copilot Guide: https://www.launchpadhq.io/platform/claude/guide Data as of: 2026-10-26 Sources: - https://claude.com/pricing (checked 2026-10-26) - https://claude.com/resources/articles/auto-mode-default-in-claude-code (checked 2026-10-26) ### Claude Code Canonical: https://www.launchpadhq.io/platform/claude-code Data: https://www.launchpadhq.io/data/v1/tools/claude-code.json Category: AI Coding & Development Company: Anthropic Website: https://docs.anthropic.com/en/docs/claude-code Cost tier: PAID Difficulty: ADVANCED Lifecycle: active Primary use: Terminal agentic coding agent Tasks: coding Free tier: no (No Claude Code on the Free plan. Available from Pro ($17–20/mo) and Max (from $100/mo)) Starting price: Pro $17/mo with annual ($20 billed monthly) Lowest paid plan: Claude Code is included on paid Claude plans, Pro $17/mo with annual ($20 billed monthly) Plans: - Claude Code is included on paid Claude plans: Pro $17/mo with annual ($20 billed monthly) - Max from: $100/mo - also Team and Enterprise. Not included on Free.: See pricing Features: - Auto mode is the default for new Claude Code sessions on Pro, Max and Team plans. Anthropic announced the switch on 2026-08-07 for new sessions starting 2026-08-14. A classifier reviews each tool call: safe actions go ahead, risky ones (irreversible, destructive or aimed outside your environment) are blocked. Shift+Tab in the CLI, or the mode menu in the desktop app, switches modes. - Auto mode stays opt-in for Claude Enterprise, the Claude API and cloud platforms for now (as of 2026-10-26). Alternatives: Cursor, GitHub Copilot, Devin Desktop, OpenAI Codex CLI Guide: https://www.launchpadhq.io/platform/claude-code/guide Data as of: 2026-10-26 Sources: - https://claude.com/resources/articles/auto-mode-default-in-claude-code (checked 2026-10-26) ### Clay Canonical: https://www.launchpadhq.io/platform/clay Data: https://www.launchpadhq.io/data/v1/tools/clay.json Category: Sales, Marketing & SEO AI Company: Clay Labs Inc. Website: https://www.clay.com Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Description: Enrichment and GTM workflows over third-party personal data (names, emails, employers, and similar)., GDPR, CAN-SPAM, and similar outreach-law compliance is the user’s responsibility — do not treat enrichment as permission to email. Official pricing (clay.com/pricing): Free 500 actions/mo; Launch listed as “$167/mo, starts at 15,000 actions/mo”. The page also shows $54–$60/mo add-on tiers; the exact billing basis for those figures is unverified. Primary use: Data enrichment and go-to-market workflows Free tier: yes (100 data credits and 500 actions/mo, 200 rows per table.) Starting price: $167/mo Lowest paid plan: Launch, $167/mo/billed annually Plans: - Launch: $167/mo/billed annually (Default annual view (Save 10%); $185/mo on monthly) - Growth: $446/mo/billed annually ($495/mo billed monthly) Guide: https://www.launchpadhq.io/platform/clay/guide Data as of: 2026-10-14 Sources: - https://www.clay.com/pricing (checked 2026-10-14) ### Cleo Canonical: https://www.launchpadhq.io/platform/cleo Data: https://www.launchpadhq.io/data/v1/tools/cleo.json Category: Finance & Real Estate AI Company: Cleo AI Ltd. Website: https://web.meetcleo.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Budgeting chat app. it can link bank accounts (account aggregation). Paid tiers include cash advances — those products have separate fees and terms; this listing is not an offer of credit. Official FAQ (meetcleo.com FAQs): Free app; Plus $5.99/mo or $44.99/yr; Pro $8.99/mo; Builder $14.99/mo. Some users pay legacy Plus prices from $1.99. Primary use: AI budgeting chat that can link bank accounts and offer cash advances Free tier: yes (Free app; bank linking and cash-advance terms on paid tiers) Starting price: $5.99/month Lowest paid plan: Plus, $5.99/month Plans: - Plus: $5.99/month (Cash advances up to $250, subject to eligibility) - Pro: $8.99/month (Cash advances up to $250, subject to eligibility) - Builder: $14.99/month (Cash advances up to $500, subject to eligibility) Data as of: 2026-10-14 Sources: - https://web.meetcleo.com/products/pricing (checked 2026-10-14) ### Cline Canonical: https://www.launchpadhq.io/platform/cline Data: https://www.launchpadhq.io/data/v1/tools/cline.json Category: AI Coding & Development Company: Cline Website: https://cline.bot Cost tier: FREE Difficulty: INTERMEDIATE Lifecycle: active Description: Free open-source VS Code extension and CLI. You pay for inference at cost or bring your own key. Client-side architecture; local models are possible so code can stay on your machine. Enterprise is contact-sales (cline.bot/pricing). Primary use: Open-source coding agent in VS Code or CLI, BYOK or pay-for-inference Tasks: coding Free tier: yes (Full client is free; you supply model access (local or a key)) Starting price: See pricing Lowest paid plan: Open source free, See pricing Plans: - Open source free: See pricing - pay for inference or BYOK: See pricing - Enterprise contact sales: See pricing Guide: https://www.launchpadhq.io/platform/cline/guide Data as of: 2026-10-13 Sources: - https://cline.bot ### Clio Manage Canonical: https://www.launchpadhq.io/platform/clio-manage Data: https://www.launchpadhq.io/data/v1/tools/clio-manage.json Category: Legal AI Company: Clio Website: https://clio.com Cost tier: FREEMIUM Difficulty: ADVANCED Lifecycle: active Primary use: AI-powered legal practice management Free tier: yes (7-day free trial; no permanent free tier) Starting price: $49/user/mo Lowest paid plan: From, $49/user/mo Plans: - From: $49/user/mo Guide: https://www.launchpadhq.io/platform/clio-manage/guide Data as of: not dated Sources: - https://clio.com ### Cohere Canonical: https://www.launchpadhq.io/platform/cohere Data: https://www.launchpadhq.io/data/v1/tools/cohere.json Category: International & Regional AI Company: Cohere Website: https://cohere.com Cost tier: FREE Difficulty: BEGINNER Lifecycle: active Primary use: Enterprise LLM API with multilingual focus Free tier: yes (Free: trial API key; rate-limited; basic models; playground access) Starting price: Free, from See pricing Lowest paid plan: Pay-per-use API, See pricing Plans: - Free: Free (trial) - Pay-per-use API: See pricing - Enterprise: Custom Alternatives: AI21 Labs (Jamba), Mistral Vibe, OpenAI API, Anthropic API Guide: https://www.launchpadhq.io/platform/cohere/guide Data as of: not dated Sources: - https://cohere.com ### ComfyUI Canonical: https://www.launchpadhq.io/platform/comfyui Data: https://www.launchpadhq.io/data/v1/tools/comfyui.json Category: Local & Open-Source AI Company: Comfy Org Website: https://www.comfy.org Cost tier: FREE Difficulty: ADVANCED Lifecycle: active Description: Free, open-source local graph UI for image and video models. Data stays on your machine when you run it locally. A capable GPU is typically required. Comfy Cloud is metered by GPU time (docs.comfy.org); the dollar rate is unverified. Primary use: Local node-based image and video generation Free tier: yes (Comfy Desktop itself is free. Cloud: Try Cloud for free, with free new-user credits (amount unpublished).) Starting price: $20/month Lowest paid plan: Standard, $20/month Plans: - Standard: $20/month ($16/mo, $192 billed yearly; per workspace) - Creator: $35/month ($28/mo, $336 yearly; per workspace) - Pro: $100/month ($80/mo on yearly billing; per workspace) - Team: $200–$2,500/month (Five tiers; monthly credit commitment, not per seat) API pricing: Comfy API billed per GPU-second: RTX PRO 6000 $4.54/hr, H100 $6.23/hr, H200 $7.71/hr, B200 $11.23/hr. Storage $0.20/GB/mo. MiniMax commercial license from $5,000/mo. Guide: https://www.launchpadhq.io/platform/comfyui/guide Data as of: 2026-10-14 Sources: - https://www.comfy.org/ (checked 2026-10-14) - https://www.comfy.org/pricing (checked 2026-10-14) - https://www.comfy.org/cloud (checked 2026-10-14) ### Connected Papers Canonical: https://www.launchpadhq.io/platform/connected-papers Data: https://www.launchpadhq.io/data/v1/tools/connected-papers.json Category: Research & Academic Tools Company: Connected Papers Website: https://connectedpapers.com Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Primary use: Visual citation network graphs for paper discovery Tasks: research-papers Free tier: yes (Free: 5 graphs/month; visual citation exploration; basic features) Starting price: Free, from $3/mo Lowest paid plan: Academic, $3/mo Plans: - Free: Free (Free: 5 graphs/mo; Academic: unlimited graphs + prior/derivative works; Team: collaboration) - Academic: $3/mo Alternatives: Research Rabbit, Litmaps, Semantic Scholar Guide: https://www.launchpadhq.io/platform/connected-papers/guide Data as of: not dated Sources: - https://connectedpapers.com ### Consensus Canonical: https://www.launchpadhq.io/platform/consensus Data: https://www.launchpadhq.io/data/v1/tools/consensus.json Category: Research & Academic Tools Company: Consensus Website: https://consensus.app Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI search engine for evidence-based scientific answers Tasks: research-papers Free tier: yes (Free: unlimited searches; AI-synthesized answers; basic paper results) Starting price: Free, from $8.99/mo Lowest paid plan: Premium, $8.99/mo Plans: - Free: Free (Free: unlimited search; Premium: enhanced AI synthesis + GPT-4 analysis + bookmarks) - Premium: $8.99/mo Alternatives: Elicit, Semantic Scholar, Scite Guide: https://www.launchpadhq.io/platform/consensus/guide Data as of: not dated Sources: - https://consensus.app ### Copy.ai Canonical: https://www.launchpadhq.io/platform/copy-ai Data: https://www.launchpadhq.io/data/v1/tools/copy-ai.json Category: Text & Conversational AI Company: Copy.ai Website: https://copy.ai Cost tier: PAID Difficulty: BEGINNER Lifecycle: active Primary use: AI marketing copy and sales automation Tasks: marketing-copy, email-writing Free tier: no (No free plan listed on copy.ai/prices; prior 2,000-words/mo claim could not be confirmed) Starting price: $29/mo billed monthly Lowest paid plan: Chat, $29/mo billed monthly Plans: - Chat: $29/mo billed monthly ($24/mo billed $288/yr) - Growth: $1,000/mo ($12,000/yr) - Expansion: $2,000/mo - Scale: $3,000/mo - Enterprise: Custom Alternatives: Jasper, Writesonic / Chatsonic, ChatGPT, Notion AI Guide: https://www.launchpadhq.io/platform/copy-ai/guide Data as of: 2026-10-14 Sources: - https://copy.ai ### Cursor Canonical: https://www.launchpadhq.io/platform/cursor Data: https://www.launchpadhq.io/data/v1/tools/cursor.json Category: AI Coding & Development Company: Anysphere Website: https://cursor.com Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI-first IDE with deep codebase understanding Tasks: coding Free tier: yes (Free $0: limited AI usage in Cursor) Starting price: Free, from $8/mo Lowest paid plan: Plus, $8/mo Plans: - Free: $0 (Each tier includes a usage pool shared across Cursor and Grok that refreshes weekly; Free: limited; Plus: basic; Premium 4x, Super 15x, Ultra 50x Plus; Teams pooled) - Plus: $8/mo (Grok models only) - Premium: $30/mo (4x Plus usage) - Super: $100/mo (15x Plus) - Ultra: $200/mo (50x Plus) - Teams from: $30/user/mo - Enterprise: Custom (. Third-party models are billed separately by usage. Sold only through X Subscription Corp) Alternatives: GitHub Copilot, Devin Desktop, Claude Code, Amazon Q Developer Guide: https://www.launchpadhq.io/platform/cursor/guide Data as of: 2026-10-25 Sources: - https://cursor.com ### Deepgram Canonical: https://www.launchpadhq.io/platform/deepgram Data: https://www.launchpadhq.io/data/v1/tools/deepgram.json Category: AI APIs & Developer Services Company: Deepgram Website: https://deepgram.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI speech-to-text and text-to-speech API Tasks: transcription Free tier: yes (Free: $200 credit for new accounts; pay-per-use after) Starting price: $0.0048/min promotional (strikethrough regular $0.0077/min). $200 credit on signup. Lowest paid plan: Nova-3 Monolingual pre-recorded currently, $0.0048/min promotional (strikethrough regular $0.0077/min). $200 credit on signup. Plans: - Pay-as-you-go: See pricing - Nova-3 Monolingual pre-recorded currently: $0.0048/min promotional (strikethrough regular $0.0077/min). $200 credit on signup. Alternatives: AssemblyAI Guide: https://www.launchpadhq.io/platform/deepgram/guide Data as of: 2026-10-14 Sources: - https://deepgram.com ### DeepL Canonical: https://www.launchpadhq.io/platform/deepl Data: https://www.launchpadhq.io/data/v1/tools/deepl.json Category: International & Regional AI Company: DeepL Website: https://www.deepl.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Translation product. Official Pro page (deepl.com/pro): Free version includes 50,000 characters/user/month and 1 file translation. Individual (Pro) €7.49 per month billed annually (VAT included as stated). Team and Business list prices did not render as euro figures on this fetch — unverified. Pro states original content is not used to train DeepL’s models. Primary use: Machine translation for text and files Tasks: translation Free tier: yes (50,000 characters/month, 1 file translation/month (up to 5 MB), 1 glossary. Plan limits are listed per user/month.) Starting price: €7.49 Lowest paid plan: Individual, €7.49/month, EUR, billed annually Plans: - Individual: €7.49/month, EUR, billed annually (300k characters, 3 files; VAT included) - Team: €24.99/user/month, EUR, billed annually (1M characters, 20 files; VAT included) - Business: €49.99/user/month, EUR, billed annually (Unlimited characters (fair use), 100 files; VAT included) - Enterprise: Custom API pricing: USD. Developer free: 1M characters, one-time. Growth: $26/mo billed annually plus usage ($27.50 per 1M extra characters). Enterprise: custom. Guide: https://www.launchpadhq.io/platform/deepl/guide Data as of: 2026-10-14 Sources: - https://www.deepl.com/en/pro (checked 2026-10-14) - https://www.deepl.com/en/pro-api (checked 2026-10-14) ### DeepSeek Canonical: https://www.launchpadhq.io/platform/deepseek Data: https://www.launchpadhq.io/data/v1/tools/deepseek.json Category: International & Regional AI Company: DeepSeek Website: https://www.deepseek.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Chat app and API from a China-based company., hosted service processes data in China — review before sending sensitive or business inputs. Open weights can be run locally so data stays on your machine. Official API pricing (deepseek-flash / DeepSeek-V4.1-Flash): $0.30 input (cache miss) / $1.20 output per 1M tokens at peak; off-peak is half. V4-Pro: $1.32 in / $3.96 out at peak. Context 1M tokens. Peak hours 01:00–04:00 and 06:00–10:00 UTC on weekdays. Free consumer chat-app limits are unverified. Primary use: Budget chat app and pay-as-you-go API Free tier: yes (Free consumer chat app; exact daily/monthly limits unverified) Starting price: Free, from $0.30/$1.20 per 1M tokens in/out Lowest paid plan: API deepseek-flash peak, $0.30/$1.20 per 1M tokens in/out Plans: - API deepseek-flash peak: $0.30/$1.20 per 1M tokens in/out (off-peak half) - V4-Pro peak: $1.32/$3.96 - Free: Free (chat app limits unverified) API pricing: Per 1M tokens at peak rate (off-peak is half). Flash: input $0.30 (cache miss) / $0.006 (cache hit); output $1.20. V4-Pro: input $1.32 (cache miss) / $0.044 (cache hit); output $3.96. Context window 1M. Search and Fetch tool APIs (added 2026-10-17): $0.75 per 1,000 successful requests each. Guide: https://www.launchpadhq.io/platform/deepseek/guide Data as of: 2026-10-24 Sources: - https://www.deepseek.com/en (checked 2026-10-14) - https://api-docs.deepseek.com/quick_start/pricing (checked 2026-10-24) - https://api-docs.deepseek.com/updates (checked 2026-10-17) ### Descript Canonical: https://www.launchpadhq.io/platform/descript Data: https://www.launchpadhq.io/data/v1/tools/descript.json Category: Voice & Speech Company: Descript Website: https://descript.com Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Description: Edit audio and video by editing text transcripts; overdub, studio sound, and collaborative timelines. Primary use: Audio/video editing with AI transcription Tasks: transcription Free tier: yes (Free $0: 60 media minutes/mo, 720p watermarked export, limited AI) Starting price: Free, from $24/person/mo monthly or $16 annual Lowest paid plan: Hobbyist, $24/person/mo monthly or $16 annual Plans: - Free: $0 (Free: 1 hr transcription + 1 export; Hobbyist $24/mo: 10 hrs; Business $33/mo: 30 hrs) - Hobbyist: $24/person/mo monthly or $16 annual - Creator: $35/$24 - Business: $65/$50 per person - Enterprise: Custom Alternatives: InVideo AI, Opus Clip, Pictory AI, CapCut Guide: https://www.launchpadhq.io/platform/descript/guide Data as of: 2026-10-14 Sources: - https://descript.com ### Docsumo Canonical: https://www.launchpadhq.io/platform/docsumo Data: https://www.launchpadhq.io/data/v1/tools/docsumo.json Category: Document & PDF Processing Company: Docsumo Website: https://docsumo.com Cost tier: FREE Difficulty: BEGINNER Lifecycle: active Primary use: AI document data extraction and processing Free tier: yes (Free: 100 pages/month; basic document extraction; email support) Starting price: Free, from See pricing Lowest paid plan: paid plans, See pricing Plans: - Free: Free (trial) - paid plans: See pricing Alternatives: Nanonets, Reducto, Upstage (Solar) Guide: https://www.launchpadhq.io/platform/docsumo/guide Data as of: not dated Sources: - https://docsumo.com ### Doximity Ask Canonical: https://www.launchpadhq.io/platform/doximity-gpt Data: https://www.launchpadhq.io/data/v1/tools/doximity-gpt.json Category: Healthcare AI Company: Doximity, Inc. Website: https://www.doximity.com/gpt Cost tier: FREE Difficulty: BEGINNER Lifecycle: active Description: Writing assistant on Doximity. Official page (doximity.com/gpt): free for verified US clinicians. Access is gated to licensed clinicians — outputs are not a substitute for clinical judgment or patient medical advice. HIPAA / BAA claims are unverified from the fetched public page. Primary use: Clinical writing assistant for verified US clinicians Free tier: yes (Free for verified US clinicians) Starting price: Free, from See pricing Lowest paid plan: no public paid consumer tier, See pricing Plans: - Free: Free (for verified US clinicians) - no public paid consumer tier: See pricing Data as of: 2026-10-14 Sources: - https://www.doximity.com/ (checked 2026-10-14) ### Dragon Medical One Canonical: https://www.launchpadhq.io/platform/dragon-medical-one Data: https://www.launchpadhq.io/data/v1/tools/dragon-medical-one.json Category: Healthcare AI Company: Microsoft (Nuance) Website: https://nuance.com Cost tier: ENTERPRISE Difficulty: ADVANCED Lifecycle: active Primary use: AI clinical documentation and medical dictation Free tier: no (No free tier; subscription via healthcare organization) Starting price: See pricing Lowest paid plan: Enterprise/subscription, See pricing Plans: - Enterprise/subscription: See pricing Guide: https://www.launchpadhq.io/platform/dragon-medical-one/guide Data as of: not dated Sources: - https://nuance.com ### Duolingo Max Canonical: https://www.launchpadhq.io/platform/duolingo-max Data: https://www.launchpadhq.io/data/v1/tools/duolingo-max.json Category: Education & Learning AI Company: Duolingo Website: https://duolingo.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI-powered language learning Free tier: yes (Duolingo free: full lessons with ads; limited hearts; basic features) Starting price: Free, from $12.99/mo Lowest paid plan: Super, $12.99/mo Plans: - Free: Free (Free: ad-supported + heart limits; Super: $7/mo no ads; Max: $14/mo adds AI features) - Super: $12.99/mo - Max: $29.99/mo Alternatives: Speak (Language AI) Guide: https://www.launchpadhq.io/platform/duolingo-max/guide Data as of: not dated Sources: - https://duolingo.com ### ElevenLabs Canonical: https://www.launchpadhq.io/platform/elevenlabs Data: https://www.launchpadhq.io/data/v1/tools/elevenlabs.json Category: Voice & Speech Company: ElevenLabs Website: https://elevenlabs.io Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Voice synthesis, TTS, voice cloning, music, sound effects Tasks: text-to-speech Free tier: yes (Free $0: 10,000 credits/month; 3 Studio projects) Starting price: Free, from $5 equivalent, Creator $18.33, Pro $82.50) Lowest paid plan: Enterprise custom. Annual = 10 months (Starter, $5 equivalent, Creator $18.33, Pro $82.50) Plans: - Free: $0 (10k credits/mo) - Starter: $6/mo (30k credits; first-month promo $1 shown until Oct 18, 2026) - Creator: $22/mo (121k; first-month $11 shown) - Pro: $99 (600k) - Scale: $299 (1.8M, 3 seats) - Business: $990 (6M, 10 seats) - Enterprise custom. Annual = 10 months (Starter: $5 equivalent, Creator $18.33, Pro $82.50) Guide: https://www.launchpadhq.io/platform/elevenlabs/guide Data as of: 2026-10-14 Sources: - https://elevenlabs.io ### Elicit Canonical: https://www.launchpadhq.io/platform/elicit Data: https://www.launchpadhq.io/data/v1/tools/elicit.json Category: Research & Academic Tools Company: Elicit (Ought) Website: https://elicit.com Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI research assistant for systematic literature review Tasks: research-papers Free tier: yes (Free: 5,000 paper results; basic extraction; limited concept maps) Starting price: Free, from $10/mo Lowest paid plan: Plus, $10/mo Plans: - Free: Free (Free: 5K results; Plus: unlimited searches + 25 extractions/mo; Pro: unlimited all) - Plus: $10/mo - Pro plans: See pricing Alternatives: Consensus, SciSpace, Semantic Scholar, Scite Guide: https://www.launchpadhq.io/platform/elicit/guide Data as of: not dated Sources: - https://elicit.com ### EvenUp Canonical: https://www.launchpadhq.io/platform/evenup Data: https://www.launchpadhq.io/data/v1/tools/evenup.json Category: Legal AI Company: EvenUp Inc Website: https://www.evenuplaw.com Cost tier: ENTERPRISE Difficulty: INTERMEDIATE Lifecycle: active Description: AI platform for personal-injury law firms. Official site (evenuplaw.com): no published list price. Contact sales (enterprise, no published price). Site states SOC 2-audited and HIPAA-attested. Treat outputs as assistance, not legal advice. Primary use: AI for personal-injury law firms Free tier: no (No public free plan) Starting price: No public pricing (demo only) Lowest paid plan: Demo, No public pricing (demo only) Plans: - Demo: No public pricing (demo only) (Schedule a Demo) Data as of: 2026-10-14 Sources: - https://www.evenuplaw.com/ (checked 2026-10-14) ### Fathom Canonical: https://www.launchpadhq.io/platform/fathom Data: https://www.launchpadhq.io/data/v1/tools/fathom.json Category: Meetings & Notes Company: Fathom Website: https://www.fathom.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Meeting recorder at fathom.ai (not fathom.video). Recording consent applies: tell attendees and follow all-party consent laws where they apply. Official pricing : Free with unlimited recordings and transcription; Premium $20/mo ($16 annual); Team $19/user ($15 annual, 2-user minimum); Business $34/user ($25 annual). Primary use: Free unlimited meeting recording and transcription Tasks: meeting-notes, transcription Free tier: yes (Free forever plan.) Starting price: Talk to sales Lowest paid plan: Enterprise, Talk to sales Plans: - Team: (2-user minimum) - Business: (2-user minimum) - Enterprise: Talk to sales Guide: https://www.launchpadhq.io/platform/fathom/guide Data as of: 2026-10-14 Sources: - https://www.fathom.ai/pricing (checked 2026-10-14) ### Figma AI Canonical: https://www.launchpadhq.io/platform/figma-ai Data: https://www.launchpadhq.io/data/v1/tools/figma-ai.json Category: AI Plugins for Business Software Company: Figma Website: https://figma.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI features in Figma design tool Free tier: yes (Figma free plan: limited AI features; 3 Figma files; basic design tools) Starting price: See pricing Lowest paid plan: Included with Figma plans, See pricing Plans: - Included with Figma plans: See pricing Guide: https://www.launchpadhq.io/platform/figma-ai/guide Data as of: not dated Sources: - https://figma.com ### Firecrawl Canonical: https://www.launchpadhq.io/platform/firecrawl Data: https://www.launchpadhq.io/data/v1/tools/firecrawl.json Category: AI APIs & Developer Services Company: Firecrawl Website: https://www.firecrawl.dev Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Description: Web crawl/scrape API. Official pricing (firecrawl.dev/pricing): Free 1,000 credits/mo, no card; Hobby $16/mo annual (5,000 credits); Standard $83/mo annual (100k); Growth $333/mo annual (500k). Open source with a self-host option. Primary use: Crawl and convert websites into LLM-ready data Free tier: yes (Free plan (credit amount unpublished).) Starting price: $16/mo Lowest paid plan: Hobby, $16/mo/billed annually Plans: - Hobby: $16/mo/billed annually (5,000 credits/mo) - Standard: $83/mo/billed annually (100,000 credits) - Growth: $333/mo/billed annually (500,000 credits) - Scale: $599/mo/billed annually (1,000,000 credits) Data as of: 2026-10-14 Sources: - https://www.firecrawl.dev/pricing (checked 2026-10-14) ### Fireflies.ai Canonical: https://www.launchpadhq.io/platform/fireflies-ai Data: https://www.launchpadhq.io/data/v1/tools/fireflies-ai.json Category: Meetings & Notes Company: Fireflies.ai Corp. Website: https://fireflies.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: A bot joins your calls and transcribes them.: some US states require all-party consent to record — tell every attendee before the bot joins and follow the law where participants are located. Official pricing (fireflies.ai/pricing): Free forever; Pro $18/mo ($10 annual); Business $29 ($19 annual); Enterprise $39 (annual only). Free-tier transcription is described as unlimited; other free limits are unverified. Primary use: Meeting transcription and searchable notes via a meeting bot Tasks: meeting-notes, transcription Free tier: yes ($0. Unlimited transcription*, limited AI summaries, 400 mins of storage per team, and 20 AI credits.) Starting price: Free, from $18 Lowest paid plan: Pro, $18/seat/month Plans: - Free: $0 (Unlimited transcription*, limited AI summaries, 400 mins of storage per team, 20 AI credits) - Pro: $18/seat/month ($10 billed annually) - Business: $29/seat/month ($19 annually) - Enterprise: $39/seat, annual only Guide: https://www.launchpadhq.io/platform/fireflies-ai/guide Data as of: 2026-10-14 Sources: - https://fireflies.ai/pricing (checked 2026-10-14) ### Fireworks AI Canonical: https://www.launchpadhq.io/platform/fireworks-ai Data: https://www.launchpadhq.io/data/v1/tools/fireworks-ai.json Category: AI APIs & Developer Services Company: Fireworks AI Website: https://fireworks.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Fast, cheap inference for open models Free tier: yes (Free: $1 credit for new accounts; generous rate limits on free models) Starting price: See pricing Lowest paid plan: Pay-per-token, See pricing Plans: - Pay-per-token: See pricing - generous free tier: See pricing Alternatives: Groq, Together AI, Replicate, OpenAI API Guide: https://www.launchpadhq.io/platform/fireworks-ai/guide Data as of: not dated Sources: - https://fireworks.ai ### FLUX Canonical: https://www.launchpadhq.io/platform/flux Data: https://www.launchpadhq.io/data/v1/tools/flux.json Category: Image Generation & Editing Company: Black Forest Labs Website: https://blackforestlabs.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: High-quality photorealistic image generation Tasks: image-generation Free tier: yes (FLUX.1 Schnell (fast) is open-source and free to self-host; hosted services vary) Starting price: ~$0.048/image (768×768, 1 image). Open-weights licensing (Builder/Platform) is separate and custom. Lowest paid plan: Pay as you go, no subscriptions. Official API calculator default, ~$0.048/image (768×768, 1 image). Open-weights licensing (Builder/Platform) is separate and custom. Plans: - Pay as you go, no subscriptions. Official API calculator default: ~$0.048/image (768×768, 1 image). Open-weights licensing (Builder/Platform) is separate and custom. Alternatives: Midjourney, Leonardo AI, Adobe Firefly, Ideogram Guide: https://www.launchpadhq.io/platform/flux/guide Data as of: 2026-10-14 Sources: - https://blackforestlabs.ai ### Magnific (formerly Freepik) Canonical: https://www.launchpadhq.io/platform/freepik-ai Data: https://www.launchpadhq.io/data/v1/tools/freepik-ai.json Category: Image Generation & Editing Company: Freepik Website: https://www.freepik.com/ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Freepik’s AI surface. Official pricing page (freepik.com/pricing) is branded Magnific (formerly Freepik) and lists paid credit-based plans; dollar sticker prices for the consumer tiers did not render as a complete table on this fetch — unverified. A free path is advertised; exact free limits unverified. Primary use: Stock plus AI image and video generation Tasks: image-generation Free tier: yes (Free path advertised; limits unverified) Starting price: $20/month Lowest paid plan: Premium, $20/month Plans: - Premium: $20/month ($14.50/mo billed annually) - Premium+: $45/month ($33.75/mo billed annually) - Pro: $110/month ($82.50/mo billed annually) Guide: https://www.launchpadhq.io/platform/freepik-ai/guide Data as of: 2026-10-17 Sources: - https://www.magnific.com/pricing (checked 2026-10-14) - https://www.magnific.com/ai/docs/ai-credits-and-limits (checked 2026-10-17) ### Gamma Canonical: https://www.launchpadhq.io/platform/gamma Data: https://www.launchpadhq.io/data/v1/tools/gamma.json Category: Document & PDF Processing Company: Gamma Tech, Inc. Website: https://gamma.app Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Presentation and document generator. Official tiers (gamma.app/pricing): Free 400 one-time credits, up to 10 slides per prompt; Plus $12/mo ($9 annual); Pro $25/mo ($18 annual); Ultra $100/mo ($90 annual); Business $40/seat (annual). SOC 2 Type II certified. Primary use: AI-generated presentations and docs Tasks: slides Free tier: yes ($0. 400 credits at signup, 10 slides per prompt.) Starting price: Free, from $12 Lowest paid plan: Plus, $12/per seat per month, billed monthly Plans: - Free: $0 (400 credits at signup, 10 slides per prompt) - Plus: $12/per seat per month, billed monthly ($9 billed annually; 1,000 credits a month) - Pro: $25/per seat per month, billed monthly ($18 billed annually; 4,000 credits a month) - Ultra: $100/per seat per month, billed monthly ($90 billed annually; 20,000 credits a month) Guide: https://www.launchpadhq.io/platform/gamma/guide Data as of: 2026-10-17 Sources: - https://gamma.app/pricing (checked 2026-10-14) - https://help.gamma.app/en/articles/7834324-how-do-credits-work-in-gamma (checked 2026-10-17) ### Gemini Canonical: https://www.launchpadhq.io/platform/gemini Data: https://www.launchpadhq.io/data/v1/tools/gemini.json Category: Text & Conversational AI Company: Google DeepMind Website: https://gemini.google.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Multimodal AI, research, images, video, long-context analysis Tasks: writing, cover-letter, resume, email-writing, image-generation Free tier: yes (Free with a Google Account: Gemini app with the Flash-Lite model, image generation and editing, Deep Research, Gemini Live, Canvas and Gems; 15 GB of cloud storage) Starting price: Free, from $4.99/month Lowest paid plan: Google AI Plus, $4.99/month Plans: - Free: $0/month (With a Google Account) - Google AI Plus: $4.99/month (2x higher usage limits than Free; 400 GB storage) - Google AI Pro: $19.99/month (4x higher usage limits than Free; 5 TB storage) - Google AI Ultra: $99.99/month (Starting price: 5x higher usage limits vs. AI Pro; $199.99/month for 20x) - Gemini Enterprise Business edition: $21/seat/month (Starting price; up to 300 seats) Features: - Gemini app usage limits are compute-based: they factor in prompt complexity, the features used and chat length, refresh every 5 hours until a weekly limit is reached, and can be extended by buying AI credits. - Paid personal plans include Google Flow credits for its AI creative studio: 200 on Google AI Plus, 1,000 on Google AI Pro, and 10,000 or 25,000 on Google AI Ultra. Alternatives: ChatGPT, Claude, Grok, Microsoft Copilot Guide: https://www.launchpadhq.io/platform/gemini/guide Data as of: 2026-10-26 Sources: - https://gemini.google/subscriptions/ (checked 2026-10-26) ### Gemini CLI Canonical: https://www.launchpadhq.io/platform/gemini-cli Data: https://www.launchpadhq.io/data/v1/tools/gemini-cli.json Category: AI Coding & Development Company: Google Website: https://github.com/google-gemini/gemini-cli Cost tier: FREE Difficulty: ADVANCED Lifecycle: active Primary use: Open-source terminal AI coding agent Tasks: coding Free tier: yes (Free and open-source; uses free Gemini API tier (1,000 requests/day)) Starting price: Free Plans: - Free: Free (generous quota) Alternatives: Claude Code, OpenAI Codex CLI, Aider Guide: https://www.launchpadhq.io/platform/gemini-cli/guide Data as of: not dated Sources: - https://github.com/google-gemini/gemini-cli ### Genspark Canonical: https://www.launchpadhq.io/platform/genspark Data: https://www.launchpadhq.io/data/v1/tools/genspark.json Category: Workflow & Automation Company: Genspark Inc. Website: https://www.genspark.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: AI workspace / super-agent product. Official pricing is unverified (JavaScript-only page). Category is Workflow & Automation because it is sold as an always-on work surface rather than a single chat box. Primary use: AI workspace that mixes search, slides, and agents Tasks: slides, web-research Free tier: yes (Free path unverified) Starting price: See pricing Lowest paid plan: Plan prices unverified (official page JavaScript-only), See pricing Plans: - Plan prices unverified (official page JavaScript-only): See pricing Guide: https://www.launchpadhq.io/platform/genspark/guide Data as of: 2026-10-14 Sources: - https://www.genspark.ai/ (checked 2026-10-14) ### GitHub Copilot Canonical: https://www.launchpadhq.io/platform/github-copilot Data: https://www.launchpadhq.io/data/v1/tools/github-copilot.json Category: AI Coding & Development Company: GitHub (Microsoft) Website: https://github.com/features/copilot Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI code completion and agent mode in existing IDEs Tasks: coding Free tier: yes (Free plan: limited chat and agent usage; up to 2,000 completions plus entry-level models per month) Starting price: Free, from $10/mo Lowest paid plan: Pro, $10/mo Plans: - Free: $0 (Free: up to 2,000 completions/mo plus limited chat and agent usage; Pro, Pro+ and Max: unlimited completions, premium usage drawn from monthly AI Credits) - Pro: $10/mo - Pro+: $39/mo - Max $100/mo. Monthly AI Credit totals: Pro $15, Pro+ $70, Max $200 (1 credit = $0.01) Alternatives: Cursor, Claude Code, Devin Desktop, Amazon Q Developer Guide: https://www.launchpadhq.io/platform/github-copilot/guide Data as of: 2026-10-25 Sources: - https://github.com/features/copilot ### Glasp Canonical: https://www.launchpadhq.io/platform/glasp Data: https://www.launchpadhq.io/data/v1/tools/glasp.json Category: Research & Academic Tools Company: Glasp Website: https://glasp.co Cost tier: FREE Difficulty: BEGINNER Lifecycle: active Primary use: AI-powered web highlighting and note-taking Free tier: yes (Completely free; all features included; social highlighting platform) Starting price: Free Plans: - Free: Free (No limits; free platform (ad/premium model anticipated)) Guide: https://www.launchpadhq.io/platform/glasp/guide Data as of: not dated Sources: - https://glasp.co ### Gong Canonical: https://www.launchpadhq.io/platform/gong Data: https://www.launchpadhq.io/data/v1/tools/gong.json Category: Sales, Marketing & SEO AI Company: Gong Website: https://gong.io Cost tier: ENTERPRISE Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI revenue intelligence and conversation analytics Free tier: no (No free tier; enterprise pricing; demo available) Starting price: Custom Lowest paid plan: Enterprise, Custom Plans: - Enterprise: Custom Alternatives: Outreach Guide: https://www.launchpadhq.io/platform/gong/guide Data as of: not dated Sources: - https://gong.io ### Google Antigravity Canonical: https://www.launchpadhq.io/platform/google-antigravity Data: https://www.launchpadhq.io/data/v1/tools/google-antigravity.json Category: AI Coding & Development Company: Google Website: https://antigravity.google/ Cost tier: FREE Difficulty: INTERMEDIATE Lifecycle: active Primary use: Multi-agent AI coding IDE Free tier: yes (Free cloud-based IDE; AI features included; generous free tier) Starting price: Free Plans: - Free: Free (preview) Alternatives: Replit Agent, Cursor, Kiro Guide: https://www.launchpadhq.io/platform/google-antigravity/guide Data as of: 2026-10-20 Sources: - https://antigravity.google/ (checked 2026-10-20) ### Google Imagen 3 Canonical: https://www.launchpadhq.io/platform/google-imagen-3 Data: https://www.launchpadhq.io/data/v1/tools/google-imagen-3.json Category: Image Generation & Editing Company: Google DeepMind Website: https://deepmind.google/technologies/imagen-3/ Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Fast image generation with strong text rendering Tasks: image-generation Free tier: yes (Available via Gemini free tier (limited); via Vertex AI (pay-per-use)) Starting price: See pricing Lowest paid plan: Via Gemini, See pricing Plans: - Via Gemini: See pricing - Vertex AI API: See pricing Alternatives: Adobe Firefly, Midjourney, FLUX Guide: https://www.launchpadhq.io/platform/google-imagen-3/guide Data as of: not dated Sources: - https://deepmind.google/technologies/imagen-3/ ### Google Lyria 3 Canonical: https://www.launchpadhq.io/platform/google-lyria-3 Data: https://www.launchpadhq.io/data/v1/tools/google-lyria-3.json Category: Music Generation Company: Google DeepMind Website: https://deepmind.google Cost tier: FREE Difficulty: BEGINNER Lifecycle: active Primary use: AI music generation and real-time adaptive music Free tier: yes (Available via YouTube Music AI tools; limited via MusicFX (Google AI Test Kitchen)) Starting price: Free, from See pricing Lowest paid plan: API via Vertex AI, See pricing Plans: - Free: Free (MusicFX) - API via Vertex AI: See pricing Alternatives: Suno, Udio, AIVA, Soundraw Guide: https://www.launchpadhq.io/platform/google-lyria-3/guide Data as of: not dated Sources: - https://deepmind.google ### Google Veo 3.1 Canonical: https://www.launchpadhq.io/platform/google-veo-3-1 Data: https://www.launchpadhq.io/data/v1/tools/google-veo-3-1.json Category: Video Creation & Editing Company: Google DeepMind Website: https://deepmind.google/technologies/veo/ Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI video with native audio and Google integration Free tier: yes (Limited access via Gemini Advanced or VideoFX; pay-per-use via Vertex AI) Starting price: See pricing Lowest paid plan: Gemini subs, See pricing Plans: - Gemini subs: See pricing - Vertex AI API: See pricing Alternatives: Runway, Pika, Luma Dream Machine (Ray3) Guide: https://www.launchpadhq.io/platform/google-veo-3-1/guide Data as of: not dated Sources: - https://deepmind.google/technologies/veo/ ### Google Vertex AI Canonical: https://www.launchpadhq.io/platform/google-vertex-ai Data: https://www.launchpadhq.io/data/v1/tools/google-vertex-ai.json Category: AI APIs & Developer Services Company: Google Cloud Website: https://cloud.google.com/vertex-ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Enterprise ML/AI platform with Gemini models Free tier: yes ($300 Google Cloud free trial credits; Gemini API free tier via AI Studio (1,500 req/day)) Starting price: See pricing Lowest paid plan: Pay-per-use, See pricing Plans: - Pay-per-use: See pricing - various model pricing: See pricing Alternatives: Amazon Bedrock, Azure OpenAI Service, OpenAI API, Anthropic API Guide: https://www.launchpadhq.io/platform/google-vertex-ai/guide Data as of: not dated Sources: - https://cloud.google.com/vertex-ai ### GPT Image 1.5 (DALL-E successor) Canonical: https://www.launchpadhq.io/platform/gpt-image-1-5-dall-e-successor Data: https://www.launchpadhq.io/data/v1/tools/gpt-image-1-5-dall-e-successor.json Category: Image Generation & Editing Company: OpenAI Website: https://chat.openai.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Text-to-image with strong prompt adherence and text rendering Tasks: image-generation Free tier: yes (3 images/day on ChatGPT free; included in ChatGPT Plus/Pro) Starting price: $20/mo Lowest paid plan: Included with ChatGPT Plus ($20/mo), $20/mo Plans: - Included with ChatGPT Plus ($20/mo): $20/mo - API: See pricing Alternatives: Midjourney, Adobe Firefly, Ideogram, Leonardo AI Guide: https://www.launchpadhq.io/platform/gpt-image-1-5-dall-e-successor/guide Data as of: not dated Sources: - https://chat.openai.com ### GPT4All Canonical: https://www.launchpadhq.io/platform/gpt4all Data: https://www.launchpadhq.io/data/v1/tools/gpt4all.json Category: Local & Open-Source AI Company: Nomic AI Website: https://gpt4all.io Cost tier: FREE Difficulty: INTERMEDIATE Lifecycle: active Primary use: Simplest desktop app for local LLMs Free tier: yes (Completely free and open-source; desktop app for anyone) Starting price: Free Plans: - Free: Free (None (hardware-limited only)) Alternatives: Ollama, LM Studio, Jan Guide: https://www.launchpadhq.io/platform/gpt4all/guide Data as of: not dated Sources: - https://gpt4all.io ### Grammarly Canonical: https://www.launchpadhq.io/platform/grammarly Data: https://www.launchpadhq.io/data/v1/tools/grammarly.json Category: Research & Academic Tools Company: Superhuman Platform Website: https://grammarly.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI writing assistant for grammar, tone, and clarity Tasks: grammar-check, writing, cover-letter, email-writing, academic-writing Free tier: yes (Write without spelling and grammar mistakes, see your writing tone, and generate text with 100 AI prompts a month.) Starting price: Free, from $12 Lowest paid plan: Pro, $12/member/month Plans: - Free: $0/month (100 AI prompts a month) - Pro: $12/member/month (Billed annually; $30 when billed monthly. 2,000 AI prompts a month; team accounts up to 149 people) - Enterprise: Contact Sales (Available through Superhuman Go; unlimited AI prompts per member a month) Features: - Grammarly's FAQ says Pro includes all of Premium's advanced writing features, such as tone suggestions and full-sentence rewrites, plus team features that were previously only in the Business plan (as of 2026-10-26). - Switching from Premium to Pro cancels the Premium subscription with a refund prorated to the remainder of it, per the same FAQ. Guide: https://www.launchpadhq.io/platform/grammarly/guide Data as of: 2026-10-26 Sources: - https://www.grammarly.com/plans (checked 2026-10-26) - https://www.grammarly.com/pro (checked 2026-10-26) ### Granola Canonical: https://www.launchpadhq.io/platform/granola Data: https://www.launchpadhq.io/data/v1/tools/granola.json Category: Meetings & Notes Company: Granola, Inc. Website: https://www.granola.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Captures device audio; no bot joins the call, so other attendees may not know they are being recorded. Recording consent is still required — tell participants and follow all-party consent laws where they apply. Official pricing (granola.ai/pricing): Basic $0 with limited meeting history; Business $14/user/mo; Enterprise $35/user/mo. Model-training opt-out is offered. Primary use: Bot-free meeting notes from device audio Tasks: meeting-notes Free tier: yes (Basic, $0 per user per month, with limited meeting history.) Starting price: Free, from $14 Lowest paid plan: Business, $14 Plans: - Basic: $0 (per user per month; limited meeting history) - Business: $14 (per user per month) - Enterprise: $35 (per user per month) Guide: https://www.launchpadhq.io/platform/granola/guide Data as of: 2026-10-14 Sources: - https://www.granola.ai/pricing (checked 2026-10-14) - https://www.granola.ai/ (checked 2026-10-14) ### Grok Canonical: https://www.launchpadhq.io/platform/grok Data: https://www.launchpadhq.io/data/v1/tools/grok.json Category: Text & Conversational AI Company: SpaceXAI Website: https://x.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Fact-checking, conversational AI, images/video on X platform Free tier: yes (limited, exact limit not published) Starting price: Free, from $8/month Lowest paid plan: Plus, $8/month Plans: - Free: $0/month (limited, exact limit not published) - Plus: $8/month ($80/yr, base usage) - Premium: $30/month ($300/yr, 4x) - Super: $100/month ($1,000/yr, 15x) - Ultra: $200/month ($2,000/yr, 50x) Alternatives: ChatGPT, Claude, Gemini, Perplexity AI Guide: https://www.launchpadhq.io/platform/grok/guide Data as of: 2026-10-14 Sources: - https://x.ai/grok (checked 2026-10-14) - https://docs.x.ai/x-subscription/pricing (checked 2026-10-14) - https://docs.x.ai/developers/models (checked 2026-10-14) ### Grok Bot Canonical: https://www.launchpadhq.io/platform/grok-bot Data: https://www.launchpadhq.io/data/v1/tools/grok-bot.json Category: Workflow & Automation Company: SpaceXSI Website: https://grokbot.com Cost tier: PAID Difficulty: BEGINNER Lifecycle: active Description: xAI's Grok Bot, a separate product from the Grok chat app, with desktop and mobile apps. Primary use: Always-on AI teammates with their own cloud computer Free tier: no (None (needs X Subscription Premium or higher)) Starting price: $30/mo ($300/yr), Super $100/mo ($1,000/yr), Ultra $200/mo ($2,000/yr) and matching team seats. Not on Free or Plus. Existing SuperGrok and Cursor Pro/Pro+/Ultra/Teams keep access, per x.ai at time of writing. Weekly pool size and Additional Usage unit rate unverified. Lowest paid plan: No standalone price. The X Subscription (docs.x.ai/x-subscription/pricing): included on Premium, $30/mo ($300/yr), Super $100/mo ($1,000/yr), Ultra $200/mo ($2,000/yr) and matching team seats. Not on Free or Plus. Existing SuperGrok and Cursor Pro/Pro+/Ultra/Teams keep access, per x.ai at time of writing. Weekly pool size and Additional Usage unit rate unverified. Plans: - No standalone price. The X Subscription (docs.x.ai/x-subscription/pricing): included on Premium: $30/mo ($300/yr), Super $100/mo ($1,000/yr), Ultra $200/mo ($2,000/yr) and matching team seats. Not on Free or Plus. Existing SuperGrok and Cursor Pro/Pro+/Ultra/Teams keep access, per x.ai at time of writing. Weekly pool size and Additional Usage unit rate unverified. Data as of: 2026-10-14 Sources: - https://x.ai/bot (checked 2026-10-14) - https://docs.x.ai/grok-bot (checked 2026-10-14) - https://x.ai/bot#download (checked 2026-10-14) ### Groq Canonical: https://www.launchpadhq.io/platform/groq Data: https://www.launchpadhq.io/data/v1/tools/groq.json Category: AI APIs & Developer Services Company: Groq Website: https://groq.com Cost tier: FREE Difficulty: BEGINNER Lifecycle: active Primary use: Ultra-fast LLM inference on custom hardware Free tier: yes (Free: generous free tier with rate limits; multiple models available; no credit card required) Starting price: Free, from See pricing Lowest paid plan: pay-per-token, See pricing Plans: - Free: Free (Free: 30 req/min (varies by model); paid: higher limits; Developer: $0; Enterprise: custom) - pay-per-token: See pricing Alternatives: Together AI, Fireworks AI, OpenAI API, Anthropic API Guide: https://www.launchpadhq.io/platform/groq/guide Data as of: not dated Sources: - https://groq.com ### Gumloop Canonical: https://www.launchpadhq.io/platform/gumloop Data: https://www.launchpadhq.io/data/v1/tools/gumloop.json Category: Workflow & Automation Company: Gumloop Website: https://gumloop.com Cost tier: FREE Difficulty: INTERMEDIATE Lifecycle: active Primary use: No-code AI workflow builder with visual editor Free tier: yes (Free: 2,000 credits; 1 trigger; 2 concurrent workflows; community support only) Starting price: Free, from See pricing Lowest paid plan: paid plans, See pricing Plans: - Free: Free (plan) - paid plans: See pricing Alternatives: Zapier, Make (Integromat), n8n, Relevance AI Guide: https://www.launchpadhq.io/platform/gumloop/guide Data as of: not dated Sources: - https://gumloop.com ### Harvey AI Canonical: https://www.launchpadhq.io/platform/harvey-ai Data: https://www.launchpadhq.io/data/v1/tools/harvey-ai.json Category: Legal AI Company: Harvey AI Website: https://harvey.ai Cost tier: ENTERPRISE Difficulty: ADVANCED Lifecycle: active Primary use: AI for legal research, drafting, and analysis Free tier: no (No free tier; enterprise contracts only) Starting price: Custom Lowest paid plan: Enterprise, Custom Plans: - Enterprise: Custom Alternatives: CoCounsel, Spellbook, Luminance Guide: https://www.launchpadhq.io/platform/harvey-ai/guide Data as of: not dated Sources: - https://harvey.ai ### Healthee Canonical: https://www.launchpadhq.io/platform/healthee Data: https://www.launchpadhq.io/data/v1/tools/healthee.json Category: Healthcare AI Company: Healthee Website: https://healthee.co Cost tier: ENTERPRISE Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI healthcare benefits navigation Free tier: no (No individual free tier; employer-provided benefit) Starting price: Custom Lowest paid plan: Enterprise, Custom Plans: - Enterprise: Custom Guide: https://www.launchpadhq.io/platform/healthee/guide Data as of: not dated Sources: - https://healthee.co ### Hedra Canonical: https://www.launchpadhq.io/platform/hedra Data: https://www.launchpadhq.io/data/v1/tools/hedra.json Category: Video Creation & Editing Company: Hedra Website: https://www.hedra.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Avatar and talking-head video. Official pricing (hedra.com/pricing): Basic $20/mo (2,000 credits); Ultra $100/mo (18,000); Teams $75/mo (14,400); Enterprise custom. Page title states a free plan with no credit card; free-plan limits are unverified. You upload faces and voices — confirm consent and likeness rights for anyone depicted. Primary use: Talking-head and avatar video from face and voice uploads Free tier: yes (Free Plan, No Credit Card. Credit amount unpublished.) Starting price: $20/month Lowest paid plan: Basic, $20/month Plans: - Basic: $20/month (2,000 credits) - Pro: $50/month (7,200 credits) - Ultra: $100/month (18,000 credits) - Teams: $75/month (14,400 credits; per-seat unpublished) - Enterprise: Custom Guide: https://www.launchpadhq.io/platform/hedra/guide Data as of: 2026-10-17 Sources: - https://www.hedra.com/pricing (checked 2026-10-14) - https://www.hedra.com/docs/pages/app/billing/how-credits-work (checked 2026-10-17) ### HeyGen Canonical: https://www.launchpadhq.io/platform/heygen Data: https://www.launchpadhq.io/data/v1/tools/heygen.json Category: Video Creation & Editing Company: HeyGen Website: https://heygen.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI avatar videos and video translation Free tier: yes (Free $0: 3 videos/mo up to 1 min, watermarked, Avatar IV access, 500+ stock avatars, 1 custom video avatar) Starting price: Free, from $29/mo Lowest paid plan: Creator, $29/mo Plans: - Free: $0 (Free: 3 videos/mo, up to 1 min each, watermarked; Creator: 600 credits/mo, 1080p, up to 30 min/video; Pro: from 1,000 credits/mo, 4K, up to 30 min/video; Business: 1,500 credits/mo, up to 60 min/video; Enterprise: custom) - Creator: $29/mo (or $24/mo billed annually) - Pro from: $49/mo - Business: $149/mo + $20/seat - Enterprise contact sales: See pricing Alternatives: Synthesia, InVideo AI, Pictory AI Guide: https://www.launchpadhq.io/platform/heygen/guide Data as of: 2026-10-25 Sources: - https://heygen.com ### Higgsfield Canonical: https://www.launchpadhq.io/platform/higgsfield Data: https://www.launchpadhq.io/data/v1/tools/higgsfield.json Category: Video Creation & Editing Company: Higgsfield, Inc. Website: https://higgsfield.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: AI video generator. Official pricing is unverified (JavaScript-only). Do not invent credit packs or monthly fees. Primary use: AI video generation Free tier: yes (Limited use) Starting price: $19/month Lowest paid plan: Starter, $19/month Plans: - Starter: $19/month (Annual; page says no difference vs monthly) - Plus: $47/month (Annual; from $59 monthly) - Ultra: $99/month (Annual; from $129 monthly) - Teams: From $65/seat/month Guide: https://www.launchpadhq.io/platform/higgsfield/guide Data as of: 2026-10-17 Sources: - https://higgsfield.ai/ (checked 2026-10-14) - https://higgsfield.ai/pricing (checked 2026-10-14) - https://higgsfield.ai/creator-hub/help-center/credits-and-usage/how-credits-work (checked 2026-10-17) ### HubSpot AI Canonical: https://www.launchpadhq.io/platform/hubspot-ai Data: https://www.launchpadhq.io/data/v1/tools/hubspot-ai.json Category: AI Plugins for Business Software Company: HubSpot Website: https://hubspot.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI tools across CRM, marketing, and sales Free tier: yes (Free CRM + limited AI features (ChatSpot, content assistant); free for up to 5 users) Starting price: See pricing Lowest paid plan: Included with HubSpot plans, See pricing Plans: - Included with HubSpot plans: See pricing Guide: https://www.launchpadhq.io/platform/hubspot-ai/guide Data as of: not dated Sources: - https://hubspot.com ### Hugging Face Canonical: https://www.launchpadhq.io/platform/hugging-face Data: https://www.launchpadhq.io/data/v1/tools/hugging-face.json Category: Local & Open-Source AI Company: Hugging Face Website: https://huggingface.co Cost tier: FREEMIUM Difficulty: ADVANCED Lifecycle: active Primary use: AI model hub, hosting, and deployment Free tier: yes (Free: model hosting, Spaces (basic), datasets, Inference API (rate-limited), community features) Starting price: Free, from $9/mo Lowest paid plan: Pro, $9/mo Plans: - Free: Free (Free: rate-limited Inference API; Pro ($9/mo): faster inference; Enterprise: dedicated; Spaces: limited GPU free) - Pro: $9/mo - Enterprise: Custom Alternatives: Replicate Guide: https://www.launchpadhq.io/platform/hugging-face/guide Data as of: not dated Sources: - https://huggingface.co ### Ideogram Canonical: https://www.launchpadhq.io/platform/ideogram Data: https://www.launchpadhq.io/data/v1/tools/ideogram.json Category: Image Generation & Editing Company: Ideogram Inc. Website: https://ideogram.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI images with best text rendering Tasks: logo-design, image-generation, graphic-design Free tier: yes (30 credits a week for accounts that sign in with Google, Apple or Microsoft (email-and-password accounts get no free credits). Everything generated on the free plan is public.) Starting price: Free, from $20/month Lowest paid plan: Plus, $20/month Plans: - Free: $0/month (30 credits a week) - Plus: $20/month ($15/month billed yearly ($180 a year); 2,400 credits a month) - Pro: $60/month ($42/month billed yearly ($504 a year); 7,500 credits a month) - Team: $30/user/month ($20 per user/month billed yearly ($240 per user a year); 3,600 credits per user; at least 2 users) - Enterprise: Custom API pricing: Ideogram Vector API (ideogram-vector-1), USD per SVG. Text to vector: Low $0.13; Medium $0.16; High $0.20 from a prompt alone, $0.32 with reference images (Low and Medium cost the same with references). Image to vector (vectorize an image): $0.10 per image. Features: - Ideogram Vector makes scalable, editable SVG files, either from a text prompt or by converting an existing image (as of 2026-10-26). Alternatives: Midjourney, Adobe Firefly, Leonardo AI, Recraft Guide: https://www.launchpadhq.io/platform/ideogram/guide Data as of: 2026-10-26 Sources: - https://docs.ideogram.ai/plans-and-billing/available-plans (checked 2026-10-26) - https://ideogram.ai/models/vector/ (checked 2026-10-26) ### InVideo AI Canonical: https://www.launchpadhq.io/platform/invideo-ai Data: https://www.launchpadhq.io/data/v1/tools/invideo-ai.json Category: Video Creation & Editing Company: invideo Website: https://invideo.io Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Script-to-video automation for beginners Free tier: yes (Unlimited AI video generation; watermarked; 10 min max; iStock media) Starting price: $10 per seat/mo Lowest paid plan: Starter, $10 per seat/mo/billed annually Plans: - Starter: $10 per seat/mo/billed annually (400 credits per seat/mo. 50% off promo on annual plans (list $20/$60/$150)) - Plus: $30 per seat/mo/billed annually (2,000 credits. 50% off promo on annual plans (list $20/$60/$150)) - Max: $75 per seat/mo/billed annually (5,000 credits; tier slider up to 15,000. 50% off promo on annual plans (list $20/$60/$150)) Alternatives: Pictory AI, Synthesia, HeyGen, Descript Guide: https://www.launchpadhq.io/platform/invideo-ai/guide Data as of: 2026-10-14 Sources: - https://invideo.io/pricing (checked 2026-10-14) ### Inworld AI Canonical: https://www.launchpadhq.io/platform/inworld-ai Data: https://www.launchpadhq.io/data/v1/tools/inworld-ai.json Category: Gaming & Creative AI Company: Inworld AI Website: https://inworld.ai Cost tier: FREE Difficulty: BEGINNER Lifecycle: active Primary use: AI NPCs and characters for games Free tier: yes (On-Demand plan: $0 to start, pay as you go. Includes voice cloning and voice design with 2 professional voice clones (a payment method on file is needed before creating a professional clone).) Starting price: Free, from $25/month Lowest paid plan: Creator, $25/month Plans: - On-Demand: $0 (Pay as you go; 2 professional voice clones) - Creator: $25/month (4 professional voice clones) - Builder: $100/month (10 professional voice clones) - Developer: $300/month (25 professional voice clones) - Growth: $1,500/month (40 professional voice clones) - Enterprise: Custom (100+ professional voice clones) Guide: https://www.launchpadhq.io/platform/inworld-ai/guide Data as of: 2026-10-25 Sources: - https://inworld.ai/pricing (checked 2026-10-25) ### Jan Canonical: https://www.launchpadhq.io/platform/jan Data: https://www.launchpadhq.io/data/v1/tools/jan.json Category: Local & Open-Source AI Company: Open Source (Apache 2.0) Website: https://jan.ai Cost tier: FREE Difficulty: INTERMEDIATE Lifecycle: active Primary use: Privacy-first local AI platform Free tier: yes (Completely free and open-source (Apache 2.0)) Starting price: Free Plans: - Free: Free (None (hardware-limited only)) Alternatives: Ollama, LM Studio, GPT4All Guide: https://www.launchpadhq.io/platform/jan/guide Data as of: not dated Sources: - https://jan.ai ### Jasper Canonical: https://www.launchpadhq.io/platform/jasper Data: https://www.launchpadhq.io/data/v1/tools/jasper.json Category: Text & Conversational AI Company: Jasper AI, INC. Website: https://jasper.ai Cost tier: PAID Difficulty: BEGINNER Lifecycle: active Description: AI writing assistant focused on marketing copy, blog posts, and brand-voice content at scale. Primary use: AI marketing copy and content generation Tasks: marketing-copy Free tier: no (No free plan; 7-day free trial of Pro) Starting price: $59/month billed yearly or $69/month billed monthly Lowest paid plan: Pro, $59/month billed yearly or $69/month billed monthly Plans: - Pro: $59/month billed yearly or $69/month billed monthly - Business: custom pricing Alternatives: Copy.ai, Writesonic / Chatsonic, Notion AI, ChatGPT Guide: https://www.launchpadhq.io/platform/jasper/guide Data as of: 2026-10-14 Sources: - https://www.jasper.ai/pricing (checked 2026-10-14) ### Jenni AI Canonical: https://www.launchpadhq.io/platform/jenni-ai Data: https://www.launchpadhq.io/data/v1/tools/jenni-ai.json Category: Research & Academic Tools Company: Jenni, Inc Website: https://jenni.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Academic writing assistant. Official pricing (jenni.ai/pricing): Free $0 (10 AI autocompletes/day, 10 PDF uploads, 3 AI edits, 5 AI chat messages); Plus $12/mo; Pro $29/mo. Page states “annual saves 60%” — exact annual dollar prices were not listed as separate figures on the fetched table. Pricing in USD. Primary use: AI writing assistant for academic papers Tasks: academic-writing Free tier: yes ($0. 10 AI autocompletes per day, 10 PDF uploads, 3 AI edits, unlimited citations.) Starting price: Free, from $12 Lowest paid plan: Plus, $12 Plans: - Free: $0 (10 AI autocompletes per day, 10 PDF uploads, 3 AI edits, unlimited citations) - Plus: $12 (5,000 autocompletes/month; billing period unconfirmed) - Pro: $29 (Unlimited autocomplete; billing period unconfirmed) Data as of: 2026-10-14 Sources: - https://jenni.ai/pricing (checked 2026-10-14) ### JetBrains AI Canonical: https://www.launchpadhq.io/platform/jetbrains-ai Data: https://www.launchpadhq.io/data/v1/tools/jetbrains-ai.json Category: AI Coding & Development Company: JetBrains s.r.o. Website: https://www.jetbrains.com/ai/ Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Description: AI features inside JetBrains IDEs. Official pricing is unverified (JavaScript-only page). Do not quote a dollar amount until the official AI pricing table is readable. Primary use: AI assistant inside JetBrains IDEs Tasks: coding Free tier: yes (AI Free: unlimited completion, 3 credits/mo) Starting price: Free, from $100/user/yr Lowest paid plan: AI Pro, $100/user/yr Plans: - AI Free: Free (Unlimited completion, 3 credits/mo) - AI Pro: $100/user/yr - AI Ultimate: $300/user/yr Data as of: 2026-10-14 Sources: - https://www.jetbrains.com/air/ (checked 2026-10-14) - https://www.jetbrains.com/store/ai/ (checked 2026-10-14) ### Julius AI Canonical: https://www.launchpadhq.io/platform/julius-ai Data: https://www.launchpadhq.io/data/v1/tools/julius-ai.json Category: Research & Academic Tools Company: Julius AI Website: https://julius.ai Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI data analysis and visualization for researchers Free tier: yes (Free: limited messages; basic data analysis; chart generation) Starting price: Free, from $20/mo Lowest paid plan: Pro, $20/mo Plans: - Free: Free (Free: ~15 messages/mo; Essential: $20/mo; Pro: $45/mo with advanced features) - Pro: $20/mo Guide: https://www.launchpadhq.io/platform/julius-ai/guide Data as of: not dated Sources: - https://julius.ai ### Khan Academy (Khanmigo) Canonical: https://www.launchpadhq.io/platform/khan-academy-khanmigo Data: https://www.launchpadhq.io/data/v1/tools/khan-academy-khanmigo.json Category: Education & Learning AI Company: Khan Academy Website: https://khanacademy.org Cost tier: FREE Difficulty: BEGINNER Lifecycle: active Primary use: AI tutor for personalized education Free tier: yes (Khanmigo free for all learners; teachers need school district partnership or $5/mo premium) Starting price: Free, from See pricing Lowest paid plan: District licensing, See pricing Plans: - Free: Free (with Khanmigo) - District licensing: See pricing Guide: https://www.launchpadhq.io/platform/khan-academy-khanmigo/guide Data as of: not dated Sources: - https://khanacademy.org ### Kickresume Canonical: https://www.launchpadhq.io/platform/kickresume Data: https://www.launchpadhq.io/data/v1/tools/kickresume.json Category: Career & Job Search Company: Kickresume s.r.o. Website: https://www.kickresume.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Kickresume is a resume and cover letter builder with an AI writer that drafts a resume from a job title and job ad, an ATS resume checker and a personal-website builder. It is based in Slovakia and says it hosts data in the EU under GDPR. Per its privacy policy, AI features send data to providers such as OpenAI and Anthropic, and anonymised user documents may be used to train its systems, with a GDPR right to object. Primary use: AI resume and cover letter builder Tasks: resume, cover-letter, job-search Free tier: yes (Free while you use only free customization options: 4 free resume templates, unlimited downloads. Free for students verified through ISIC, ITIC or UNiDAYS.) Starting price: $8 Lowest paid plan: Yearly, $8/month, billed every 12 months Plans: - Yearly: $8/month, billed every 12 months (Includes vouchers for 5 friends) - Quarterly: $18/month, billed every 3 months (Includes vouchers for 3 friends) - Monthly: $24/month Features: - AI Writer that generates a resume from your job title and the job ad. - AI cover letter writer and 40+ cover letter templates matching the resume templates. - ATS Resume Checker that scores your resume for applicant tracking systems. - Career Map that suggests career paths from your resume. - LinkedIn and PDF import. - Personal website builder with 7 templates on paid plans. Alternatives: Teal, Rezi Data as of: 2026-10-26 Sources: - https://www.kickresume.com/terms/ (checked 2026-10-26) - https://www.kickresume.com/en/pricing/sale/ (checked 2026-10-26) ### Kimi Canonical: https://www.launchpadhq.io/platform/kimi Data: https://www.launchpadhq.io/data/v1/tools/kimi.json Category: International & Regional AI Company: Moonshot AI Website: https://www.kimi.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Moonshot AI's Kimi assistant (K3.1 Pro is the newest model). China-based company. Membership : Plus $15/mo, Pro $31/mo, Max $79/mo, Ultra $159/mo, billed annually; the Free plan includes limited K3.1 Pro use. Pay-as-you-go API at platform.kimi.ai. Review data-handling before sending sensitive inputs. Primary use: Long-document and agentic chat Free tier: yes (Free plan: limited use of K3.1 Pro on the web and in Kimi Work, basic agent credits, 1 agent task at a time, 2 scheduled tasks, 2 projects with 500MB storage. No K3.1 Pro Highspeed or Kimi Code.) Starting price: Free, from $15 Lowest paid plan: Plus, $15/month, billed annually Plans: - Free: $0 (Limited K3.1 Pro use; basic agent credits) - Plus: $15/month, billed annually (Monthly-billing price not shown) - Pro: $31/month, billed annually (Monthly-billing price not shown) - Max: $79/month, billed annually (Monthly-billing price not shown) - Ultra: $159/month, billed annually (Monthly-billing price not shown) API pricing: Pay-as-you-go, USD per 1M tokens, input / output. kimi-k3.1-pro: $2 / $10, cached input $0.20. kimi-k3.1-pro-highspeed: $4 / $20, cached input $0.20. kimi-k3: $3 / $15, cached input $0.30. kimi-k2.7-code: $0.95 / $4, cached input $0.19. kimi-k2.7-code-highspeed: $1.90 / $8, cached input $0.38. kimi-k2.6: $0.95 / $4, cached input $0.16. K3 series context 1,048,576 tokens; K2 series 262,144 tokens. K2 series: closed to newly registered users from 2026-10-31 (existing users not affected); Kimi plans to retire all K2 models by the end of December 2026 (stated plan, as of 2026-10-26). Features: - K3.1 Pro and K3.1 Pro Highspeed are available on Kimi.ai and the Kimi API (as of 2026-10-26). Kimi says Highspeed runs 2-3x faster than K3.1 Pro. - K3.1 Pro has a 1-million-token context window. - Kimi says the full K3.1 Pro model weights will be released by 2026-11-04 (stated plan, as of 2026-10-26). - Kimi Enterprise: K3.1 Pro is off by default at launch; administrators can enable it. Guide: https://www.launchpadhq.io/platform/kimi/guide Data as of: 2026-10-26 Sources: - https://www.kimi.com/ (checked 2026-10-14) - https://www.kimi.ai/resources/kimi-k3-1-pro-pricing (checked 2026-10-26) - https://platform.kimi.ai/docs/pricing/chat (checked 2026-10-26) - https://www.kimi.ai/blog/kimi-k3-1-pro (checked 2026-10-26) ### Kira Canonical: https://www.launchpadhq.io/platform/kira-systems Data: https://www.launchpadhq.io/data/v1/tools/kira-systems.json Category: Legal AI Company: Litera Website: https://www.litera.com/capabilities/review Cost tier: ENTERPRISE Difficulty: INTERMEDIATE Lifecycle: acquired Status: Kira is now sold by Litera. kirasystems.com redirects to Litera's review page, where the product is still called Kira. Primary use: AI contract analysis and due diligence Free tier: no (No free tier; enterprise contracts only) Starting price: Custom Lowest paid plan: Enterprise, Custom Plans: - Enterprise: Custom Alternatives: Luminance, Harvey AI, Spellbook Guide: https://www.launchpadhq.io/platform/kira-systems/guide Data as of: 2026-10-20 Sources: - https://www.litera.com/capabilities/review (checked 2026-10-20) ### Kiro Canonical: https://www.launchpadhq.io/platform/kiro Data: https://www.launchpadhq.io/data/v1/tools/kiro.json Category: AI Coding & Development Company: Amazon (AWS) Website: https://kiro.dev Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Primary use: Spec-driven agentic coding IDE Free tier: yes (Perpetual Free tier: 50 credits/mo (do not roll over); rate-limited models. Not “free during preview.”) Starting price: Free, from $20/user/mo Lowest paid plan: Pro, $20/user/mo Plans: - Free: $0 (Beta: generous limits; expected to have tiered pricing at GA) - Pro: $20/user/mo - Pro+: $40 - Pro Max: $100 - Power: $200 per user/mo - Enterprise via AWS: See pricing Alternatives: Cursor, Amazon Q Developer, GitHub Copilot, Devin Desktop Guide: https://www.launchpadhq.io/platform/kiro/guide Data as of: 2026-10-14 Sources: - https://kiro.dev ### Kling AI Canonical: https://www.launchpadhq.io/platform/kling-ai Data: https://www.launchpadhq.io/data/v1/tools/kling-ai.json Category: Video Creation & Editing Company: kling.ai Website: https://klingai.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Kuaishou's Kling video and image studio. China-based company; hosted generations are processed by that service — review before uploading faces, brands, or confidential footage. Official international membership (kling.ai/llms.txt citing re-checked 2026-10-14): Basic $0 (no monthly credits); Standard $8.80/mo after $6.99 first month (660 credits); Pro $32.56/mo (3,000 credits); Premier $80.96/mo (8,000); Ultra $159.99/mo (26,000). Official credits policy: $1 USD = 66 credits. Official blog states Basic-generated content is not for commercial use and carries a watermark. Primary use: AI video and image generation Free tier: yes (Basic free: no monthly credits; watermark; not for commercial use) Starting price: Free, from $1 Lowest paid plan: $1 = 66 credits, $1 Plans: - Basic: $0 - Standard: $8.80/mo ($6.99 first month, 660 credits) - Pro: $32.56/mo - Premier: $80.96/mo - Ultra: $159.99/mo - $1 = 66 credits: $1 Guide: https://www.launchpadhq.io/platform/kling-ai/guide Data as of: 2026-10-17 Sources: - https://kling.ai/ (checked 2026-10-14) - https://kling.ai/docs/payment-policy (checked 2026-10-17) - https://kling.ai/app/membership/membership-plan, ### Krea Canonical: https://www.launchpadhq.io/platform/krea Data: https://www.launchpadhq.io/data/v1/tools/krea.json Category: Image Generation & Editing Company: Krea Website: https://www.krea.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Multi-model image and video studio with LoRA training and an API. Official pricing (krea.ai/pricing): free 100 units/day; paid list/discounted monthly prices shown as $9/$5, $35/$21, $105/$63, $200/$160. The billing basis behind the two numbers is unverified. Primary use: Many image and video models in one studio Tasks: image-generation Free tier: yes (100 units/day, $0/month.) Starting price: $9/month Lowest paid plan: Basic, $9/month Plans: - Basic: $9/month ($5.25/mo billed annually; 5,000 units/mo) - Pro: $35/month ($21/mo billed annually; 20,000 units/mo) - Max: $105/month ($63/mo billed annually; 60,000 units/mo) - Business: $200/month ($160/mo billed annually; 80k units/month) Guide: https://www.launchpadhq.io/platform/krea/guide Data as of: 2026-10-24 Sources: - https://www.krea.ai/pricing (checked 2026-10-14) ### Krisp Canonical: https://www.launchpadhq.io/platform/krisp Data: https://www.launchpadhq.io/data/v1/tools/krisp.json Category: Meetings & Notes Company: Krisp Technologies, Inc. Website: https://krisp.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Noise cancellation (on-device) with an optional AI notetaker. Noise cancellation does not require sending meeting audio to the cloud; the notetaker does capture meeting audio., some US states require all-party consent to record — tell attendees before using notes features and follow the law where participants are located. Official pricing (krisp.ai/pricing): Free trial, no card (exact trial limits unverified); Core $16/user/mo ($8 annual); Advanced $30 ($15 annual); call-center $15/agent/mo annual. Primary use: On-device noise cancellation plus an optional AI notetaker Tasks: meeting-notes Free tier: yes (Free Trial, $0, Free for 7 days, no credit card required. This is a 7-day trial, not a permanent free plan.) Starting price: $15 Lowest paid plan: AI Agent Core, $15/per agent/mo, billed annually Plans: - Core: $16/mo/user ($8/mo/user on the Annually toggle (Save 50%)) - Advanced: $30/mo/user ($15/mo/user on the Annually toggle) - AI Agent Core: $15/per agent/mo, billed annually (Call-center plan) Data as of: 2026-10-14 Sources: - https://krisp.ai/pricing/ (checked 2026-10-14) ### Legora Canonical: https://www.launchpadhq.io/platform/legora Data: https://www.launchpadhq.io/data/v1/tools/legora.json Category: Legal AI Company: Legora Website: https://legora.com Cost tier: ENTERPRISE Difficulty: INTERMEDIATE Lifecycle: active Description: Legal work platform (agentic OS for lawyers). Official site (legora.com): book a demo / sales-led. No published list price. Contact sales (enterprise, no published price). Treat outputs as assistance, not legal advice. Site lists SOC 2 Type II, ISO 27001, GDPR, HIPAA, and states zero AI training on your data. Primary use: Collaborative legal AI for firms and in-house teams Free tier: no (No public free plan; demo via sales) Starting price: No public pricing (demo only) Lowest paid plan: Demo, No public pricing (demo only) Plans: - Demo: No public pricing (demo only) (Book a demo) Data as of: 2026-10-14 Sources: - https://legora.com/ (checked 2026-10-14) ### Leonardo AI Canonical: https://www.launchpadhq.io/platform/leonardo-ai Data: https://www.launchpadhq.io/data/v1/tools/leonardo-ai.json Category: Image Generation & Editing Company: Leonardo AI (Canva) Website: https://leonardo.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI images with custom model fine-tuning Tasks: image-generation Free tier: yes (150 tokens/day; basic image generation; community models; limited features) Starting price: Free, from $28/mo Lowest paid plan: Pro, $28/mo Plans: - Free: Free (150 tokens/day) - Pro: $28/mo - Professional: $56/mo Alternatives: Midjourney, Adobe Firefly, Ideogram, FLUX Guide: https://www.launchpadhq.io/platform/leonardo-ai/guide Data as of: not dated Sources: - https://leonardo.ai ### Lexis+ with Protégé Canonical: https://www.launchpadhq.io/platform/lexis-plus-ai Data: https://www.launchpadhq.io/data/v1/tools/lexis-plus-ai.json Category: Legal AI Company: LexisNexis Website: https://www.lexisnexis.com/en-us/products/lexis-plus-ai.page Cost tier: ENTERPRISE Difficulty: INTERMEDIATE Lifecycle: renamed Status: LexisNexis renamed Lexis+ AI to Lexis+ with Protégé in February 2026. Description: Legal research AI from LexisNexis. Official product page : “Free trial available; pricing varies… contact for a quote.” No public list price. Typical use is enterprise legal data — treat outputs as research assistance, not legal advice, and follow your firm’s confidentiality rules. Primary use: Enterprise legal research AI (quote-based) Free tier: no (Free trial available; duration and limits contact-sales) Starting price: Free, from See pricing Lowest paid plan: Quote only, See pricing Plans: - Quote only: See pricing - Free: Free (trial available) Data as of: 2026-10-20 Sources: - https://www.lexisnexis.com/en-us/products/lexis-plus-ai.page (checked 2026-10-20) ### Lindy AI Canonical: https://www.launchpadhq.io/platform/lindy-ai Data: https://www.launchpadhq.io/data/v1/tools/lindy-ai.json Category: Workflow & Automation Company: Lindy AI Website: https://lindy.ai Cost tier: PAID Difficulty: INTERMEDIATE Lifecycle: active Primary use: Custom AI agent builder for business workflows Free tier: no (No ongoing free plan. $50 in free credits, last 7 days, no credit card) Starting price: from $29.99/mo per active user (3,000 credits each) Lowest paid plan: Team, from $29.99/mo per active user (3,000 credits each) Plans: - Team: from $29.99/mo per active user (3,000 credits each) (Every active user in the workspace is billed) - Enterprise: custom Alternatives: Gumloop, Relevance AI Guide: https://www.launchpadhq.io/platform/lindy-ai/guide Data as of: 2026-10-14 Sources: - https://www.lindy.ai/pricing (checked 2026-10-14) ### Litmaps Canonical: https://www.launchpadhq.io/platform/litmaps Data: https://www.launchpadhq.io/data/v1/tools/litmaps.json Category: Research & Academic Tools Company: Litmaps Website: https://litmaps.com Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Primary use: Visual literature mapping and discovery Tasks: research-papers Free tier: yes (Free: 1 seed map; limited articles; basic visualization) Starting price: Free, from $10/mo Lowest paid plan: Pro, $10/mo Plans: - Free: Free (Free: 1 map; Pro: unlimited maps + monitoring + team features) - Pro: $10/mo Alternatives: Connected Papers, Research Rabbit, Semantic Scholar Guide: https://www.launchpadhq.io/platform/litmaps/guide Data as of: not dated Sources: - https://litmaps.com ### llama.cpp Canonical: https://www.launchpadhq.io/platform/llama-cpp Data: https://www.launchpadhq.io/data/v1/tools/llama-cpp.json Category: Local & Open-Source AI Company: Open Source (Georgi Gerganov) Website: https://github.com/ggerganov/llama.cpp Cost tier: FREE Difficulty: ADVANCED Lifecycle: active Primary use: C/C++ LLM inference engine Free tier: yes (Completely free and open-source (MIT)) Starting price: Free Plans: - Free: Free (None (hardware-limited)) Alternatives: LocalAI Guide: https://www.launchpadhq.io/platform/llama-cpp/guide Data as of: not dated Sources: - https://github.com/ggerganov/llama.cpp ### LM Studio Canonical: https://www.launchpadhq.io/platform/lm-studio Data: https://www.launchpadhq.io/data/v1/tools/lm-studio.json Category: Local & Open-Source AI Company: Element Labs Website: https://lmstudio.ai Cost tier: FREE Difficulty: INTERMEDIATE Lifecycle: active Primary use: Desktop GUI for local LLMs Free tier: yes (Completely free for personal use; all features included) Starting price: Free Plans: - Free: Free (None (hardware-limited only); free for personal use; business license separate) Alternatives: Ollama, Jan, GPT4All Guide: https://www.launchpadhq.io/platform/lm-studio/guide Data as of: not dated Sources: - https://lmstudio.ai ### LocalAI Canonical: https://www.launchpadhq.io/platform/localai Data: https://www.launchpadhq.io/data/v1/tools/localai.json Category: Local & Open-Source AI Company: Open Source Website: https://localai.io Cost tier: FREE Difficulty: ADVANCED Lifecycle: active Primary use: OpenAI API-compatible local inference Free tier: yes (Completely free and open-source (MIT)) Starting price: Free Plans: - Free: Free (None (hardware-limited)) Alternatives: Ollama, vLLM Guide: https://www.launchpadhq.io/platform/localai/guide Data as of: not dated Sources: - https://localai.io ### Lovable Canonical: https://www.launchpadhq.io/platform/lovable Data: https://www.launchpadhq.io/data/v1/tools/lovable.json Category: AI Coding & Development Company: Lovable Website: https://lovable.dev Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI app builder for non-developers Tasks: coding Free tier: yes (Free $0, no card: 5 build credits per day (up to 30/month), 20 Cloud credits/month, up to 5 lovable.app domains) Starting price: Free, from $25/mo Lowest paid plan: Pro, $25/mo Plans: - Free: $0 (Free: 5 build credits/day (max 30/month); Pro: 100 monthly credits with rollover; Business: 100 monthly credits; Enterprise: volume-based) - Pro: $25/mo (100 monthly credits, rollover) - Business: $50/mo (100 monthly credits, SSO) - Enterprise volume-based. Yearly billing gives 2 months free: See pricing - students up to 50% off Pro: See pricing Alternatives: Bolt.new, v0, Replit Agent Guide: https://www.launchpadhq.io/platform/lovable/guide Data as of: 2026-10-25 Sources: - https://lovable.dev ### Ludo AI Canonical: https://www.launchpadhq.io/platform/ludo-ai Data: https://www.launchpadhq.io/data/v1/tools/ludo-ai.json Category: Gaming & Creative AI Company: Ludo AI Website: https://ludo.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI game design and concept generation Free tier: yes (Free: basic game concept generation; limited daily ideas; community access) Starting price: Free, from $16/mo Lowest paid plan: Pro, $16/mo Plans: - Free: Free (Free: limited daily generations; Pro: higher limits; Teams: collaboration) - Pro: $16/mo Guide: https://www.launchpadhq.io/platform/ludo-ai/guide Data as of: not dated Sources: - https://ludo.ai ### Luma Dream Machine (Ray3) Canonical: https://www.launchpadhq.io/platform/luma-dream-machine-ray3 Data: https://www.launchpadhq.io/data/v1/tools/luma-dream-machine-ray3.json Category: Video Creation & Editing Company: Luma AI Website: https://lumalabs.ai Cost tier: PAID Difficulty: BEGINNER Lifecycle: active Primary use: 3D-aware video with physics simulation Free tier: no (No free plan listed on lumalabs.ai/pricing; homepage still offers a try-free path whose allowance is unpublished) Starting price: $30/mo Lowest paid plan: Plus, $30/mo Plans: - Plus: $30/mo ($25/mo billed yearly, $300/yr) - Pro: $90/mo ($75/mo yearly, $900/yr) - Ultra: $300/mo ($250/mo yearly, $3,000/yr) - Team and Enterprise contact sales: See pricing Alternatives: Runway, Pika, Google Veo 3.1 Guide: https://www.launchpadhq.io/platform/luma-dream-machine-ray3/guide Data as of: 2026-10-14 Sources: - https://lumalabs.ai ### Luminance Canonical: https://www.launchpadhq.io/platform/luminance Data: https://www.launchpadhq.io/data/v1/tools/luminance.json Category: Legal AI Company: Luminance Website: https://luminance.com Cost tier: ENTERPRISE Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI document review and contract analysis Free tier: no (No free tier; enterprise/firm contracts only) Starting price: Custom Lowest paid plan: Enterprise, Custom Plans: - Enterprise: Custom Alternatives: Harvey AI, CoCounsel, Kira Guide: https://www.launchpadhq.io/platform/luminance/guide Data as of: not dated Sources: - https://luminance.com ### MagicSchool Canonical: https://www.launchpadhq.io/platform/magicschool Data: https://www.launchpadhq.io/data/v1/tools/magicschool.json Category: Education & Learning AI Company: Magic School, Inc. Website: https://www.magicschool.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Teacher toolkit. Tools can handle student work and student data —, check your district privacy policy and FERPA/COPPA obligations before pasting student names, work, or identifiers. Official pricing (magicschool.ai/pricing): Free forever for individual teachers; Plus $12.99/mo or $8.33/mo billed annually; school/district plans available. Primary use: AI toolkit for teachers (lesson plans, feedback, and similar) Free tier: yes ($0 Forever Free, for individual teachers exploring AI.) Starting price: Free, from $8.33 Lowest paid plan: Plus, $8.33/user/mo billed annually Plans: - Forever Free: $0 (For individual teachers exploring AI) - Plus: $8.33/user/mo billed annually ($99.96 yearly, or $12.99/mo billed monthly) - Enterprise: custom (For schools and districts) Guide: https://www.launchpadhq.io/platform/magicschool/guide Data as of: 2026-10-14 Sources: - https://www.magicschool.ai/pricing (checked 2026-10-14) ### Mailchimp AI Canonical: https://www.launchpadhq.io/platform/mailchimp-ai Data: https://www.launchpadhq.io/data/v1/tools/mailchimp-ai.json Category: AI Plugins for Business Software Company: Intuit (Mailchimp) Website: https://mailchimp.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI email marketing assistant Free tier: yes (Free Mailchimp plan includes basic AI features; 500 contacts; 1,000 emails/mo) Starting price: See pricing Lowest paid plan: Included with Mailchimp plans, See pricing Plans: - Included with Mailchimp plans: See pricing Guide: https://www.launchpadhq.io/platform/mailchimp-ai/guide Data as of: not dated Sources: - https://mailchimp.com ### Make (Integromat) Canonical: https://www.launchpadhq.io/platform/make-integromat Data: https://www.launchpadhq.io/data/v1/tools/make-integromat.json Category: Workflow & Automation Company: Celonis (Make) Website: https://make.com Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Primary use: Visual AI workflow automation Tasks: workflow-automation Free tier: yes (Free: 1,000 credits/mo (credits are Make's billing unit)) Starting price: Free, from See pricing Lowest paid plan: paid plans Core, Pro, Teams and Enterprise. Paid prices did not render on the official page, so none are quoted, See pricing Plans: - Free: $0 (1,000 credits/mo) - paid plans Core, Pro, Teams and Enterprise. Paid prices did not render on the official page, so none are quoted: See pricing Alternatives: Zapier, n8n, Activepieces, Pabbly Connect Guide: https://www.launchpadhq.io/platform/make-integromat/guide Data as of: 2026-10-25 Sources: - https://make.com ### Manus Canonical: https://www.launchpadhq.io/platform/manus Data: https://www.launchpadhq.io/data/v1/tools/manus.json Category: Workflow & Automation Company: Manus Website: https://manus.im Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Autonomous agent that acts in a cloud browser, so it can see pages and data you give it. Official help-center pricing : Free 300 credits/day (capped 1,500/mo), Manus 2.0 Lite only; Pro from $20/mo (4,000 credits) or from $40/mo (8,000, 7-day trial); Team from $20/seat; annual billing saves 17%. Originally a China-founded company; current ownership and data jurisdiction are unverified. Primary use: Autonomous cloud-browser agent for multi-step tasks Free tier: yes (300 credits/day, max 1,500/mo, Manus 2.0 Lite only) Starting price: Free, from $20/mo Lowest paid plan: Pro from, $20/mo Plans: - Free: Free (300 credits/day (max 1,500/mo)) - Pro from: $20/mo - Pro 8,000 credits from: $40/mo - Team from: $20/seat - annual −17%: See pricing Guide: https://www.launchpadhq.io/platform/manus/guide Data as of: 2026-10-13 Sources: - https://manus.im ### Mem Canonical: https://www.launchpadhq.io/platform/mem Data: https://www.launchpadhq.io/data/v1/tools/mem.json Category: Meetings & Notes Company: Mem Labs Website: https://mem.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI-powered self-organizing notes Free tier: yes (Free: 25 messages/mo, 25 new notes, task tracking on those notes, 15 open tasks/projects, 3 hours Voice Mode) Starting price: Free, from $9/mo Lowest paid plan: Plus, $9/mo Plans: - Free: $0 (Free: limited AI queries/month; Pro $10/mo: unlimited AI; team features) - Plus: $9/mo - Pro: $29/mo - Pro 2×: $49 - Pro 5×: $99 - Pro 10×: $149 - Pro 20×: $199 - Extra Usage +$50 on Pro 20×: $50 Alternatives: Notion AI Guide: https://www.launchpadhq.io/platform/mem/guide Data as of: 2026-10-14 Sources: - https://mem.ai ### Meshy Canonical: https://www.launchpadhq.io/platform/meshy Data: https://www.launchpadhq.io/data/v1/tools/meshy.json Category: Gaming & Creative AI Company: Meshy Website: https://meshy.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI 3D model generation (text/image to 3D) Free tier: yes (Free: 5 credits/day; basic 3D model generation from text/image) Starting price: Free, from $20/mo Lowest paid plan: Pro, $20/mo Plans: - Free: Free (Free: 5 credits/day (few models); Pro: 200 credits/mo; Max: 500; Unlimited available) - Pro: $20/mo Guide: https://www.launchpadhq.io/platform/meshy/guide Data as of: not dated Sources: - https://meshy.ai ### Meta AI Canonical: https://www.launchpadhq.io/platform/meta-ai Data: https://www.launchpadhq.io/data/v1/tools/meta-ai.json Category: Text & Conversational AI Company: Meta Website: https://www.meta.ai Cost tier: FREE Difficulty: BEGINNER Lifecycle: active Description: Meta’s consumer assistant at meta.ai., published paid pricing did not render on the official page. Data handling is tied to Meta accounts — review Meta’s privacy policy before pasting anything you would not share with a Meta product. Usual cloud processing. Primary use: Free chatbot and image helper tied to a Meta account Free tier: yes (Meta AI is free to use for everyday use. Meta is testing usage limits, and a subscription is available for more.) Starting price: Free, from See pricing Lowest paid plan: paid-plan prices unverified (official page did not list them), See pricing Plans: - Free: Free (access advertised at meta.ai) - paid-plan prices unverified (official page did not list them): See pricing Guide: https://www.launchpadhq.io/platform/meta-ai/guide Data as of: 2026-10-17 Sources: - https://ai.meta.com/meta-ai/assistant/ (checked 2026-10-14) - https://research.meta.ai/blog/muse-spark-open/ (checked 2026-10-17) ### Microsoft Copilot Canonical: https://www.launchpadhq.io/platform/microsoft-copilot Data: https://www.launchpadhq.io/data/v1/tools/microsoft-copilot.json Category: Text & Conversational AI Company: Microsoft Website: https://copilot.microsoft.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI assistant in Microsoft 365, search, and Windows Tasks: writing, cover-letter, resume, email-writing, meeting-notes Free tier: yes (Copilot has a free tier: the paid Microsoft 365 plans are described as giving higher Copilot usage than free. The page does not list the free limits.) Starting price: $9.99/month Lowest paid plan: Microsoft 365 Personal, $9.99/month Plans: - Microsoft 365 Personal: $9.99/month (Or $99.99/year; 1 person; higher Copilot usage than free) - Microsoft 365 Family: $12.99/month (Or $129.99/year; 1 to 6 people, AI for the subscription owner only) - Microsoft 365 Premium: $19.99/month (Or $199.99/year; extensive Copilot usage) - Microsoft 365 Pro: $99.99/month (Monthly billing only; highest Copilot usage; 5 TB storage) Features: - Microsoft 365 Premium includes extensive Copilot usage limits for advanced features, image creation and chat, and Microsoft 365 Pro has the highest usage limits. - Microsoft 365 Personal and the plans above it include the Word, Excel, PowerPoint and Outlook desktop apps with Copilot. - The page lists Researcher (find and synthesize information) and Analyst (analyze data) among Copilot's features. Alternatives: ChatGPT, Claude, Gemini, Grok Guide: https://www.launchpadhq.io/platform/microsoft-copilot/guide Data as of: 2026-10-26 Sources: - https://www.microsoft.com/en-us/copilot/solutions/individuals (checked 2026-10-26) ### Microsoft Designer Canonical: https://www.launchpadhq.io/platform/microsoft-designer Data: https://www.launchpadhq.io/data/v1/tools/microsoft-designer.json Category: Image Generation & Editing Company: Microsoft Website: https://designer.microsoft.com Cost tier: FREE Difficulty: BEGINNER Lifecycle: active Primary use: AI design tool with image generation Tasks: graphic-design, image-generation Free tier: yes (Full access free via Microsoft account; image generation; design templates; social media posts) Starting price: Free Plans: - Free: Free (with Microsoft account) Alternatives: Adobe Firefly, Ideogram Guide: https://www.launchpadhq.io/platform/microsoft-designer/guide Data as of: not dated Sources: - https://designer.microsoft.com ### Microsoft Power Automate Canonical: https://www.launchpadhq.io/platform/microsoft-power-automate Data: https://www.launchpadhq.io/data/v1/tools/microsoft-power-automate.json Category: Workflow & Automation Company: Microsoft Website: https://powerautomate.microsoft.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Workflow automation within Microsoft 365 ecosystem Tasks: workflow-automation Free tier: yes (Included in M365 licenses; limited free flows; basic connectors) Starting price: $15/user/mo Lowest paid plan: From, $15/user/mo Plans: - From: $15/user/mo - included in some M365 plans: See pricing Alternatives: Zapier, Make (Integromat), n8n Guide: https://www.launchpadhq.io/platform/microsoft-power-automate/guide Data as of: not dated Sources: - https://powerautomate.microsoft.com ### Midjourney Canonical: https://www.launchpadhq.io/platform/midjourney Data: https://www.launchpadhq.io/data/v1/tools/midjourney.json Category: Image Generation & Editing Company: Midjourney Inc. Website: https://midjourney.com Cost tier: ENTERPRISE Difficulty: BEGINNER Lifecycle: active Primary use: Artistic and photorealistic AI image generation Tasks: image-generation Free tier: no (No free tier (was briefly available; currently requires paid plan)) Starting price: $10/mo Lowest paid plan: Basic, $10/mo Plans: - Basic: $10/mo - Standard: $30/mo - Pro: $60/mo - Mega: $120/mo Alternatives: Adobe Firefly, Ideogram, Leonardo AI, FLUX Guide: https://www.launchpadhq.io/platform/midjourney/guide Data as of: not dated Sources: - https://midjourney.com ### MiniMax Canonical: https://www.launchpadhq.io/platform/minimax Data: https://www.launchpadhq.io/data/v1/tools/minimax.json Category: AI APIs & Developer Services Company: MiniMax Website: https://www.minimax.io Cost tier: PAID Difficulty: INTERMEDIATE Lifecycle: active Description: MiniMax's model platform. M3.1 Flash, released 2026-10-19 per MiniMax, costs $0.15 input / $0.60 output per 1M tokens on the MiniMax API. Its weights are on Hugging Face as open weights under the MiniMax Community License, a custom license rather than open source: commercial use needs a "Built with MiniMax M3.1-Flash" credit plus a notice to MiniMax, or MiniMax's written authorization for products over US$20M in yearly revenue. Per the API platform's privacy policy, the service is operated by Nanonoble Pte. Ltd. (Singapore) and personal data is stored in a data center in the United States; the policy does not say whether API inputs are used to train models. Primary use: Low-cost model API, plus open-weight models you can run on your own hardware Free tier: no (No free API tier listed with M3.1 Flash; the open weights download free under MiniMax's community license) Starting price: $0.15/$0.60 per 1M tokens in/out, $0.015 cached input Lowest paid plan: API M3.1 Flash, $0.15/$0.60 per 1M tokens in/out, $0.015 cached input Plans: - API M3.1 Flash: $0.15/$0.60 per 1M tokens in/out, $0.015 cached input - other models and the M Plan are priced on platform.minimax.io: See pricing API pricing: MiniMax M3.1 Flash, USD per 1M tokens, flat: $0.15 input, $0.60 output, $0.015 cached input. Features: - M3.1 Flash weights are downloadable from MiniMax's Hugging Face page (as of 2026-10-26) as open weights under the MiniMax Community License, a custom license, not an open-source one. - Non-commercial use is permitted. Commercial use requires showing "Built with MiniMax M3.1-Flash", plus a one-time notice to MiniMax, or, for products with more than US$20 million in yearly revenue, MiniMax's prior written authorization. Alternatives: DeepSeek, Qwen, Kimi, OpenRouter Data as of: 2026-10-26 Sources: - https://platform.minimax.io/protocol/privacy-policy (checked 2026-10-26) - https://www.minimax.io/blog/minimax-m31f (checked 2026-10-26) - https://huggingface.co/MiniMaxAI/MiniMax-M3.1-Flash/blob/main/LICENSE (checked 2026-10-26) ### Mistral Vibe Canonical: https://www.launchpadhq.io/platform/mistral-vibe Data: https://www.launchpadhq.io/data/v1/tools/mistral-vibe.json Category: Text & Conversational AI Company: Mistral AI Website: https://chat.mistral.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: renamed Status: Mistral renamed its Le Chat assistant to Vibe. Mistral says all conversations, settings and plans carry over. Primary use: Conversational AI, multilingual, coding Free tier: yes (Free plan with limited messages, web searches and coding sessions.) Starting price: Free, from $14.99/month Lowest paid plan: Pro, $14.99/month Plans: - Free: $0 (Limited messages and web searches) - Pro: $14.99/month (Excluding taxes; verified students $5.99/month for up to 12 months) - Team: $24.99/user/month (Excluding taxes) - Enterprise: Custom (Contact sales) Alternatives: ChatGPT, Claude, Gemini, Grok Guide: https://www.launchpadhq.io/platform/mistral-vibe/guide Data as of: 2026-10-25 Sources: - https://mistral.ai/products/le-chat (checked 2026-10-25) - https://mistral.ai/pricing (checked 2026-10-25) ### Monarch Canonical: https://www.launchpadhq.io/platform/monarch Data: https://www.launchpadhq.io/data/v1/tools/monarch.json Category: Finance & Real Estate AI Company: Monarch Money, Inc. Website: https://www.monarch.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Budgeting and net-worth tracker with AI features. it aggregates linked bank and investment accounts — review who in the household can see those connections. Official pricing (monarch.com/pricing): $14.99/mo or $99.99/yr. A trial is advertised; trial length is unverified. Primary use: Household budgeting and net-worth tracker with AI features Free tier: yes (Trial advertised; length unverified) Starting price: $14.99/month Lowest paid plan: Monthly subscription, $14.99/month Plans: - Yearly subscription: $99.99/year - Monthly subscription: $14.99/month Data as of: 2026-10-14 Sources: - https://www.monarch.com/pricing (checked 2026-10-14) ### Mubert Canonical: https://www.launchpadhq.io/platform/mubert Data: https://www.launchpadhq.io/data/v1/tools/mubert.json Category: Music Generation Company: Mubert Website: https://mubert.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Continuous AI background music streams Free tier: yes (Ambassador free: 25 generations/mo, 5 downloads/mo, non-commercial) Starting price: $14/mo Lowest paid plan: Creator, $14/mo Plans: - Ambassador Free: See pricing - Creator: $14/mo ($11.69 shown as discounted) - Pro: $39/mo commercial ($32.49 discounted) - Business: $199/mo ($149.29 discounted) Alternatives: Soundraw, AIVA, Beatoven.ai, Suno Guide: https://www.launchpadhq.io/platform/mubert/guide Data as of: 2026-10-14 Sources: - https://mubert.com ### Murf AI Canonical: https://www.launchpadhq.io/platform/murf-ai Data: https://www.launchpadhq.io/data/v1/tools/murf-ai.json Category: Voice & Speech Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: AI voice generator for voiceovers, e-learning, and ads with multiple voices and languages. Primary use: Text-to-speech, voice overs Free tier: yes (Limited free voice generation minutes) Starting price: Free, from $29/mo Lowest paid plan: $29/mo, $29/mo Plans: - Free: Free - $29/mo: $29/mo Guide: https://www.launchpadhq.io/platform/murf-ai/guide Data as of: not dated Sources: - https://www.launchpadhq.io/platform/murf-ai ### n8n Canonical: https://www.launchpadhq.io/platform/n8n Data: https://www.launchpadhq.io/data/v1/tools/n8n.json Category: Workflow & Automation Company: n8n GmbH Website: https://n8n.io Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Primary use: Open-source low-code AI workflow automation Tasks: workflow-automation Free tier: yes (Free only if you run it yourself: the self-hosted Community edition is free with almost the full feature set. n8n Cloud has no free plan, only a trial with no card required.) Starting price: Free, from €20/month Lowest paid plan: Starter (Cloud), €20/month Plans: - Community (self-hosted): Free (You host and maintain it yourself) - Starter (Cloud): €20/month (Billed annually; 2.5K workflow executions) - Pro (Cloud): €50/month (Billed annually; 10K workflow executions) - Business (self-hosted): €667/month (Billed annually; 40K workflow executions) - Enterprise: Custom (Hosted by n8n or self-hosted) Alternatives: Zapier, Make (Integromat), Activepieces, Microsoft Power Automate Guide: https://www.launchpadhq.io/platform/n8n/guide Data as of: 2026-10-25 Sources: - https://docs.n8n.io/hosting/community-edition-features/ (checked 2026-10-25) - https://n8n.io/pricing/ (checked 2026-10-25) ### Nanonets Canonical: https://www.launchpadhq.io/platform/nanonets Data: https://www.launchpadhq.io/data/v1/tools/nanonets.json Category: Document & PDF Processing Company: Nanonets Website: https://nanonets.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI-powered document processing and OCR Free tier: yes (Free: 500 pages/month; pre-trained models; basic features) Starting price: Free, from $499/mo Lowest paid plan: Pro, $499/mo Plans: - Free: Free (Free: 500 pages/mo; Pro: 5,000; Enterprise: custom; additional pages purchasable) - Pro: $499/mo Alternatives: Docsumo, Reducto, Upstage (Solar) Guide: https://www.launchpadhq.io/platform/nanonets/guide Data as of: not dated Sources: - https://nanonets.com ### Napkin AI Canonical: https://www.launchpadhq.io/platform/napkin-ai Data: https://www.launchpadhq.io/data/v1/tools/napkin-ai.json Category: Document & PDF Processing Company: Napkin AI Website: https://www.napkin.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Text-to-diagram visuals. Official pricing (napkin.ai/pricing): Free 500 AI credits/week with Napkin branding on visuals; Plus $9/person/mo; Pro $22/mo. Yearly billing saves 25%. Usual SaaS data-handling; no extra privacy flag beyond that. Primary use: Turn text into diagrams for docs and slides Free tier: yes (Free Forever, $0, 500 AI credits per week, with Napkin branding.) Starting price: Free, from $12 Lowest paid plan: Plus, $12/per person per month, billed monthly Plans: - Free Forever: $0 (500 AI credits per week, with Napkin branding) - Plus: $12/per person per month, billed monthly ($9 billed yearly; 10,000 credits/mo) - Pro: $30/per person per month, billed monthly ($22 billed yearly; 30,000 credits/mo) Guide: https://www.launchpadhq.io/platform/napkin-ai/guide Data as of: 2026-10-14 Sources: - https://www.napkin.ai/pricing/ (checked 2026-10-14) ### Naver (HyperCLOVA X / Cue:) Canonical: https://www.launchpadhq.io/platform/naver-hyperclova-x-cue Data: https://www.launchpadhq.io/data/v1/tools/naver-hyperclova-x-cue.json Category: International & Regional AI Company: Naver Website: https://clova.ai Cost tier: FREE Difficulty: BEGINNER Lifecycle: active Primary use: Korean AI assistant and search Free tier: yes (Free: Cue: chatbot for Korean users; integrated into Naver ecosystem) Starting price: Free, from See pricing Lowest paid plan: Enterprise via Naver Cloud, See pricing Plans: - Free: Free (Free for consumer use; API pricing for developers) - Enterprise via Naver Cloud: See pricing Alternatives: Gemini Guide: https://www.launchpadhq.io/platform/naver-hyperclova-x-cue/guide Data as of: not dated Sources: - https://clova.ai ### Gemini Notebook Canonical: https://www.launchpadhq.io/platform/notebooklm Data: https://www.launchpadhq.io/data/v1/tools/notebooklm.json Category: Text & Conversational AI Company: Google Website: https://notebook.google/ Cost tier: FREE Difficulty: BEGINNER Lifecycle: renamed Status: Google renamed NotebookLM to Gemini Notebook on July 16, 2026. Primary use: AI Q&A grounded in your uploaded documents Free tier: yes (Full access; upload up to 50 sources per notebook; Audio Overview (podcast generation); sharing) Starting price: Free Plans: - Free: Free (50 sources/notebook; 500K words/source; limited Audio Overview generations/day) Alternatives: ChatPDF, Anara Guide: https://www.launchpadhq.io/platform/notebooklm/guide Data as of: 2026-10-20 Sources: - https://blog.google/innovation-and-ai/products/gemini-notebook/notebooklm-gemini-notebook/ (checked 2026-10-20) - https://notebook.google/ (checked 2026-10-20) ### Notion AI Canonical: https://www.launchpadhq.io/platform/notion-ai Data: https://www.launchpadhq.io/data/v1/tools/notion-ai.json Category: Text & Conversational AI Company: Notion Labs, Inc. Website: https://www.notion.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI writing and Q&A within Notion workspace Tasks: writing, meeting-notes, summarization Free tier: yes (Free for individuals: a trial of Notion AI, unlimited pages and blocks for one person (limited for 2+ members), file uploads up to 5 MB, 7 days of page history and up to 10 guests.) Starting price: Free, from $10 Lowest paid plan: Plus, $10/seat/month Plans: - Free: $0/seat/month (Trial of Notion AI) - Plus: $10/seat/month (Trial of Notion AI) - Business: $20/seat/month (Notion Agent and AI Meeting Notes included; 1,500 Standard and 500 Select Notion credits per seat a month) - Enterprise: Custom pricing (Zero data retention with Notion's LLM providers) Features: - Notion AI runs on Notion credits: Free and Plus get a limited trial, and Business includes 1,500 Standard credits (Notion Agent on Auto, AI Meeting Notes, Database Autofill) and 500 Select credits (work where you pick the model, plus Custom Agents you own) per seat a month, per the plan comparison (as of 2026-10-26). - Admins can buy extra Notion credits to cover additional AI usage, per the pricing page (as of 2026-10-26). - Notion's pricing FAQ says Notion AI will not use your data to train its models unless you opt in, and that AI subprocessors are contractually barred from training on customer data. Alternatives: ChatGPT, Copy.ai, Jasper, Mem Guide: https://www.launchpadhq.io/platform/notion-ai/guide Data as of: 2026-10-26 Sources: - https://www.notion.com/pricing (checked 2026-10-26) ### Ollama Canonical: https://www.launchpadhq.io/platform/ollama Data: https://www.launchpadhq.io/data/v1/tools/ollama.json Category: Local & Open-Source AI Company: Ollama (Open Source) Website: https://ollama.com Cost tier: FREE Difficulty: INTERMEDIATE Lifecycle: active Primary use: CLI to run local LLMs (Docker-like) Free tier: yes (Completely free and open-source; all features included) Starting price: Free Plans: - Free: Free (None (hardware-limited only); runs on your machine) Alternatives: LM Studio, Jan, GPT4All, llama.cpp Guide: https://www.launchpadhq.io/platform/ollama/guide Data as of: not dated Sources: - https://ollama.com ### Open WebUI Canonical: https://www.launchpadhq.io/platform/open-webui Data: https://www.launchpadhq.io/data/v1/tools/open-webui.json Category: Local & Open-Source AI Company: Open Source Website: https://openwebui.com Cost tier: FREE Difficulty: ADVANCED Lifecycle: active Primary use: Web interface for local LLMs (pairs with Ollama) Free tier: yes (Completely free and open-source (MIT)) Starting price: Free Plans: - Free: Free (None (self-hosted, hardware-limited)) Alternatives: Text Generation WebUI (oobabooga) Guide: https://www.launchpadhq.io/platform/open-webui/guide Data as of: not dated Sources: - https://openwebui.com ### OpenAI API Canonical: https://www.launchpadhq.io/platform/openai-api Data: https://www.launchpadhq.io/data/v1/tools/openai-api.json Category: AI APIs & Developer Services Company: OpenAI Website: https://platform.openai.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: GPT models, embeddings, whisper, TTS via API Free tier: yes (Free: $5 initial credits (new accounts); rate-limited; GPT-4o-mini free tier) Starting price: $2.50/$10 per 1M tokens) Lowest paid plan: Pay-per-token (GPT-4o, $2.50/$10 per 1M tokens) Plans: - Pay-per-token (GPT-4o: $2.50/$10 per 1M tokens) API pricing: Decisions API (GPT-6 Luna), USD per 1M tokens: $0.10 input, $0.01 cached input. You only pay for input tokens; there are no output-token charges. Features: - Decisions API: generally available 2026-10-22, runs on GPT-6 Luna via its own Decisions endpoint. It answers questions with bounded answers, returning a probability, a choice or a score, with up to 400 questions per request. Regional processing is available in the US and EU. Alternatives: Anthropic API, Google Vertex AI, Amazon Bedrock, Azure OpenAI Service Guide: https://www.launchpadhq.io/platform/openai-api/guide Data as of: 2026-10-26 Sources: - https://openai.com/index/decisions-api (checked 2026-10-26) - https://developers.openai.com/api/docs/models (checked 2026-10-14) ### OpenAI Codex CLI Canonical: https://www.launchpadhq.io/platform/openai-codex-cli Data: https://www.launchpadhq.io/data/v1/tools/openai-codex-cli.json Category: AI Coding & Development Company: OpenAI Website: https://github.com/openai/codex Cost tier: FREE Difficulty: ADVANCED Lifecycle: active Primary use: Open-source terminal coding agent Tasks: coding Free tier: yes (Free and open-source; uses your own OpenAI API key) Starting price: Free Plans: - Free: Free (open source) Alternatives: Claude Code, Aider, Gemini CLI Guide: https://www.launchpadhq.io/platform/openai-codex-cli/guide Data as of: not dated Sources: - https://github.com/openai/codex ### OpenEvidence Canonical: https://www.launchpadhq.io/platform/openevidence Data: https://www.launchpadhq.io/data/v1/tools/openevidence.json Category: Healthcare AI Company: OpenEvidence Website: https://www.openevidence.com Cost tier: FREE Difficulty: INTERMEDIATE Lifecycle: active Description: Cited clinical Q&A, free for verified US healthcare professionals. Access is gated to licensed clinicians: it is not for patients or self-diagnosis and is not a consumer symptom checker. Its privacy policy (updated September 17, 2026) says it secures protected health information (PHI) in compliance with HIPAA and does not train AI models on PHI. Its security page says HIPAA covered entities may send PHI under its standard Business Associate Agreement. Availability outside the US was not confirmed on the official site. Primary use: Cited clinical answers for verified US clinicians Free tier: yes (Free for verified US healthcare professionals) Starting price: Free Plans: - Free: $0, no paid plans listed Guide: https://www.launchpadhq.io/platform/openevidence/guide Data as of: 2026-10-25 Sources: - https://www.openevidence.com ### OpenRouter Canonical: https://www.launchpadhq.io/platform/openrouter Data: https://www.launchpadhq.io/data/v1/tools/openrouter.json Category: AI APIs & Developer Services Company: OpenRouter, Inc Website: https://openrouter.ai Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Description: Multi-provider LLM router. Official pricing (openrouter.ai/pricing): 25+ free models, 50 requests/day; Standard pay-as-you-go plus 5.5% platform fee; Business 8% fee plus EU/US in-region routing; Enterprise custom; BYOK $25k/mo fee-free then 5%. Data-policy routing is paid-only. Whether free-model providers log prompts is unverified. Primary use: One API for many models, including free-tier models Free tier: yes (25+ free models; free-model rate limits are 50 requests/day, rising to 1,000/day once you've bought at least $10 of credits.) Starting price: No fee up to $25,000 of list-price inference per month, then 5% Lowest paid plan: BYOK, No fee up to $25,000 of list-price inference per month, then 5% Plans: - Standard: 5.5% fee on credit purchases (Never on individual requests) - Business: 8% fee on credit purchases (Never on individual requests) - BYOK: No fee up to $25,000 of list-price inference per month, then 5% - Enterprise: Custom Guide: https://www.launchpadhq.io/platform/openrouter/guide Data as of: 2026-10-14 Sources: - https://openrouter.ai/pricing (checked 2026-10-14) - https://openrouter.ai/models (checked 2026-10-14) ### Opus Clip Canonical: https://www.launchpadhq.io/platform/opus-clip Data: https://www.launchpadhq.io/data/v1/tools/opus-clip.json Category: Gaming & Creative AI Company: Opus Website: https://opus.pro Cost tier: FREEMIUM Difficulty: ADVANCED Lifecycle: active Primary use: AI long-video to short-clip repurposing Free tier: yes (Free: 60 minutes/month upload; basic AI clipping; watermarked) Starting price: Free, from $15/mo Lowest paid plan: Starter, $15/mo Plans: - Free: Free (Free: 60 min/mo; Starter: 200 min; Growth: 400 min; Business: 600 min) - Starter: $15/mo Alternatives: Descript, InVideo AI, Pictory AI Guide: https://www.launchpadhq.io/platform/opus-clip/guide Data as of: not dated Sources: - https://opus.pro ### Otter AI Canonical: https://www.launchpadhq.io/platform/otter-ai Data: https://www.launchpadhq.io/data/v1/tools/otter-ai.json Category: Meetings & Notes Company: Otter.ai Website: https://otter.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI meeting transcription and real-time notes Tasks: meeting-notes, transcription Free tier: yes (Basic (free): 300 transcription minutes/mo, 30 min per conversation, 3 lifetime file imports) Starting price: Free, from $16.99/user/mo billed monthly Lowest paid plan: Pro, $16.99/user/mo billed monthly Plans: - Basic: $0 - Pro: $16.99/user/mo billed monthly - Business: $30/user/mo billed monthly ($24 billed annually) Alternatives: tl;dv, Tactiq, Descript Guide: https://www.launchpadhq.io/platform/otter-ai/guide Data as of: 2026-10-25 Sources: - https://otter.ai ### Outreach Canonical: https://www.launchpadhq.io/platform/outreach Data: https://www.launchpadhq.io/data/v1/tools/outreach.json Category: Sales, Marketing & SEO AI Company: Outreach Website: https://outreach.io Cost tier: ENTERPRISE Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI-powered sales engagement platform Free tier: no (No free tier; enterprise pricing; demo available) Starting price: Custom Lowest paid plan: Enterprise, Custom Plans: - Enterprise: Custom Alternatives: Gong Guide: https://www.launchpadhq.io/platform/outreach/guide Data as of: not dated Sources: - https://outreach.io ### Pabbly Connect Canonical: https://www.launchpadhq.io/platform/pabbly-connect Data: https://www.launchpadhq.io/data/v1/tools/pabbly-connect.json Category: Workflow & Automation Company: Pabbly Website: https://pabbly.com/connect/ Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Budget workflow automation Free tier: yes (Free trial (100 tasks); then paid plans from $25/mo lifetime deals available) Starting price: $25/mo Lowest paid plan: From, $25/mo Plans: - From: $25/mo (lifetime deals available) Alternatives: Zapier, Make (Integromat), n8n, Activepieces Guide: https://www.launchpadhq.io/platform/pabbly-connect/guide Data as of: not dated Sources: - https://pabbly.com/connect/ ### Paperpal Canonical: https://www.launchpadhq.io/platform/paperpal Data: https://www.launchpadhq.io/data/v1/tools/paperpal.json Category: Research & Academic Tools Company: Cactus Communications Website: https://paperpal.com Cost tier: FREE Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI academic writing assistant Tasks: grammar-check, academic-writing Free tier: yes (Free: limited word checks; basic grammar/language; Chrome extension) Starting price: Free, from See pricing Lowest paid plan: Pro plans, See pricing Plans: - Free: Free (Free: limited checks/month; Prime: 50K words/mo; Pro: unlimited; Premium: unlimited + checks) - Pro plans: See pricing Alternatives: Grammarly, Trinka AI Guide: https://www.launchpadhq.io/platform/paperpal/guide Data as of: not dated Sources: - https://paperpal.com ### PathAI Canonical: https://www.launchpadhq.io/platform/pathai Data: https://www.launchpadhq.io/data/v1/tools/pathai.json Category: Healthcare AI Company: PathAI Website: https://pathai.com Cost tier: ENTERPRISE Difficulty: ADVANCED Lifecycle: active Primary use: AI-powered pathology and diagnostics Free tier: no (No free tier; enterprise/pharma contracts) Starting price: Custom Lowest paid plan: Enterprise, Custom Plans: - Enterprise: Custom Guide: https://www.launchpadhq.io/platform/pathai/guide Data as of: not dated Sources: - https://pathai.com ### Perplexity AI Canonical: https://www.launchpadhq.io/platform/perplexity-ai Data: https://www.launchpadhq.io/data/v1/tools/perplexity-ai.json Category: Text & Conversational AI Company: Perplexity AI Website: https://perplexity.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI-powered search engine with cited answers Tasks: web-research Free tier: yes (Practically unlimited basic searches, 3 Pro Searches a day, 1 Research query a month and limited basic file uploads; no advanced AI models or image generation.) Starting price: $20/month Lowest paid plan: Pro, $20/month Plans: - Pro: $20/month (Or $200/year) - Max: $200/month (Or $2,000/year; annual billing only on the perplexity.ai web app) - Enterprise Pro: $40/seat/month (Or $400/year per seat) - Enterprise Max: $325/seat/month (Or $3,250/year per seat) Features: - Pro and Max subscribers can opt out of data collection in settings; Enterprise Pro and Enterprise Max data is never logged or used for training, per Perplexity's plan comparison (as of 2026-10-26). - The plan comparison table lists Pro Searches at 3 a day on Free, weekly limits on Pro and Max, 400 a week on Enterprise Pro and 4,000 a week on Enterprise Max (as of 2026-10-26). Alternatives: ChatGPT, Gemini, Grok Guide: https://www.launchpadhq.io/platform/perplexity-ai/guide Data as of: 2026-10-26 Sources: - https://www.perplexity.ai/help-center/en/articles/11187416-which-perplexity-subscription-plan-is-right-for-you (checked 2026-10-26) - https://www.perplexity.ai/help-center/en/articles/12167980-using-the-connector-for-slack (checked 2026-10-26) ### Perplexity Comet Canonical: https://www.launchpadhq.io/platform/perplexity-comet Data: https://www.launchpadhq.io/data/v1/tools/perplexity-comet.json Category: Text & Conversational AI Company: Perplexity Website: https://www.perplexity.ai/comet Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Perplexity’s AI browser (Comet). Official pricing is unverified (official page blocked). Do not assume it is included in every Perplexity plan. Review Perplexity’s privacy policy — the browser can see pages you visit. Primary use: AI browser from Perplexity Free tier: yes (Will always provide a free version; includes Comet.) Starting price: $17/mo Lowest paid plan: Pro, $17/mo/billed annually Plans: - Pro: $17/mo/billed annually - Max: $167/mo/billed annually Guide: https://www.launchpadhq.io/platform/perplexity-comet/guide Data as of: 2026-10-24 Sources: - https://www.perplexity.ai/comet (checked 2026-10-14) - https://www.perplexity.ai/pro (checked 2026-10-14) - https://www.perplexity.ai/help-center/en/articles/10354919-what-advanced-ai-models-are-included-in-my-subscription (checked 2026-10-24) ### Photoroom Canonical: https://www.launchpadhq.io/platform/photoroom Data: https://www.launchpadhq.io/data/v1/tools/photoroom.json Category: Sales, Marketing & SEO AI Company: Photoroom Website: https://photoroom.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI background removal and product photography Free tier: yes (A free path is not priced on the paid-plan cards fetched 2026-10-14; do not quote the old $12.99/mo Pro figure.) Starting price: $7.50/mo billed yearly (8,000 AI credits, 1,000 exports). Max and Ultra are listed as yearly billed plans Lowest paid plan: Pro, $7.50/mo billed yearly (8,000 AI credits, 1,000 exports). Max and Ultra are listed as yearly billed plans Plans: - Pro: $7.50/mo billed yearly (8,000 AI credits, 1,000 exports). Max and Ultra are listed as yearly billed plans - their dollar amounts did not render on the fetched page — unverified. Enterprise custom. Monthly (non-yearly) prices unverified.: See pricing Alternatives: Canva AI (Magic Studio), Adobe Firefly Guide: https://www.launchpadhq.io/platform/photoroom/guide Data as of: 2026-10-14 Sources: - https://photoroom.com ### Pictory AI Canonical: https://www.launchpadhq.io/platform/pictory-ai Data: https://www.launchpadhq.io/data/v1/tools/pictory-ai.json Category: Video Creation & Editing Company: Pictory Website: https://pictory.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Convert blogs/long-form content to video Free tier: yes (No free plan; 14-day trial (3 video projects) on the pricing page) Starting price: $29/mo Lowest paid plan: Starter, $29/mo Plans: - Starter: $29/mo ($25 billed annually) - Professional: $59/mo ($35 annual) - Team: $199/mo ($119 annual) - Enterprise: Custom Alternatives: InVideo AI, Descript, Synthesia, HeyGen Guide: https://www.launchpadhq.io/platform/pictory-ai/guide Data as of: 2026-10-14 Sources: - https://pictory.ai ### Pika Canonical: https://www.launchpadhq.io/platform/pika Data: https://www.launchpadhq.io/data/v1/tools/pika.json Category: Video Creation & Editing Company: Pika Labs Website: https://pika.art Cost tier: FREEMIUM Difficulty: ADVANCED Lifecycle: active Primary use: Fast AI video for social media Free tier: yes (Free $0 with 0 monthly credits (credit packs only); no-watermark listed on Free in the compare table) Starting price: Free, from $10/mo Lowest paid plan: Starter, $10/mo Plans: - Free: $0 (Free: 150 credits/mo (about 30 videos); Standard: 700/mo; Pro: 2000/mo; Unlimited option) - Starter: $10/mo ($8 annual, 900 credits/mo) - Creator: $35/mo ($28 annual, 3,150 credits) - Fancy: $95–$880/mo ($76–$704 annual) Alternatives: Runway, Luma Dream Machine (Ray3), Google Veo 3.1 Guide: https://www.launchpadhq.io/platform/pika/guide Data as of: 2026-10-14 Sources: - https://pika.art ### Pinecone Canonical: https://www.launchpadhq.io/platform/pinecone Data: https://www.launchpadhq.io/data/v1/tools/pinecone.json Category: AI APIs & Developer Services Company: Pinecone Website: https://pinecone.io Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Vector database for AI/RAG applications Free tier: yes (Starter Free: on-demand database, inference, assistant; usage caps (e.g. ~2 GB storage)) Starting price: $20/mo flat Lowest paid plan: Builder, $20/mo flat Plans: - Starter Free: See pricing - Builder: $20/mo flat - Standard: $50/mo minimum then pay-as-you-go - Enterprise: $500/mo minimum Guide: https://www.launchpadhq.io/platform/pinecone/guide Data as of: 2026-10-14 Sources: - https://pinecone.io ### PlayHT Canonical: https://www.launchpadhq.io/platform/play-ht Data: https://www.launchpadhq.io/data/v1/tools/play-ht.json Category: Voice & Speech Cost tier: DISCONTINUED Difficulty: BEGINNER Lifecycle: discontinued Status: PlayHT is no longer available. Bloomberg reported in July 2025 that Meta acquired PlayAI, the company behind PlayHT, and that its team was joining Meta. Description: PlayHT was an AI text-to-speech and voice-cloning service for voiceovers, podcasts, e-learning and apps. Primary use: Discontinued: former AI text-to-speech and voice-cloning service Status: discontinued (no current pricing) Data as of: 2026-10-20 Sources: - https://www.bloomberg.com/news/articles/2025-07-11/meta-acquires-voice-ai-startup-playai-continuing-to-add-talent (checked 2026-10-20) ### Poe Canonical: https://www.launchpadhq.io/platform/poe Data: https://www.launchpadhq.io/data/v1/tools/poe.json Category: Text & Conversational AI Company: Quora Website: https://poe.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Quora's multi-model chat app. Official about page : free for most usage; plans from $4.99/mo. Prompts are sent to whichever third-party model provider you pick. Exact free daily points are unverified (third parties cite about 300 points/day). Third-party paid point allowances: $4.99 (10k points/day), $19.99 (660k points/mo), then $49.99, $99.99, $249.99. Primary use: One app to try many third-party models Free tier: yes (Free for most usage; exact daily points unverified) Starting price: Free, from $4.99/mo Lowest paid plan: Plans from, $4.99/mo Plans: - Plans from: $4.99/mo - higher tiers: $19.99 / $49.99 / $99.99 / $249.99 (third-party point tables) - Free: Free (allowance unverified) Guide: https://www.launchpadhq.io/platform/poe/guide Data as of: 2026-10-13 Sources: - https://poe.com ### QuillBot Canonical: https://www.launchpadhq.io/platform/quillbot Data: https://www.launchpadhq.io/data/v1/tools/quillbot.json Category: Text & Conversational AI Company: QuillBot, a Learneo Inc. business Website: https://quillbot.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Paraphrasing and writing helper. Official pricing is unverified (official page blocked). Treat plan names and dollar amounts as unpublished in this catalog pass. Usual SaaS data-handling — do not paste confidential drafts without reading the privacy policy. Primary use: Paraphrasing and writing helper Tasks: paraphrase, grammar-check, writing, summarization Free tier: yes (Paraphrase up to 100 words, summarize up to 1,200 words.) Starting price: $8.33/mo Lowest paid plan: Premium, $8.33/mo/billed annually Plans: - Premium: $8.33/mo/billed annually Guide: https://www.launchpadhq.io/platform/quillbot/guide Data as of: 2026-10-14 Sources: - https://quillbot.com/premium (checked 2026-10-14) ### Quizlet Q-Chat Canonical: https://www.launchpadhq.io/platform/quizlet-q-chat Data: https://www.launchpadhq.io/data/v1/tools/quizlet-q-chat.json Category: Education & Learning AI Company: Quizlet Website: https://quizlet.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI study assistant and flashcard generation Free tier: yes (Free: basic flashcards; limited AI features; ad-supported) Starting price: Free, from $35.99/year Lowest paid plan: Plus, $35.99/year Plans: - Free: Free (Free: limited AI explanations; Plus: $8/mo enhanced AI + no ads; Teachers: $4/mo) - Plus: $35.99/year Alternatives: Khan Academy (Khanmigo) Guide: https://www.launchpadhq.io/platform/quizlet-q-chat/guide Data as of: not dated Sources: - https://quizlet.com ### Qwen Canonical: https://www.launchpadhq.io/platform/qwen Data: https://www.launchpadhq.io/data/v1/tools/qwen.json Category: International & Regional AI Company: Powered by Alibaba Cloud Website: https://qwen.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Alibaba's Qwen family. The hosted chat/API is operated by Alibaba. Open-weight models can be run locally (for example through Ollama or LM Studio), in which case prompts stay on your machine. Official Model Studio international API (re-checked 2026-10-14 at alibabacloud.com/help/en/model-studio/model-pricing): qwen-flash $0.05 input / $0.40 output per 1M tokens (≤256K); qwen3.8-max $2 / $6 per 1M tokens. Qwen Chat is described as free with no subscription; exact chat limits remain unverified. Primary use: Hosted Qwen Chat and Alibaba Cloud Model Studio API; open weights for local use Free tier: yes (free to use, open to all) Starting price: $0.05/$0.40 per 1M tokens Lowest paid plan: API from, $0.05/$0.40 per 1M tokens Plans: - API from: $0.05/$0.40 per 1M tokens - qwen3.8-max: $2/$6 per 1M - Qwen Chat free (limits unverified): See pricing Guide: https://www.launchpadhq.io/platform/qwen/guide Data as of: 2026-10-14 Sources: - https://qwen.ai/ (checked 2026-10-14) ### Ramp Intelligence Canonical: https://www.launchpadhq.io/platform/ramp-intelligence Data: https://www.launchpadhq.io/data/v1/tools/ramp-intelligence.json Category: Finance & Real Estate AI Company: Ramp Website: https://ramp.com Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI-powered corporate card and expense management Free tier: yes (Free for Ramp card customers (no card required for basic spend management)) Starting price: Free, from $12/user/mo Lowest paid plan: Ramp Plus, $12/user/mo Plans: - Free: Free (Free with Ramp account; advanced features on Ramp Plus ($12/user/mo)) - Ramp Plus: $12/user/mo Guide: https://www.launchpadhq.io/platform/ramp-intelligence/guide Data as of: not dated Sources: - https://ramp.com ### Read AI Canonical: https://www.launchpadhq.io/platform/read-ai Data: https://www.launchpadhq.io/data/v1/tools/read-ai.json Category: Meetings & Notes Company: Read AI Website: https://www.read.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Meeting notes and cross-meeting search., some US states require all-party consent to record — tell every attendee before recording and follow the law where participants are located. Official pricing (read.ai/pricing): Free 5 meeting transcripts/mo, no card; paid per user per month $19.75 ($15 annual), $29.75 ($22.50 annual), $39.75 ($29.75 annual). Primary use: Meeting notes plus search across past meetings Tasks: meeting-notes Free tier: yes (Free, no card: 5 meeting transcripts and 30 measured meetings per month, meetings up to 1 hour, no audio or video uploads.) Starting price: Free, from $15 Lowest paid plan: Pro, $15/user/month Plans: - Free: $0 (5 meeting transcripts/month; 30 measured meetings/month; 1-hour max meeting length) - Pro: $15/user/month (Billed annually; $19.75/user/month billed monthly. Unlimited transcripts, 100 upload min/month, 4-hour meetings) - Enterprise: $22.50/user/month (Billed annually; $29.75/user/month billed monthly. 200 upload min/month) - Enterprise+: $29.75/user/month (Billed annually, requires 5+ licenses; $39.75/user/month billed monthly. 300 upload min/month, 8-hour meetings, HIPAA, SAML/SCIM) Guide: https://www.launchpadhq.io/platform/read-ai/guide Data as of: 2026-10-26 Sources: - https://www.read.ai/pricing (checked 2026-10-25) - https://www.read.ai/plans-pricing (checked 2026-10-26) ### Recraft Canonical: https://www.launchpadhq.io/platform/recraft Data: https://www.launchpadhq.io/data/v1/tools/recraft.json Category: Image Generation & Editing Company: Recraft AI Website: https://recraft.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI images for design and branding Tasks: logo-design, graphic-design, image-generation Free tier: yes (Free plan available. Images made on it are owned by Recraft, shown publicly in the community gallery, and not licensed for commercial use.) Starting price: Free, from See pricing Lowest paid plan: Pro plans available, See pricing Plans: - Free: Free - Pro plans available: See pricing Features: - Top-up credits are available on the Basic, Pro and Teams plans; on Teams workspaces they are shared across all seats. - Plans can be paid monthly or annually, with annual billing described as saving 20%. - Paid-plan images stay private and are not shown in the community gallery. Alternatives: Ideogram, Adobe Firefly, Leonardo AI, Midjourney Guide: https://www.launchpadhq.io/platform/recraft/guide Data as of: 2026-10-26 Sources: - https://www.recraft.ai/pricing (checked 2026-10-26) ### Reducto Canonical: https://www.launchpadhq.io/platform/reducto Data: https://www.launchpadhq.io/data/v1/tools/reducto.json Category: Document & PDF Processing Company: Reducto Website: https://reducto.ai Cost tier: FREE Difficulty: BEGINNER Lifecycle: active Primary use: AI document parsing and extraction for developers Free tier: yes (Free: 200 pages/month; basic parsing; API access) Starting price: Free, from See pricing Lowest paid plan: paid plans, See pricing Plans: - Free: Free (Free: 200 pages/mo; Starter: 2,000; Pro: 10,000; Enterprise: custom) - paid plans: See pricing Alternatives: Upstage (Solar), Nanonets, Docsumo Guide: https://www.launchpadhq.io/platform/reducto/guide Data as of: not dated Sources: - https://reducto.ai ### Relevance AI Canonical: https://www.launchpadhq.io/platform/relevance-ai Data: https://www.launchpadhq.io/data/v1/tools/relevance-ai.json Category: Workflow & Automation Company: Relevance AI Website: https://relevanceai.com Cost tier: FREE Difficulty: INTERMEDIATE Lifecycle: active Primary use: No-code AI agent and workflow builder Free tier: yes (Free tier: 100 credits/day; basic agent building; limited tools) Starting price: Free, from See pricing Lowest paid plan: Pro plans available, See pricing Plans: - Free: Free (Free: 100 credits/day; Pro: 10K/mo; Business: 100K/mo; Enterprise: custom) - Pro plans available: See pricing Alternatives: Gumloop, Lindy AI Guide: https://www.launchpadhq.io/platform/relevance-ai/guide Data as of: not dated Sources: - https://relevanceai.com ### Replicate Canonical: https://www.launchpadhq.io/platform/replicate Data: https://www.launchpadhq.io/data/v1/tools/replicate.json Category: AI APIs & Developer Services Company: Replicate Website: https://replicate.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Run open-source AI models via API Free tier: yes (Free: some models have free predictions; hardware billing per second of compute) Starting price: See pricing Lowest paid plan: Pay-per-second of compute, See pricing Plans: - Pay-per-second of compute: See pricing Alternatives: Together AI, Fireworks AI Guide: https://www.launchpadhq.io/platform/replicate/guide Data as of: not dated Sources: - https://replicate.com ### Replit Agent Canonical: https://www.launchpadhq.io/platform/replit-agent Data: https://www.launchpadhq.io/data/v1/tools/replit-agent.json Category: AI Coding & Development Company: Replit Website: https://replit.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Cloud IDE for building apps from prompts Tasks: coding Free tier: yes (Free Replit account with limited compute; Agent available on paid plans) Starting price: $20/mo ($18/mo billed annually) including $20 toward more powerful models Lowest paid plan: Core, $20/mo ($18/mo billed annually) including $20 toward more powerful models Plans: - Core: $20/mo ($18/mo billed annually) including $20 toward more powerful models - higher individual tier: $100/mo ($90 annual) with $100 of model credits - Enterprise: Custom Alternatives: Lovable, Bolt.new, v0, Google Antigravity Guide: https://www.launchpadhq.io/platform/replit-agent/guide Data as of: 2026-10-14 Sources: - https://replit.com ### Research Rabbit Canonical: https://www.launchpadhq.io/platform/research-rabbit Data: https://www.launchpadhq.io/data/v1/tools/research-rabbit.json Category: Research & Academic Tools Company: Research Rabbit Website: https://researchrabbit.ai Cost tier: FREE Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI-powered paper recommendations and similarity clustering Tasks: research-papers Free tier: yes (Completely free; unlimited paper discovery; collections; recommendations; collaboration) Starting price: Free Plans: - Free: Free (No limits; fully free platform) Alternatives: Connected Papers, Litmaps, Semantic Scholar Guide: https://www.launchpadhq.io/platform/research-rabbit/guide Data as of: not dated Sources: - https://researchrabbit.ai ### Restb.ai Canonical: https://www.launchpadhq.io/platform/restb-ai Data: https://www.launchpadhq.io/data/v1/tools/restb-ai.json Category: Finance & Real Estate AI Company: Restb.ai Website: https://restb.ai Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI computer vision for real estate Free tier: yes (Free API trial with limited calls) Starting price: See pricing Lowest paid plan: API pay-per-use, See pricing Plans: - API pay-per-use: See pricing Guide: https://www.launchpadhq.io/platform/restb-ai/guide Data as of: not dated Sources: - https://restb.ai ### Retell AI Canonical: https://www.launchpadhq.io/platform/retell-ai Data: https://www.launchpadhq.io/data/v1/tools/retell-ai.json Category: AI APIs & Developer Services Company: Retell AI Website: https://www.retellai.com Cost tier: FREEMIUM Difficulty: ADVANCED Lifecycle: active Description: Voice-agent platform. Official pricing (retellai.com/pricing): starts at $0, pay for minutes; compare table lists voice AI at $0.07–$0.31/minute; 20 concurrent calls included. A $10 free-credit figure was not on the fetched page — unverified. Voice-agent calls may require recording-consent / disclosure compliance by the user depending on jurisdiction. Primary use: Pay-as-you-go voice AI phone agents Free tier: yes ($10 in free credits; 20 free concurrent calls) Starting price: $0.07–$0.31/min Lowest paid plan: Pay as you go, $0.07–$0.31/min Plans: - Pay as you go: $0.07–$0.31/min (AI voice agents; AI chat agents $0.002+/msg) - Enterprise: Custom Data as of: 2026-10-14 Sources: - https://www.retellai.com/pricing (checked 2026-10-14) ### Rezi Canonical: https://www.launchpadhq.io/platform/rezi Data: https://www.launchpadhq.io/data/v1/tools/rezi.json Category: Career & Job Search Company: Rezi Website: https://www.rezi.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Rezi is an AI resume builder that scores a resume on 23 metrics (Rezi Score) and tailors it to a job description's keywords. It also writes cover letters and resignation letters, offers AI interview practice, and has a Chrome extension that auto-fills job applications. Its privacy policy does not describe how resume content is handled by AI features or whether it is used for training. Primary use: AI resume builder with ATS scoring Tasks: resume, cover-letter, interview-prep, job-search Free tier: yes (1 resume, 1 AI interview, 3 PDF downloads, limited keyword targeting and content analysis, unlimited cover and resignation letters. No card required.) Starting price: Free, from $29/month Lowest paid plan: Pro, $29/month Plans: - Free: $0/month (1 resume, 3 PDF downloads) - Pro: $29/month (Unlimited resumes, AI and downloads; 1 expert resume review per month) - Lifetime: $149/one-time (Pro features; no free monthly expert review) - Enterprise: $99/month (Per 200 users; for universities, agencies and career sites; SSO, webhooks, team roles) Features: - Rezi Score rates a resume on 23 metrics for ATS and hiring readiness. - AI Keyword Targeting tailors a resume to a job description's keywords. - AI resume writer, editor and summary writer. - AI cover letter and resignation letter writers. - AI Interview practice. - Exports to PDF, Microsoft Word (DOCX) and Google Drive. Alternatives: Teal, Kickresume Data as of: 2026-10-26 Sources: - https://www.rezi.ai/legal (checked 2026-10-26) - https://www.rezi.ai/pricing (checked 2026-10-26) - https://www.rezi.ai/rezi-chrome-extension (checked 2026-10-26) - https://www.rezi.ai/ (checked 2026-10-26) ### Runway Canonical: https://www.launchpadhq.io/platform/runway Data: https://www.launchpadhq.io/data/v1/tools/runway.json Category: Video Creation & Editing Company: Runway AI Website: https://runwayml.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Professional AI video generation and filmmaking Free tier: yes (Free plan: one-time 125 credits that do not expire) Starting price: $15/mo monthly or $12/mo billed annually Lowest paid plan: Standard, $15/mo monthly or $12/mo billed annually Plans: - Standard: $15/mo monthly or $12/mo billed annually - Pro: $35/$28 - Max: $95/$76. The old “Pro $76” figure is Max billed annually. Alternatives: Pika, Google Veo 3.1, Luma Dream Machine (Ray3) Guide: https://www.launchpadhq.io/platform/runway/guide Data as of: 2026-10-14 Sources: - https://runwayml.com ### Runway for Education Canonical: https://www.launchpadhq.io/platform/runway-for-education Data: https://www.launchpadhq.io/data/v1/tools/runway-for-education.json Category: Education & Learning AI Company: Runway AI Website: https://runway.com/educators Cost tier: PAID Difficulty: INTERMEDIATE Lifecycle: active Description: Runway's education programme is a discount, not a free plan: verified students and educators get 25% off any paid Runway plan, and schools buying 3 or more classroom licenses get up to 25% off. Primary use: AI creative tools for educational institutions Free tier: no (No free education plan. Students and educators get 25% off a paid Runway plan instead.) Starting price: 25% off Lowest paid plan: Students & educators, 25% off Plans: - Students & educators: 25% off (Any paid Runway plan; eligibility verified through SheerID) - Classroom licensing: Up to 25% off (Classes of 3 or more licenses; a school-issued email for each account) Data as of: 2026-10-25 Sources: - https://runway.com/educators (checked 2026-10-20) ### Sakana AI Canonical: https://www.launchpadhq.io/platform/sakana-ai Data: https://www.launchpadhq.io/data/v1/tools/sakana-ai.json Category: International & Regional AI Company: Sakana AI Website: https://sakana.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Japanese AI research lab — nature-inspired AI Free tier: yes (Research models available; no consumer product yet; API in development) Starting price: See pricing Lowest paid plan: Research, See pricing Plans: - Research: See pricing - API partnerships: See pricing Alternatives: Cohere, AI21 Labs (Jamba) Guide: https://www.launchpadhq.io/platform/sakana-ai/guide Data as of: not dated Sources: - https://sakana.ai ### Salesforce Einstein Canonical: https://www.launchpadhq.io/platform/salesforce-einstein Data: https://www.launchpadhq.io/data/v1/tools/salesforce-einstein.json Category: AI Plugins for Business Software Company: Salesforce Website: https://salesforce.com/einstein Cost tier: ENTERPRISE Difficulty: BEGINNER Lifecycle: active Primary use: AI-powered CRM features Free tier: no (Included in Salesforce editions (Sales Cloud from $25/user/mo); AI features vary by tier) Starting price: $500/mo Lowest paid plan: Einstein 1 from, $500/mo Plans: - Included with Salesforce: See pricing - Einstein 1 from: $500/mo Guide: https://www.launchpadhq.io/platform/salesforce-einstein/guide Data as of: not dated Sources: - https://salesforce.com/einstein ### Sarvam AI Canonical: https://www.launchpadhq.io/platform/sarvam-ai Data: https://www.launchpadhq.io/data/v1/tools/sarvam-ai.json Category: International & Regional AI Company: Axonwise Private Limited Website: https://www.sarvam.ai Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Description: India-based models for Indic-language speech, TTS, and chat., official API pricing (sarvam.ai/api-pricing) is pay-as-you-go after starter credits. Example: Bulbul V4 Flash TTS ₹2.00 per 1k characters, dropping to ₹1.70 on higher volume tiers. Voice agents are listed “as low as ₹3/min”. Starter credit amount is unverified. Review data-handling for inputs processed in India. Primary use: Indian-language speech, TTS, and chat models (API) Free tier: yes (Start with free credits (amount unpublished). Credits never expire, with no monthly commitment.) Starting price: See pricing Lowest paid plan: Pay-as-you-go (e.g. Bulbul V4 Flash TTS ₹2.00 per 1k chars, ₹1.70 higher tiers), See pricing Plans: - Pay-as-you-go (e.g. Bulbul V4 Flash TTS ₹2.00 per 1k chars, ₹1.70 higher tiers): See pricing - voice agents from ₹3/min: See pricing - starter credits amount unverified: See pricing API pricing: INR. Bulbul V4 Flash text-to-speech: ₹2.00 per 1,000 characters (pay as you go) down to ₹1.70 (Growth/Business). Bulbul V3: ₹3.00 → ₹2.70 per 1,000 characters. Speech-to-text: ₹0.49/min (pay as you go) down to ₹0.41/min; with diarization ₹0.59 → ₹0.50/min. Sarvam 105B LLM: ₹15 input / ₹5 cached / ₹60 output per 1M tokens. Voice agents as low as ₹3 a minute. Data as of: 2026-10-14 Sources: - https://www.sarvam.ai/ (checked 2026-10-14) - https://www.sarvam.ai/api-pricing (checked 2026-10-14) ### Scale AI Canonical: https://www.launchpadhq.io/platform/scale-ai Data: https://www.launchpadhq.io/data/v1/tools/scale-ai.json Category: AI APIs & Developer Services Company: Scale AI Website: https://scale.com Cost tier: ENTERPRISE Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI data labeling and model evaluation platform Free tier: no (No free tier; enterprise/project-based pricing) Starting price: Custom Lowest paid plan: Enterprise, Custom Plans: - Enterprise: Custom Guide: https://www.launchpadhq.io/platform/scale-ai/guide Data as of: not dated Sources: - https://scale.com ### Scenario Canonical: https://www.launchpadhq.io/platform/scenario Data: https://www.launchpadhq.io/data/v1/tools/scenario.json Category: Gaming & Creative AI Company: Scenario Website: https://scenario.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI game asset generation Free tier: yes (Free: 100 images/month; basic models; community generators) Starting price: Free, from $29/mo Lowest paid plan: Pro, $29/mo Plans: - Free: Free (Free: 100 images/mo; Pro: 2,000; Team: 10,000; Enterprise: custom) - Pro: $29/mo Alternatives: Leonardo AI, Midjourney Guide: https://www.launchpadhq.io/platform/scenario/guide Data as of: not dated Sources: - https://scenario.com ### Scholarcy Canonical: https://www.launchpadhq.io/platform/scholarcy Data: https://www.launchpadhq.io/data/v1/tools/scholarcy.json Category: Research & Academic Tools Company: Scholarcy Website: https://scholarcy.com Cost tier: FREE Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI paper summarization and flashcard generation Tasks: research-papers, summarization Free tier: yes (Free Chrome extension: summarize any article; flashcard generation; limited features) Starting price: Free, from See pricing Lowest paid plan: paid plans, See pricing Plans: - Free: Free (Free extension: individual articles only; Library plans: bulk processing + reference extraction) - paid plans: See pricing Alternatives: SciSpace, Gemini Notebook, Anara, ChatPDF Guide: https://www.launchpadhq.io/platform/scholarcy/guide Data as of: not dated Sources: - https://scholarcy.com ### SciSpace Canonical: https://www.launchpadhq.io/platform/scispace Data: https://www.launchpadhq.io/data/v1/tools/scispace.json Category: Research & Academic Tools Company: SciSpace (Typeset) Website: https://typeset.io Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI paper reading, literature review, and writing Tasks: research-papers Free tier: yes (Free: limited paper reads; basic Q&A; limited highlights; Copilot queries) Starting price: Free, from $20/mo Lowest paid plan: Premium, $20/mo Plans: - Free: Free (Free: limited Copilot queries/day; Premium: unlimited Copilot; bulk processing; higher limits) - Premium: $20/mo - Advanced: $70/mo Alternatives: Semantic Scholar, Elicit, Scholarcy, Anara Guide: https://www.launchpadhq.io/platform/scispace/guide Data as of: not dated Sources: - https://typeset.io ### Scite Canonical: https://www.launchpadhq.io/platform/scite Data: https://www.launchpadhq.io/data/v1/tools/scite.json Category: Research & Academic Tools Company: Scite Website: https://scite.ai Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI citation analysis (support vs. contrast) Tasks: research-papers Free tier: yes (Free: basic search; limited Smart Citation results; community dashboard) Starting price: Free, from $20/mo Lowest paid plan: Premium, $20/mo Plans: - Free: Free (Free: limited full-text access; Individual: unlimited; Institutional: campus-wide) - Premium: $20/mo Alternatives: Semantic Scholar, Elicit, Connected Papers, Consensus Guide: https://www.launchpadhq.io/platform/scite/guide Data as of: not dated Sources: - https://scite.ai ### Semantic Scholar Canonical: https://www.launchpadhq.io/platform/semantic-scholar Data: https://www.launchpadhq.io/data/v1/tools/semantic-scholar.json Category: Research & Academic Tools Company: Allen Institute for AI Website: https://semanticscholar.org Cost tier: FREE Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI-powered academic paper discovery Tasks: research-papers Free tier: yes (Completely free; unlimited search; TLDR summaries; API access; citation data) Starting price: Free Plans: - Free: Free (API: 1 request/second default; authenticated: higher; web: unlimited) Alternatives: Elicit, Consensus, SciSpace, Connected Papers Guide: https://www.launchpadhq.io/platform/semantic-scholar/guide Data as of: not dated Sources: - https://semanticscholar.org ### Semrush AI Canonical: https://www.launchpadhq.io/platform/semrush-ai Data: https://www.launchpadhq.io/data/v1/tools/semrush-ai.json Category: Sales, Marketing & SEO AI Company: Semrush Website: https://semrush.com Cost tier: ENTERPRISE Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI-enhanced SEO and marketing analytics Free tier: no (Free account: 10 analytics reports/day; limited keyword tracking; basic features) Starting price: $129.95/mo Lowest paid plan: Pro, $129.95/mo Plans: - Pro: $129.95/mo - Guru: $249.95/mo Guide: https://www.launchpadhq.io/platform/semrush-ai/guide Data as of: not dated Sources: - https://semrush.com ### Shopify Magic / Sidekick Canonical: https://www.launchpadhq.io/platform/shopify-magic-sidekick Data: https://www.launchpadhq.io/data/v1/tools/shopify-magic-sidekick.json Category: AI Plugins for Business Software Company: Shopify Website: https://shopify.com/magic Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI tools for Shopify merchants Free tier: yes (Included free with all Shopify plans (starting $39/mo)) Starting price: See pricing Lowest paid plan: Included with Shopify plans, See pricing Plans: - Included with Shopify plans: See pricing Alternatives: Wix AI, Squarespace AI Guide: https://www.launchpadhq.io/platform/shopify-magic-sidekick/guide Data as of: not dated Sources: - https://shopify.com/magic ### Slack AI Canonical: https://www.launchpadhq.io/platform/slack-ai Data: https://www.launchpadhq.io/data/v1/tools/slack-ai.json Category: AI Plugins for Business Software Company: Salesforce (Slack) Website: https://slack.com/features/ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI search and summarization in Slack Free tier: yes (No free tier for AI features; requires Slack Business+ ($12.50/user/mo) or Enterprise) Starting price: $10/user Lowest paid plan: $10/user/mo add-on, $10/user Plans: - $10/user/mo add-on: $10/user Guide: https://www.launchpadhq.io/platform/slack-ai/guide Data as of: not dated Sources: - https://slack.com/features/ai ### Sora 2 Canonical: https://www.launchpadhq.io/platform/sora-2 Data: https://www.launchpadhq.io/data/v1/tools/sora-2.json Category: Video Creation & Editing Company: OpenAI Website: https://sora.com Cost tier: DISCONTINUED Difficulty: BEGINNER Lifecycle: discontinued Status: OpenAI discontinued the Sora web and app experiences on April 26, 2026, and the Sora API on September 24, 2026. Description: Sora was OpenAI's text-to-video app and API for making short video clips, with sound, from a written prompt. Primary use: Discontinued: former OpenAI text-to-video app and API Status: discontinued (no current pricing) Alternatives: Runway, Pika, Google Veo 3.1, Luma Dream Machine (Ray3) Data as of: 2026-10-20 Sources: - https://help.openai.com/en/articles/20001152-what-to-know-about-the-sora-discontinuation (checked 2026-10-20) ### Soundraw Canonical: https://www.launchpadhq.io/platform/soundraw Data: https://www.launchpadhq.io/data/v1/tools/soundraw.json Category: Music Generation Company: Soundraw Website: https://soundraw.io Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Customizable royalty-free AI background music Free tier: yes (Unlimited song creation and previews; download requires paid plan) Starting price: $16.99/mo Lowest paid plan: Creator, $16.99/mo Plans: - Creator: $16.99/mo Alternatives: Mubert, AIVA, Beatoven.ai, Suno Guide: https://www.launchpadhq.io/platform/soundraw/guide Data as of: not dated Sources: - https://soundraw.io ### Speak (Language AI) Canonical: https://www.launchpadhq.io/platform/speak-language-ai Data: https://www.launchpadhq.io/data/v1/tools/speak-language-ai.json Category: Education & Learning AI Company: Speak Website: https://speak.com Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI-powered English conversation practice Free tier: yes (Free: limited lessons; basic AI conversation; sample content) Starting price: $13.99/mo Lowest paid plan: Premium, $13.99/mo Plans: - Premium: $13.99/mo Alternatives: Duolingo Max Guide: https://www.launchpadhq.io/platform/speak-language-ai/guide Data as of: not dated Sources: - https://speak.com ### Speechify Canonical: https://www.launchpadhq.io/platform/speechify Data: https://www.launchpadhq.io/data/v1/tools/speechify.json Category: Voice & Speech Company: Speechify Inc Website: https://speechify.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Text-to-speech reader. Official pricing (speechify.com/pricing): Free includes 10 basic (“robotic”) voices and up to 1.5x speed; Premium $29/mo billed monthly. The site advertises annual billing that “saves 60%”; the exact annual dollar price is unverified. Usual SaaS data-handling. Primary use: Text-to-speech reader for documents and web pages Tasks: text-to-speech Free tier: yes (Speeds up to 1.5x, 10 robotic-sounding voices, text-to-speech features only.) Starting price: $29/Month Lowest paid plan: Premium, $29/Month Plans: - Premium: $29/Month (Prices shown per month, billed monthly) Guide: https://www.launchpadhq.io/platform/speechify/guide Data as of: 2026-10-14 Sources: - https://speechify.com/pricing/ (checked 2026-10-14) ### Spellbook Canonical: https://www.launchpadhq.io/platform/spellbook Data: https://www.launchpadhq.io/data/v1/tools/spellbook.json Category: Legal AI Company: Spellbook (Rally Legal) Website: https://spellbook.legal Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI contract drafting and review for lawyers Free tier: yes (Free trial; no permanent free tier) Starting price: $99/user/mo Lowest paid plan: From, $99/user/mo Plans: - From: $99/user/mo Alternatives: Harvey AI, Luminance, Kira Guide: https://www.launchpadhq.io/platform/spellbook/guide Data as of: not dated Sources: - https://spellbook.legal ### Squarespace AI Canonical: https://www.launchpadhq.io/platform/squarespace-ai Data: https://www.launchpadhq.io/data/v1/tools/squarespace-ai.json Category: AI Plugins for Business Software Company: Squarespace Website: https://squarespace.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI content and design for websites Free tier: yes (14-day free trial; no permanent free tier; plans from $16/mo) Starting price: $16/mo Lowest paid plan: Included with plans from, $16/mo Plans: - Included with plans from: $16/mo Alternatives: Wix AI, WordPress AI (Jetpack AI) Guide: https://www.launchpadhq.io/platform/squarespace-ai/guide Data as of: not dated Sources: - https://squarespace.com ### Stability AI Canonical: https://www.launchpadhq.io/platform/stability-ai Data: https://www.launchpadhq.io/data/v1/tools/stability-ai.json Category: Local & Open-Source AI Company: Stability AI Website: https://stability.ai Cost tier: FREE Difficulty: ADVANCED Lifecycle: active Description: Publisher of Stable Diffusion and the Stable Audio models. There are three ways in: download the open models and run them yourself (for example with ComfyUI), which keeps your inputs on your machine; call them from code through the Developer Platform API, which offers image, audio and 3D models but no video (Stability's video models are download-only); or make music and sound effects in the Stable Audio web app at stableaudio.com, its remaining consumer app. Its consumer image app, Stable Assistant, is no longer on Stability's site. Primary use: Open image, audio and 3D models, a pay-per-use API, and the Stable Audio web app Free tier: yes (New Developer Platform accounts get 25 free credits) Starting price: Free, from US$0.01 per credit, pay as you go Lowest paid plan: Developer Platform API, US$0.01 per credit, pay as you go Plans: - Open models (Community License): $0 for individuals and organizations under US$1M annual revenue, commercial users must register - Developer Platform API: US$0.01 per credit, pay as you go - Enterprise License (over US$1M annual revenue): custom - Stable Audio web app: separate plans at stableaudio.com/pricing Guide: https://www.launchpadhq.io/platform/stability-ai/guide Data as of: 2026-10-25 Sources: - https://stability.ai ### Suno Canonical: https://www.launchpadhq.io/platform/suno Data: https://www.launchpadhq.io/data/v1/tools/suno.json Category: Music Generation Company: Suno Inc. Website: https://suno.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI full-song generation with vocals Free tier: yes (10 songs/day (50 credits); 2 concurrent generations; non-commercial use) Starting price: Free, from $10/mo Lowest paid plan: Pro, $10/mo Plans: - Free: Free (Free: 50 credits/day (10 songs); Pro: 2,500 credits/mo; Premier: 10,000 credits/mo) - Pro: $10/mo - Premier: $30/mo Alternatives: Udio, AIVA, Soundraw, Beatoven.ai Guide: https://www.launchpadhq.io/platform/suno/guide Data as of: not dated Sources: - https://suno.com ### Surfer SEO Canonical: https://www.launchpadhq.io/platform/surfer-seo Data: https://www.launchpadhq.io/data/v1/tools/surfer-seo.json Category: Sales, Marketing & SEO AI Company: Surfer Website: https://surferseo.com Cost tier: ENTERPRISE Difficulty: INTERMEDIATE Lifecycle: acquired Status: Surfer has joined Group Positive. Surfer's announcement says the product keeps its name. Primary use: AI-powered SEO content optimization Free tier: no (No full-product free plan on surferseo.com/pricing; Keyword Surfer extension is separate and free) Starting price: $49/mo billed yearly Lowest paid plan: Discovery, $49/mo billed yearly Plans: - Discovery: $49/mo billed yearly - AI Search Analytics: $82/mo billed yearly - Standard: $99/mo billed yearly - next tier: $182/mo billed yearly - Peace of Mind: $299/mo billed yearly - Enterprise: Custom (. Monthly sticker prices were not the figures shown on the fetched cards.) Guide: https://www.launchpadhq.io/platform/surfer-seo/guide Data as of: 2026-10-20 Sources: - https://surferseo.com/blog/surfer-acquisition (checked 2026-10-20) ### Synthesia Canonical: https://www.launchpadhq.io/platform/synthesia Data: https://www.launchpadhq.io/data/v1/tools/synthesia.json Category: Video Creation & Editing Company: Synthesia Website: https://synthesia.io Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI avatar videos for training and marketing Free tier: yes (Basic $0: 10 min of video, 3 survey responses and 500 credits every month) Starting price: Free, from $29/mo billed monthly Lowest paid plan: Starter, $29/mo billed monthly Plans: - Basic: $0/mo - Starter: $29/mo billed monthly - Pro: $89/mo billed monthly ($64/mo billed yearly) - Enterprise: Custom (. Starter's yearly price is shown inconsistently on the official page, so it is not quoted) Alternatives: HeyGen, InVideo AI, Pictory AI Guide: https://www.launchpadhq.io/platform/synthesia/guide Data as of: 2026-10-25 Sources: - https://synthesia.io ### Tabnine Canonical: https://www.launchpadhq.io/platform/tabnine Data: https://www.launchpadhq.io/data/v1/tools/tabnine.json Category: AI Coding & Development Company: Tabnine Website: https://tabnine.com Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: acquired Status: Tricentis announced on July 30, 2026 that it has acquired Tabnine and will build Tabnine's Enterprise Context Engine into its Agentic Quality Engineering Platform. Primary use: Privacy-first AI code completion Tasks: coding Free tier: yes (Free: basic code completions; limited AI chat; single IDE) Starting price: Free, from $12/user/mo Lowest paid plan: Pro, $12/user/mo Plans: - Free: Free (Free: basic completions; Dev: full AI chat + better completions; Enterprise: custom + on-premise) - Pro: $12/user/mo Alternatives: GitHub Copilot, Cursor, Devin Desktop, Amazon Q Developer Guide: https://www.launchpadhq.io/platform/tabnine/guide Data as of: 2026-10-20 Sources: - https://www.tricentis.com/news/tricentis-acquires-tabnine (checked 2026-10-20) ### Tactiq Canonical: https://www.launchpadhq.io/platform/tactiq Data: https://www.launchpadhq.io/data/v1/tools/tactiq.json Category: Meetings & Notes Company: Tactiq Website: https://tactiq.io Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI meeting transcription browser extension Tasks: meeting-notes, transcription Free tier: yes (Free forever, no card: 10 transcripts/mo and 5 AI credits/mo) Starting price: Free, from $8/user/mo billed annually Lowest paid plan: Pro, $8/user/mo billed annually Plans: - Free: $0 (Free: 10 transcripts/mo and 5 AI credits/mo; Pro: unlimited transcripts and 10 AI credits/mo; Team and above: unlimited transcripts and AI credits) - Pro: $8/user/mo billed annually - Team: $16.67/user/mo billed annually - Business: $29.17/user/mo billed annually - Enterprise: Custom (. Monthly (non-annual) prices unverified) Alternatives: tl;dv, Otter AI, Zoom AI Companion Guide: https://www.launchpadhq.io/platform/tactiq/guide Data as of: 2026-10-25 Sources: - https://tactiq.io ### Teal Canonical: https://www.launchpadhq.io/platform/teal Data: https://www.launchpadhq.io/data/v1/tools/teal.json Category: Career & Job Search Company: Teal Labs, Inc. Website: https://www.tealhq.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Teal is a job-search workspace that combines an AI resume builder, a job application tracker and a Chrome extension for saving jobs from job boards. It scores a resume against a job description's keywords and drafts resume bullets, summaries and cover letters with AI. Per its privacy policy, AI features may send content to third-party providers such as OpenAI, and career content may be used in anonymized, aggregated form to improve its AI models. Primary use: AI resume builder and job tracker Tasks: resume, cover-letter, job-search, linkedin-profile Free tier: yes (Free Forever: unlimited resumes and job tracking, 10 resume templates, top 5 job-description keywords, basic resume analysis, 1 email template per job stage, and AI credits (5 resume bullets, 2 professional summaries, 2 cover letters).) Starting price: Free, from $13 Lowest paid plan: Teal+ 7-day, $13/7 days Plans: - Free Forever: $0 (Limited templates, keyword matching and AI credits) - Teal+ 7-day: $13/7 days (Unlimited AI credits, templates and keyword matching) - Teal+ 30-day: $29/30 days (Unlimited AI credits, templates and keyword matching) - Teal+ 90-day: $79/90 days (Unlimited AI credits, templates and keyword matching) Features: - AI resume builder with multiple tailored resume versions. - Job application tracker for bookmarking jobs, notes and application statuses. - Job description keyword matching with a Match Score against your resume. - AI cover letter, professional summary and resume bullet generators. - Email templates for each stage of the job search. - LinkedIn profile review and LinkedIn summary generator. Alternatives: Kickresume, Rezi Data as of: 2026-10-26 Sources: - https://www.tealhq.com/privacy-policy (checked 2026-10-26) - https://www.tealhq.com/pricing (checked 2026-10-26) - https://www.tealhq.com/ (checked 2026-10-26) ### Text Generation WebUI (oobabooga) Canonical: https://www.launchpadhq.io/platform/text-generation-webui-oobabooga Data: https://www.launchpadhq.io/data/v1/tools/text-generation-webui-oobabooga.json Category: Local & Open-Source AI Company: Open Source Website: https://github.com/oobabooga/text-generation-webui Cost tier: FREE Difficulty: ADVANCED Lifecycle: active Primary use: Feature-rich web interface for LLMs Free tier: yes (Completely free and open-source (AGPL-3.0)) Starting price: Free Plans: - Free: Free (None (hardware-limited)) Alternatives: Open WebUI, Ollama, LM Studio Guide: https://www.launchpadhq.io/platform/text-generation-webui-oobabooga/guide Data as of: not dated Sources: - https://github.com/oobabooga/text-generation-webui ### tl;dv Canonical: https://www.launchpadhq.io/platform/tl-dv Data: https://www.launchpadhq.io/data/v1/tools/tl-dv.json Category: Meetings & Notes Company: tl;dv Website: https://tldv.io Cost tier: FREEMIUM Difficulty: ADVANCED Lifecycle: active Primary use: AI meeting recording, transcription, and clips Tasks: meeting-notes, transcription Free tier: yes (Free $0, no card: unlimited video recordings and transcription in 30+ languages; AI notes for 10 meetings and 10 Ask AI queries included) Starting price: Free, from $18/seat/mo billed annually Lowest paid plan: Pro, $18/seat/mo billed annually Plans: - Free: $0 (Free: 10 meetings with AI notes and 10 Ask AI queries; Pro: unlimited AI notes; Business: unlimited multi-meeting AI insights) - Pro: $18/seat/mo billed annually - Business: $29/seat/mo billed annually - monthly USD prices not confirmed: See pricing Alternatives: Otter AI, Tactiq, Zoom AI Companion Guide: https://www.launchpadhq.io/platform/tl-dv/guide Data as of: 2026-10-25 Sources: - https://tldv.io ### Together AI Canonical: https://www.launchpadhq.io/platform/together-ai Data: https://www.launchpadhq.io/data/v1/tools/together-ai.json Category: AI APIs & Developer Services Company: Together AI Website: https://together.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Fast inference for open-source models Free tier: yes ($5 free credits for new accounts; pay-per-use after) Starting price: Free, from $0.04/M input tokens Lowest paid plan: Serverless chat from, $0.04/M input tokens Plans: - Serverless chat from: $0.04/M input tokens (Tev1 4B Experimental) - some listed models at: $0.00 - prepaid credit,: $5 minimum historically Alternatives: Groq, Fireworks AI, Replicate, OpenAI API Guide: https://www.launchpadhq.io/platform/together-ai/guide Data as of: 2026-10-14 Sources: - https://together.ai ### Topaz Labs Canonical: https://www.launchpadhq.io/platform/topaz-labs Data: https://www.launchpadhq.io/data/v1/tools/topaz-labs.json Category: Gaming & Creative AI Company: Topaz Labs Website: https://topazlabs.com Cost tier: ENTERPRISE Difficulty: BEGINNER Lifecycle: active Primary use: AI video/image upscaling and enhancement Free tier: no (Free trial; no permanent free tier; one-time purchase or subscription) Starting price: $199 Lowest paid plan: Photo AI, $199 Plans: - Video AI: $299 - Photo AI: $199 (perpetual) Guide: https://www.launchpadhq.io/platform/topaz-labs/guide Data as of: not dated Sources: - https://topazlabs.com ### Tray.ai Canonical: https://www.launchpadhq.io/platform/tray-ai Data: https://www.launchpadhq.io/data/v1/tools/tray-ai.json Category: Workflow & Automation Company: Tray.ai Website: https://tray.ai Cost tier: ENTERPRISE Difficulty: INTERMEDIATE Lifecycle: active Primary use: Enterprise workflow orchestration with AI Free tier: no (No free tier; demo/trial available on request) Starting price: Custom Lowest paid plan: Enterprise, Custom Plans: - Enterprise: Custom Alternatives: Workato, Microsoft Power Automate Guide: https://www.launchpadhq.io/platform/tray-ai/guide Data as of: not dated Sources: - https://tray.ai ### Trinka AI Canonical: https://www.launchpadhq.io/platform/trinka-ai Data: https://www.launchpadhq.io/data/v1/tools/trinka-ai.json Category: Research & Academic Tools Company: Crimson AI Website: https://trinka.ai Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI grammar checker for academic/scientific writing Tasks: grammar-check, academic-writing Free tier: yes (Free: 5,000 words/month; basic grammar; limited suggestions) Starting price: Free, from $20/mo Lowest paid plan: Premium, $20/mo Plans: - Free: Free (Free: 5K words/mo; Premium: unlimited; Enterprise: unlimited + API) - Premium: $20/mo Alternatives: Grammarly, Paperpal Guide: https://www.launchpadhq.io/platform/trinka-ai/guide Data as of: not dated Sources: - https://trinka.ai ### Udio Canonical: https://www.launchpadhq.io/platform/udio Data: https://www.launchpadhq.io/data/v1/tools/udio.json Category: Music Generation Company: Udio Website: https://udio.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI music with producer-level control Free tier: yes (10 songs/day; basic features; non-commercial) Starting price: Free, from $10/mo Lowest paid plan: Standard, $10/mo Plans: - Free: Free (600/mo) - Standard: $10/mo - Premium: $30/mo Alternatives: Suno, AIVA, Soundraw, Beatoven.ai Guide: https://www.launchpadhq.io/platform/udio/guide Data as of: not dated Sources: - https://udio.com ### Anara Canonical: https://www.launchpadhq.io/platform/unriddle Data: https://www.launchpadhq.io/data/v1/tools/unriddle.json Category: Document & PDF Processing Company: Unriddle Website: https://anara.com/ Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: renamed Status: Unriddle announced on March 23, 2025 that it changed its name to Anara. Primary use: AI research reading assistant Free tier: yes (Free: 5 documents; basic AI Q&A; note-taking; limited features) Starting price: Free, from $16/mo Lowest paid plan: Pro, $16/mo Plans: - Free: Free (Free: 5 docs; Starter: 30 docs/mo; Pro: unlimited; Team: unlimited + collaboration) - Pro: $16/mo Alternatives: ChatPDF, Gemini Notebook, SciSpace Guide: https://www.launchpadhq.io/platform/unriddle/guide Data as of: 2026-10-20 Sources: - https://anara.com/blog/the-beginning-of-a-new-story (checked 2026-10-20) ### Upstage (Solar) Canonical: https://www.launchpadhq.io/platform/upstage-solar Data: https://www.launchpadhq.io/data/v1/tools/upstage-solar.json Category: Document & PDF Processing Company: Upstage Website: https://upstage.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: Document AI and layout analysis API Free tier: yes (Free API credits for testing; Document Parse API trial) Starting price: See pricing Lowest paid plan: API pay-per-use, See pricing Plans: - API pay-per-use: See pricing Alternatives: Reducto, Nanonets, Docsumo Guide: https://www.launchpadhq.io/platform/upstage-solar/guide Data as of: not dated Sources: - https://upstage.ai ### v0 Canonical: https://www.launchpadhq.io/platform/v0 Data: https://www.launchpadhq.io/data/v1/tools/v0.json Category: AI Coding & Development Company: Vercel Website: https://v0.dev Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI UI component generation from prompts Tasks: coding Free tier: yes (Free $0: 7 messages/day limit) Starting price: Free, from $30/user/mo Lowest paid plan: Plus, $30/user/mo Plans: - Free: $0 (Free: limited generations/mo; Premium: 200 messages/mo; Team: custom) - Plus: $30/user/mo - Business: $100/user/mo. The old “Premium $20” name is gone. Alternatives: Bolt.new, Lovable, Replit Agent Guide: https://www.launchpadhq.io/platform/v0/guide Data as of: 2026-10-14 Sources: - https://v0.dev ### Vapi Canonical: https://www.launchpadhq.io/platform/vapi Data: https://www.launchpadhq.io/data/v1/tools/vapi.json Category: AI APIs & Developer Services Company: Vapi Website: https://vapi.ai Cost tier: FREEMIUM Difficulty: ADVANCED Lifecycle: active Description: Voice-agent API. Official pricing (vapi.ai/pricing): pay for minutes; $5 free credits on the no-success-package path; Vapi hosting listed at $0.05/min prepaid; Pro success package from $999/mo minimum. A Core success-package dollar amount did not render as $29 on this fetch — unverified. Voice-agent calls may require recording-consent / disclosure compliance by the user depending on jurisdiction. Primary use: API for building voice AI phone and web agents Free tier: yes ($5 free credits, 1 included phone number, 4 concurrent calls, 14-day data retention) Starting price: Free, from $29/month Lowest paid plan: Core, $29/month Plans: - Free: $0/month ($5 free credits; 1 included phone number; 4 concurrent calls; 14-day data retention) - Usage: Pay per minute (Provider costs passed through with no markup) - Core: $29/month - Pro: $999/mo minimum; 10% of Vapi hosting fee - Premier: Contact sales Data as of: 2026-10-14 Sources: - https://vapi.ai/pricing (checked 2026-10-14) ### VEED Canonical: https://www.launchpadhq.io/platform/veed Data: https://www.launchpadhq.io/data/v1/tools/veed.json Category: Video Creation & Editing Company: VEED Website: https://www.veed.io Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Browser video editor. Official pricing is unverified (JavaScript-only). The site advertises starting for free; paid dollar amounts are unpublished in this catalog pass. Primary use: Browser video editor with AI captions and cleanup Free tier: yes (Start for free advertised; limits unverified) Starting price: $25/mo monthly or $147/yr Lowest paid plan: Creator, $25/mo monthly or $147/yr Plans: - Creator: $25/mo monthly or $147/yr (per user) - Pro: $45/mo or $265/yr (per user) - Studio: $79/mo or $465/yr (per user) - Enterprise: custom Guide: https://www.launchpadhq.io/platform/veed/guide Data as of: 2026-10-14 Sources: - https://www.veed.io/pricing (checked 2026-10-14) ### Viz.ai Canonical: https://www.launchpadhq.io/platform/viz-ai Data: https://www.launchpadhq.io/data/v1/tools/viz-ai.json Category: Healthcare AI Company: Viz.ai Website: https://viz.ai Cost tier: ENTERPRISE Difficulty: ADVANCED Lifecycle: active Primary use: AI clinical decision support for stroke and disease detection Free tier: no (No free tier; FDA-cleared medical device (hospital purchase)) Starting price: Custom Lowest paid plan: Enterprise, Custom Plans: - Enterprise: Custom Guide: https://www.launchpadhq.io/platform/viz-ai/guide Data as of: not dated Sources: - https://viz.ai ### vLLM Canonical: https://www.launchpadhq.io/platform/vllm Data: https://www.launchpadhq.io/data/v1/tools/vllm.json Category: Local & Open-Source AI Company: Open Source (UC Berkeley) Website: https://vllm.ai Cost tier: FREE Difficulty: ADVANCED Lifecycle: active Primary use: High-throughput LLM serving engine Free tier: yes (Completely free and open-source (Apache 2.0)) Starting price: Free Plans: - Free: Free (None (hardware-limited; designed for GPU servers)) Alternatives: LocalAI Guide: https://www.launchpadhq.io/platform/vllm/guide Data as of: not dated Sources: - https://vllm.ai ### Warp Canonical: https://www.launchpadhq.io/platform/warp Data: https://www.launchpadhq.io/data/v1/tools/warp.json Category: AI Coding & Development Company: Warp Website: https://www.warp.dev Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: active Description: Modern terminal with Warp Agent. Official pricing (warp.dev/pricing): Free $0/month (core terminal, Warp Agent CLI, limited cloud agents). Build includes $20 of agent usage each month. Max includes $200 of agent usage each month. Business includes $20 of agent usage per seat each month plus SSO. Enterprise custom. The subscription sticker price for Build/Max (apart from included usage) was not shown as a single figure in the fetched HTML — unverified. Warp states SOC 2 and zero data retention with contracted LLM providers. Primary use: AI-enabled terminal with agents Free tier: yes (Terminal; pay-as-you-go or bring-your-own AI; limited cloud agents) Starting price: Free, from $20/month Lowest paid plan: Build, $20/month Plans: - Free: $0/month (Terminal; pay-as-you-go or bring-your-own AI; limited cloud agents) - Build: $20/month (Includes $20 of agent usage) - Max: $200/month (Includes $200 of usage) - Business: $50/user/month (Up to 25 seats) Data as of: 2026-10-14 Sources: - https://www.warp.dev/download (checked 2026-10-14) - https://www.warp.dev/pricing (checked 2026-10-14) ### Devin Desktop Canonical: https://www.launchpadhq.io/platform/windsurf Data: https://www.launchpadhq.io/data/v1/tools/windsurf.json Category: AI Coding & Development Company: Codeium Website: https://devin.ai/desktop Cost tier: FREEMIUM Difficulty: INTERMEDIATE Lifecycle: renamed Status: Cognition renamed Windsurf to Devin Desktop in June 2026. Existing users get the change as a routine update, and Cognition says plans, pricing and extensions stay the same. Primary use: Agentic AI IDE with flow-state coding Tasks: coding Free tier: yes (Free tier with AI completions; limited premium chat credits; tab completions) Starting price: Free, from $15-20/mo Lowest paid plan: Pro, $15-20/mo Plans: - Free: Free (Free: limited premium credits + unlimited basic completions; Pro: higher credits; Team/Enterprise: custom) - Pro: $15-20/mo - Max: $200/mo Alternatives: Cursor, GitHub Copilot, Claude Code, Amazon Q Developer Guide: https://www.launchpadhq.io/platform/windsurf/guide Data as of: 2026-10-20 Sources: - https://devin.ai/blog/windsurf-is-now-devin-desktop (checked 2026-10-20) ### Wispr Flow Canonical: https://www.launchpadhq.io/platform/wispr-flow Data: https://www.launchpadhq.io/data/v1/tools/wispr-flow.json Category: Voice & Speech Company: Wispr Website: https://wisprflow.ai Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Description: Dictation app that inserts transcribed speech into whatever you have focused., dictation audio is processed in the cloud — review before dictating sensitive or regulated content. Official pricing (wisprflow.ai/pricing): Free plan $0; Pro $15/user/mo ($12 billed annually); Growth $33/user/mo ($26 annual), which adds a notetaker and admin controls. Exact Free-plan usage limits are unverified. Primary use: AI dictation that types into any app Tasks: dictation Free tier: yes (2,000 words/week on desktop, 1,000/week on mobile.) Starting price: Free, from $15/user/mo Lowest paid plan: Pro, $15/user/mo Plans: - Free: $0 - Pro: $15/user/mo ($12 annual) - Growth: $33/user/mo ($26 annual) - Free: Free (plan limits unverified) Guide: https://www.launchpadhq.io/platform/wispr-flow/guide Data as of: 2026-10-14 Sources: - https://wisprflow.ai/pricing (checked 2026-10-14) ### Wix AI Canonical: https://www.launchpadhq.io/platform/wix-ai Data: https://www.launchpadhq.io/data/v1/tools/wix-ai.json Category: AI Plugins for Business Software Company: Wix Website: https://wix.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI website builder and content generation Free tier: yes (Wix free plan includes basic AI; limited features; Wix branding; no custom domain) Starting price: $17/mo Lowest paid plan: Included with Wix plans from, $17/mo Plans: - Included with Wix plans from: $17/mo Alternatives: Squarespace AI, WordPress AI (Jetpack AI) Guide: https://www.launchpadhq.io/platform/wix-ai/guide Data as of: not dated Sources: - https://wix.com ### Autodesk Flow Studio Canonical: https://www.launchpadhq.io/platform/wonder-dynamics-wonder-studio Data: https://www.launchpadhq.io/data/v1/tools/wonder-dynamics-wonder-studio.json Category: Gaming & Creative AI Company: Autodesk Website: https://www.autodesk.com/products/flow-studio/overview Cost tier: FREE Difficulty: BEGINNER Lifecycle: renamed Status: Autodesk now sells Wonder Studio as Autodesk Flow Studio. Autodesk's product page describes Flow Studio as formerly Wonder Studio. Primary use: AI VFX and 3D character animation Free tier: yes (Free: limited credits; basic VFX shots; watermarked) Starting price: Free, from See pricing Lowest paid plan: Pro plans, See pricing Plans: - Free: Free (Free: limited credits; Pro/Enterprise: higher credits; per-project pricing) - Pro plans: See pricing Guide: https://www.launchpadhq.io/platform/wonder-dynamics-wonder-studio/guide Data as of: 2026-10-20 Sources: - https://www.autodesk.com/products/flow-studio/overview (checked 2026-10-20) ### Wondershare PDFelement AI Canonical: https://www.launchpadhq.io/platform/wondershare-pdfelement-ai Data: https://www.launchpadhq.io/data/v1/tools/wondershare-pdfelement-ai.json Category: Document & PDF Processing Company: Wondershare Website: https://pdf.wondershare.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI-enhanced PDF editor with chat and summarization Free tier: yes (Free trial with limited AI features; full PDF editing free (watermarked)) Starting price: $79.99/year Lowest paid plan: From, $79.99/year Plans: - From: $79.99/year Alternatives: Adobe Acrobat AI Assistant, ChatPDF, Anara Guide: https://www.launchpadhq.io/platform/wondershare-pdfelement-ai/guide Data as of: not dated Sources: - https://pdf.wondershare.com ### WordPress AI (Jetpack AI) Canonical: https://www.launchpadhq.io/platform/wordpress-ai-jetpack-ai Data: https://www.launchpadhq.io/data/v1/tools/wordpress-ai-jetpack-ai.json Category: AI Plugins for Business Software Company: Automattic Website: https://jetpack.com/ai/ Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI content generation for WordPress Free tier: yes (Jetpack free plan includes some features; AI Assistant requires Jetpack paid plan or separate AI add-on) Starting price: See pricing Lowest paid plan: Included with Jetpack plans, See pricing Plans: - Included with Jetpack plans: See pricing Alternatives: Wix AI, Squarespace AI Guide: https://www.launchpadhq.io/platform/wordpress-ai-jetpack-ai/guide Data as of: not dated Sources: - https://jetpack.com/ai/ ### Workato Canonical: https://www.launchpadhq.io/platform/workato Data: https://www.launchpadhq.io/data/v1/tools/workato.json Category: Workflow & Automation Company: Workato Website: https://workato.com Cost tier: ENTERPRISE Difficulty: INTERMEDIATE Lifecycle: active Primary use: Enterprise integration and automation platform Free tier: no (No free tier; free trial available; enterprise pricing) Starting price: Custom Lowest paid plan: Enterprise, Custom Plans: - Enterprise: Custom Alternatives: Tray.ai, Microsoft Power Automate Guide: https://www.launchpadhq.io/platform/workato/guide Data as of: not dated Sources: - https://workato.com ### Writesonic / Chatsonic Canonical: https://www.launchpadhq.io/platform/writesonic-chatsonic Data: https://www.launchpadhq.io/data/v1/tools/writesonic-chatsonic.json Category: Text & Conversational AI Company: Writesonic, Inc. Website: https://writesonic.com Cost tier: PAID Difficulty: BEGINNER Lifecycle: active Description: Writesonic is an SEO and GEO (AI-search visibility) tool: it tracks whether ChatGPT, Gemini, and Google AI Overviews mention your brand, then offers AI articles, site audits, and agent workflows to act on the gaps. Chatsonic, the old chatbot, is no longer what it sells. Official pricing (writesonic.com/pricing, billed annually at the listed monthly rates): Starter $79/mo, Basic $199/mo, Growth $399/mo, Enterprise custom. Annual billing saves 20%. No $16 plan on the pricing table. Start free, no credit card required. Primary use: SEO + GEO (AI-search visibility) platform Free tier: no (No free plan. The pricing page offers a free trial with no credit card needed.) Starting price: $79/month Lowest paid plan: Starter, $79/month Plans: - Starter: $79/month (Billed annually; 1 user, 1 project) - Basic: $199/month (Billed annually; 2 users, 1 project) - Growth: $399/month (Billed annually; 3 users, 2 projects) - Enterprise: Custom (Custom users, projects and access control) Alternatives: Jasper, Copy.ai, ChatGPT, Notion AI Guide: https://www.launchpadhq.io/platform/writesonic-chatsonic/guide Data as of: 2026-10-25 Sources: - https://writesonic.com/pricing (checked 2026-10-14) ### Zapier Canonical: https://www.launchpadhq.io/platform/zapier Data: https://www.launchpadhq.io/data/v1/tools/zapier.json Category: Workflow & Automation Company: Zapier Website: https://zapier.com Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: No-code app-to-app automation with AI agents Tasks: workflow-automation Free tier: yes (Free: 100 tasks/mo, unlimited two-step Zaps, 15-minute update time) Starting price: Free, from $19.99/mo billed annually Lowest paid plan: Professional from, $19.99/mo billed annually Plans: - Free: $0 (100 tasks/mo) - Professional from: $19.99/mo billed annually ($29.99 monthly, 750 tasks) - Team from: $69/mo billed annually ($103.50 monthly, 2,000 tasks, 25 users) - Enterprise: Custom Alternatives: Make (Integromat), n8n, Microsoft Power Automate, Pabbly Connect Guide: https://www.launchpadhq.io/platform/zapier/guide Data as of: 2026-10-25 Sources: - https://zapier.com ### Zillow AI / Zestimate Canonical: https://www.launchpadhq.io/platform/zillow-ai-zestimate Data: https://www.launchpadhq.io/data/v1/tools/zillow-ai-zestimate.json Category: Finance & Real Estate AI Company: Zillow Website: https://zillow.com Cost tier: FREE Difficulty: INTERMEDIATE Lifecycle: active Primary use: AI property valuation and real estate insights Free tier: yes (Completely free for consumers; Zestimate on all properties) Starting price: Free, from See pricing Lowest paid plan: Premier Agent paid, See pricing Plans: - Free: Free (consumer) - Premier Agent paid: See pricing Guide: https://www.launchpadhq.io/platform/zillow-ai-zestimate/guide Data as of: not dated Sources: - https://zillow.com ### Zoom AI Companion Canonical: https://www.launchpadhq.io/platform/zoom-ai-companion Data: https://www.launchpadhq.io/data/v1/tools/zoom-ai-companion.json Category: AI Plugins for Business Software Company: Zoom Website: https://zoom.us Cost tier: FREEMIUM Difficulty: BEGINNER Lifecycle: active Primary use: AI meeting assistant in Zoom Tasks: meeting-notes Free tier: yes (Included free with most paid Zoom plans (no extra cost); available on Zoom Workplace) Starting price: See pricing Lowest paid plan: Included with paid Zoom plans, See pricing Plans: - Included with paid Zoom plans: See pricing Guide: https://www.launchpadhq.io/platform/zoom-ai-companion/guide Data as of: not dated Sources: - https://zoom.us ## Comparisons Canonical: https://www.launchpadhq.io/compare ### Free Otter.ai alternatives for meeting notes (2026) Canonical: https://www.launchpadhq.io/alternatives/free-otter-ai-alternatives Data: https://www.launchpadhq.io/data/v1/compare/free-otter-ai-alternatives.json As of: 2026-10-14 Free meeting note takers to try instead of Otter.ai, with each free plan, its limits and current pricing pulled from our catalog and dated. Otter's free Basic plan caps how many minutes it transcribes each month and how long each conversation can be. These tools also record or transcribe meetings and write up notes, and each has a free plan. The free plans differ in what they cap: transcripts, meetings with AI notes, AI credits or how much history you keep. #### Otter AI (baseline) Page: https://www.launchpadhq.io/platform/otter-ai Free tier: yes (Basic (free): 300 transcription minutes/mo, 30 min per conversation, 3 lifetime file imports) Starting price: Free, from $16.99/user/mo billed monthly Lowest paid plan: Pro, $16.99/user/mo billed monthly Source (Pricing): https://otter.ai/pricing, as of 2026-10-25 Source (Free plan): https://otter.ai/pricing, as of 2026-10-25 Source (Usage limits): https://otter.ai/pricing, as of 2026-10-25 #### Fathom (alternative) Page: https://www.launchpadhq.io/platform/fathom Pick this if you mostly meet on Zoom, Google Meet or Teams and want unlimited recordings and transcripts on the free plan. Free tier: yes (Free forever plan.) Starting price: Talk to sales Lowest paid plan: Enterprise, Talk to sales Source (Pricing): https://www.fathom.ai/pricing, as of 2026-10-14 Source (Free plan): https://www.fathom.ai/pricing, as of 2026-10-14 #### Tactiq (alternative) Page: https://www.launchpadhq.io/platform/tactiq Pick this if you would rather work from a live transcript in the browser than have a bot join your calls. Free tier: yes (Free forever, no card: 10 transcripts/mo and 5 AI credits/mo) Starting price: Free, from $8/user/mo billed annually Lowest paid plan: Pro, $8/user/mo billed annually Source (Pricing): https://tactiq.io/pricing, as of 2026-10-14 Source (Free plan): https://tactiq.io/pricing, as of 2026-10-25 Source (Usage limits): https://tactiq.io/pricing, as of 2026-10-25 #### Read AI (alternative) Page: https://www.launchpadhq.io/platform/read-ai Pick this if you want meeting reports with participant-level metrics such as talk time. Free tier: yes (Free, no card: 5 meeting transcripts and 30 measured meetings per month, meetings up to 1 hour, no audio or video uploads.) Starting price: Free, from $15 Lowest paid plan: Pro, $15/user/month Source (Pricing): https://www.read.ai/plans-pricing, as of 2026-10-26 Source (Free plan): https://www.read.ai/pricing, as of 2026-10-25 #### Granola (alternative) Page: https://www.launchpadhq.io/platform/granola Pick this if you like taking your own rough notes and want AI to fill them out from the meeting afterwards. Free tier: yes (Basic, $0 per user per month, with limited meeting history.) Starting price: Free, from $14 Lowest paid plan: Business, $14 Source (Pricing): https://www.granola.ai/pricing, as of 2026-10-14 Source (Free plan): https://www.granola.ai/pricing, as of 2026-10-14 #### tl;dv (alternative) Page: https://www.launchpadhq.io/platform/tl-dv Pick this if you want to turn parts of recorded calls into clips you can share. Free tier: yes (Free $0, no card: unlimited video recordings and transcription in 30+ languages; AI notes for 10 meetings and 10 Ask AI queries included) Starting price: Free, from $18/seat/mo billed annually Lowest paid plan: Pro, $18/seat/mo billed annually Source (Pricing): https://tldv.io/pricing/, as of 2026-10-25 Source (Free plan): https://tldv.io/pricing/, as of 2026-10-25 Source (Usage limits): https://tldv.io/pricing/, as of 2026-10-25 What you hit first: - Free plans cap different things. Otter caps minutes, Tactiq caps transcripts and AI credits, and tl;dv caps how many meetings get AI notes. Check which one matches how often you meet. - Some cap history instead: Granola's free plan keeps only a limited meeting history. - Let people know when a meeting is being recorded. Some of these tools can join a call as a visible participant, while others work without a bot, so nobody sees a recorder join. Q: Can I switch from Otter without losing my old transcripts? A: Yes, if you export them first. Otter's free Basic plan can export audio and text files, so save what you want to keep before you move. Q: Do these tools need a bot to join my meeting? A: Not always. Granola, Tactiq and tl;dv each say they work without a bot joining, and Fathom offers both options. ### Free Zapier alternatives: automation without the task bill Canonical: https://www.launchpadhq.io/alternatives/free-zapier-alternatives Data: https://www.launchpadhq.io/data/v1/compare/free-zapier-alternatives.json As of: 2026-10-25 Free and cheaper alternatives to Zapier for connecting apps, with each free plan, its limits and current pricing pulled from our catalog and dated. Zapier connects apps so one thing can trigger another, and it bills by the task. Each plan includes a monthly task allowance, and going over it costs more per task than the plan's base rate. These alternatives do the same job with different pricing, and each has a free way in. One of them is only free if you host it yourself, and that is labelled plainly below. #### Zapier (baseline) Page: https://www.launchpadhq.io/platform/zapier Free tier: yes (Free: 100 tasks/mo, unlimited two-step Zaps, 15-minute update time) Starting price: Free, from $19.99/mo billed annually Lowest paid plan: Professional from, $19.99/mo billed annually Source (Pricing): https://zapier.com/pricing, as of 2026-10-25 Source (Free plan): https://zapier.com/pricing, as of 2026-10-25 Source (Usage limits): https://zapier.com/pricing, as of 2026-10-25 #### Make (Integromat) (alternative) Page: https://www.launchpadhq.io/platform/make-integromat Pick this if you want a hosted tool with a visual builder and are happy to learn how it charges credits for each step. Free tier: yes (Free: 1,000 credits/mo (credits are Make's billing unit)) Starting price: Free, from See pricing Lowest paid plan: paid plans Core, Pro, Teams and Enterprise. Paid prices did not render on the official page, so none are quoted, See pricing Source (Pricing): https://www.make.com/en/pricing, as of 2026-10-25 Source (Free plan): https://www.make.com/en/pricing, as of 2026-10-25 Source (Usage limits): https://www.make.com/en/pricing, as of 2026-10-25 #### Activepieces (alternative) Page: https://www.launchpadhq.io/platform/activepieces Pick this if you want a simple hosted builder with an open-source core you could move to your own server later. Free tier: yes (Free forever, no card: 1,000 credits, unlimited flows, up to 5 users. The official page lists the Free reset as daily in its comparison table but monthly in its FAQ) Starting price: Free, from $20/mo flat Lowest paid plan: Plus, $20/mo flat Source (Pricing): https://www.activepieces.com/pricing, as of 2026-10-25 Source (Free plan): https://www.activepieces.com/pricing, as of 2026-10-25 Source (Usage limits): https://www.activepieces.com/pricing, as of 2026-10-25 #### n8n (alternative) Page: https://www.launchpadhq.io/platform/n8n Pick this only if you are comfortable running software on your own server, or you are willing to pay for the hosted cloud plan. Free option is self-hosted: the Community Edition is software you install and run yourself, on your own server. The cloud plans are paid. Free tier: yes (Free only if you run it yourself: the self-hosted Community edition is free with almost the full feature set. n8n Cloud has no free plan, only a trial with no card required.) Starting price: Free, from €20/month Lowest paid plan: Starter (Cloud), €20/month Source (Pricing): https://n8n.io/pricing/, as of 2026-10-25 Source (Free plan): https://docs.n8n.io/hosting/community-edition-features/, as of 2026-10-25 What you hit first: - Free plans cap how many tasks or credits you get each month, and the way each tool counts a single step differs. - Zapier's free plan checks for new trigger data less often than its paid plans, so an automation may not fire right away. - Self-hosting avoids per-task bills, but you install, configure, secure and update the software yourself. Q: Is self-hosting really free? A: The software is free to run as a Community Edition, but you supply the server and keep it configured, secure and updated yourself. If that sounds like a chore, choose a hosted plan. ### Free ChatGPT alternatives worth using in 2026 Canonical: https://www.launchpadhq.io/alternatives/free-chatgpt-alternatives Data: https://www.launchpadhq.io/data/v1/compare/free-chatgpt-alternatives.json As of: 2026-10-24 Free AI chat assistants to try alongside or instead of ChatGPT, with each free plan, its limits and current pricing pulled from our catalog and dated. ChatGPT is not the only AI chat assistant with a free plan. Each tool below has one you can use without paying. The free plans limit use in different ways. Some refresh an allowance every few hours, and some cap extras like uploads and image creation more tightly than plain chat. #### ChatGPT (baseline) Page: https://www.launchpadhq.io/platform/chatgpt Free tier: yes (Unlimited text chats with GPT-6 (abuse guardrails apply)) Starting price: Free, from $8/month Lowest paid plan: Go, $8/month Source (Pricing): https://learn.chatgpt.com/docs/pricing, as of 2026-10-24 Source (Free plan): https://chatgpt.com/pricing, as of 2026-10-24 #### Gemini (alternative) Page: https://www.launchpadhq.io/platform/gemini Pick this if you already have a Google account. The paid plans add Gemini inside apps like Gmail and Docs. Free tier: yes (Free with a Google Account: Gemini app with the Flash-Lite model, image generation and editing, Deep Research, Gemini Live, Canvas and Gems; 15 GB of cloud storage) Starting price: Free, from $4.99/month Lowest paid plan: Google AI Plus, $4.99/month Source (Pricing): https://gemini.google/subscriptions/, as of 2026-10-26 Source (Free plan): https://gemini.google/subscriptions/, as of 2026-10-26 #### Mistral Vibe (alternative) Page: https://www.launchpadhq.io/platform/mistral-vibe Pick this if you want an assistant from a European company for quick answers, web search and simple tasks. Free tier: yes (Free plan with limited messages, web searches and coding sessions.) Starting price: Free, from $14.99/month Lowest paid plan: Pro, $14.99/month Source (Pricing): https://mistral.ai/pricing, as of 2026-10-25 Source (Free plan): https://mistral.ai/pricing, as of 2026-10-25 #### Claude (alternative) Page: https://www.launchpadhq.io/platform/claude Pick this if you want to group chats into projects and have the assistant create files and run code on the free plan. Free tier: yes (Chat on web, desktop and mobile; search the web, create files and run code; memory across conversations; connect apps and tools; Artifacts; Claude Design, Slides and Docs; up to 5 Projects.) Starting price: Free, from $17/month Lowest paid plan: Pro, $17/month Source (Pricing): https://claude.com/pricing, as of 2026-10-26 Source (Free plan): https://claude.com/pricing, as of 2026-10-26 #### Microsoft Copilot (alternative) Page: https://www.launchpadhq.io/platform/microsoft-copilot Pick this if you browse with Microsoft Edge, which has Copilot built in, or already pay for a Microsoft subscription that includes Copilot in apps like Word and Excel. Free tier: yes (Copilot has a free tier: the paid Microsoft 365 plans are described as giving higher Copilot usage than free. The page does not list the free limits.) Starting price: $9.99/month Lowest paid plan: Microsoft 365 Personal, $9.99/month Source (Pricing): https://www.microsoft.com/en-us/copilot/solutions/individuals, as of 2026-10-26 Source (Free plan): https://www.microsoft.com/en-us/copilot/solutions/individuals, as of 2026-10-26 What you hit first: - Usage windows come first on some plans. Gemini and Claude refresh free usage every five hours, and Gemini also has a weekly cap. - Free plans usually leave out each company's most capable models, which are kept for paid plans. - Extras are capped separately. ChatGPT's free plan, for example, limits uploads, image creation, voice and deep research even though everyday text chat is not capped. Q: Which free assistant is the smartest? A: There is no single answer, and the vendors change their models often. Try the same real question in two or three of them and keep the one whose answers you trust most. Q: Where can I compare the underlying models? A: Our Claude vs GPT page compares the flagship and budget models from each family, with sourced API details. ### HeyGen vs Synthesia: which AI avatar video tool should you start with? Canonical: https://www.launchpadhq.io/compare/heygen-vs-synthesia Data: https://www.launchpadhq.io/data/v1/compare/heygen-vs-synthesia.json As of: 2026-10-25 HeyGen and Synthesia side by side for beginners: free plans, limits and current pricing pulled from our catalog and dated, plus who should pick which. HeyGen and Synthesia both turn a script into a video of a talking AI presenter, so you can make explainers without filming yourself. They overlap a lot. The differences that matter to a beginner are how much the free plan lets you make and which kind of video each one is built around. #### HeyGen (contender) Page: https://www.launchpadhq.io/platform/heygen Pick this if you make marketing or creator videos and want a large library of ready-made presenters. Free tier: yes (Free $0: 3 videos/mo up to 1 min, watermarked, Avatar IV access, 500+ stock avatars, 1 custom video avatar) Starting price: Free, from $29/mo Lowest paid plan: Creator, $29/mo Source (Pricing): https://www.heygen.com/pricing, as of 2026-10-25 Source (Free plan): https://www.heygen.com/pricing, as of 2026-10-25 Source (Usage limits): https://www.heygen.com/pricing, as of 2026-10-25 #### Synthesia (contender) Page: https://www.launchpadhq.io/platform/synthesia Pick this if you make training, compliance or how-to videos for a business and want a shared team workspace. Free tier: yes (Basic $0: 10 min of video, 3 survey responses and 500 credits every month) Starting price: Free, from $29/mo billed monthly Lowest paid plan: Starter, $29/mo billed monthly Source (Pricing): https://www.synthesia.io/pricing, as of 2026-10-25 Source (Free plan): https://www.synthesia.io/pricing, as of 2026-10-25 Source (Usage limits): https://www.synthesia.io/pricing, as of 2026-10-25 What you hit first: - Free plans cap how much video you can make each month: HeyGen by the number and length of videos, Synthesia by total minutes. - Free videos carry a watermark on both tools. HeyGen's free plan does include one custom video avatar. - Making an avatar of a real person requires that person's consent. Both tools have written rules about this, so read them before you start. Q: Can I try both for free? A: Yes. HeyGen has a free plan and Synthesia has a free Basic plan, both with watermarked videos. The current allowances are in the rows above, with the date they were checked. ### Bolt.new vs Lovable: build your first app without coding Canonical: https://www.launchpadhq.io/compare/bolt-new-vs-lovable Data: https://www.launchpadhq.io/data/v1/compare/bolt-new-vs-lovable.json As of: 2026-10-25 Bolt.new and Lovable side by side for first-time app builders: free plans, limits and current pricing pulled from our catalog and dated. Bolt.new and Lovable both build a working web app from a plain-language description. You describe what you want, they write the code, and you keep refining by chatting. Both run in the browser, and both meter AI use with a daily and a monthly allowance: Bolt counts tokens and Lovable counts credits. #### Bolt.new (contender) Page: https://www.launchpadhq.io/platform/bolt-new Pick this if you want to start from a Figma design or an existing GitHub project, not just a prompt. Free tier: yes (Free: 1M tokens/month with a 300K daily limit, Bolt branding, 10MB upload limit) Starting price: Free, from $25/mo billed monthly Lowest paid plan: Pro, $25/mo billed monthly Source (Pricing): https://bolt.new/pricing, as of 2026-10-25 Source (Free plan): https://bolt.new/pricing, as of 2026-10-25 Source (Usage limits): https://bolt.new/pricing, as of 2026-10-25 #### Lovable (contender) Page: https://www.launchpadhq.io/platform/lovable Pick this if you want a chat-first builder whose free plan tops up a few build credits every day. Free tier: yes (Free $0, no card: 5 build credits per day (up to 30/month), 20 Cloud credits/month, up to 5 lovable.app domains) Starting price: Free, from $25/mo Lowest paid plan: Pro, $25/mo Source (Pricing): https://lovable.dev/pricing, as of 2026-10-25 Source (Free plan): https://lovable.dev/pricing, as of 2026-10-25 Source (Usage limits): https://lovable.dev/pricing, as of 2026-10-25 What you hit first: - The daily and monthly allowances are the first wall. Every change you ask for draws on them, and Lovable charges more for bigger changes than for small, focused edits. - Bolt says the larger the project, the more tokens each message uses. - Connecting your own custom domain is a paid feature on both. Q: Do I need to know how to code? A: Not to get started. Both build from a description you type in plain language. Q: Can I take my app elsewhere later? A: Yes. Both can connect a project to GitHub, so the code sits in a repository you control. Also consider: - v0 (https://www.launchpadhq.io/platform/v0): starting price Free, from $30/user/mo; free tier yes; data as of 2026-10-14 ### Cursor vs GitHub Copilot: which AI coding assistant for a beginner? Canonical: https://www.launchpadhq.io/compare/cursor-vs-github-copilot Data: https://www.launchpadhq.io/data/v1/compare/cursor-vs-github-copilot.json As of: 2026-10-25 Cursor and GitHub Copilot side by side for new coders: free plans, limits and current pricing pulled from our catalog and dated, plus who should pick which. To be upfront: both of these are for people ready to work in a code editor. If you want an app without touching code, an app builder that works from a chat, like Bolt.new or Lovable, is a gentler first step. Cursor is a code editor with AI built in. GitHub Copilot is an assistant that plugs into editors you may already use, such as VS Code or JetBrains IDEs. #### Cursor (contender) Page: https://www.launchpadhq.io/platform/cursor Pick this if you are happy to switch to a new editor in exchange for AI agents that can change several files at once. Free tier: yes (Free $0: limited AI usage in Cursor) Starting price: Free, from $8/mo Lowest paid plan: Plus, $8/mo Source (Pricing): https://cursor.com/pricing, as of 2026-10-25 Source (Free plan): https://cursor.com/pricing, as of 2026-10-25 Source (Usage limits): https://cursor.com/pricing, as of 2026-10-25 #### GitHub Copilot (contender) Page: https://www.launchpadhq.io/platform/github-copilot Pick this if you already use VS Code or another supported editor and want to add AI without changing tools. Free tier: yes (Free plan: limited chat and agent usage; up to 2,000 completions plus entry-level models per month) Starting price: Free, from $10/mo Lowest paid plan: Pro, $10/mo Source (Pricing): https://github.com/features/copilot/plans, as of 2026-10-25 Source (Free plan): https://github.com/features/copilot/plans, as of 2026-10-25 Source (Usage limits): https://github.com/features/copilot/plans, as of 2026-10-25 What you hit first: - Both free plans are limited. Copilot Free caps completions and chat and agent use each month, and Cursor's usage pool refreshes weekly. - Copilot Free comes with entry-level models, and on Cursor third-party models are billed separately by usage. - GitHub's own guidance is to review and test the code Copilot writes before you rely on it. The same habit applies to any AI coding tool. Q: Which is easier for a complete beginner? A: If you already use VS Code or another supported editor, Copilot adds AI to the editor you know. Cursor is a separate desktop app you download, with the AI built in. Q: Are students eligible for anything extra? A: GitHub gives free Copilot access to verified students and teachers through GitHub Education. Cursor points students to promotions at its campus and online events. Also consider: - Claude Code (https://www.launchpadhq.io/platform/claude-code): starting price Pro $17/mo with annual ($20 billed monthly); free tier no; data as of 2026-10-26 - Devin Desktop (https://www.launchpadhq.io/platform/windsurf): starting price Free, from $15-20/mo; free tier yes; data as of 2026-10-20 ## Guides Canonical: https://www.launchpadhq.io/tutorials ### Getting Started with Acrobat AI Assistant Canonical: https://www.launchpadhq.io/platform/adobe-acrobat-ai-assistant/guide Data: https://www.launchpadhq.io/data/v1/guides/adobe-acrobat-ai-assistant.json Last reviewed: 2026-07-27 Tool: Adobe Acrobat AI Assistant (https://www.launchpadhq.io/platform/adobe-acrobat-ai-assistant): starting price $22.99/mo; free tier yes; data as of not dated Ask questions of a long PDF and get cited answers — inside the reader you probably already have. Open a document in Acrobat or Reader and ask about it in plain language. It answers, with citations pointing to the passage each answer came from; you click a citation to check it, then ask the next question. What it is: - A question-and-answer layer over PDFs and other documents, built into software most people already have installed. - Available as a paid add-on to both the free Reader and paid Acrobat subscriptions. - Grounded in your documents, with citations back to source passages rather than free-form generation. - Able to work across multiple documents at once, which is where it beats reading them individually. Before you start: - This is a paid add-on. It is not included with the free Reader or with a standard Acrobat subscription — it is billed separately on top. - Pricing has changed since launch and is bundled differently across several Adobe tiers. Check Adobe's current plans page for the figure that applies to you rather than relying on a quoted price from an article. - It works on scanned documents by running text recognition first, but quality drops on poor scans. A bad photograph of a page produces bad answers. - It answers about your documents, not about the world. It is not a general assistant and does not try to be. Security: - This is the reassuring page in a category where most are not. Adobe states that it does not train its generative models on customer document content, and that it contractually prohibits the third-party models it uses from training on that content either. Document content is processed to answer your question and then discarded. Security: What that does and does not mean: - Your documents are still transmitted for processing — "not used for training" is not the same as "never leaves your device". - Organisations with stricter requirements can arrange processing inside their own cloud tenancy, which removes the residency objection but costs more. - Administrators can disable the feature, though the granularity of that control varies by plan. - Genuinely regulated material still deserves an explicit policy decision rather than an assumption. Security: - Worth noticing why this page reads calmer than its neighbours: the vendor made a narrow, checkable commitment and stated it plainly. That is what a good data posture looks like written down, and it is a useful benchmark to hold other tools against when their answer is longer and less specific. Best at: - Finding specific information in a long document quickly - Summarising a dense report before deciding whether to read it - Comparing several documents that cover the same ground - Answering with citations you can jump to and verify Okay at: - Poor-quality scans, where text recognition limits everything downstream - Complex tables and figures, which it reads inconsistently - Documents in languages outside its supported set Avoid: - Relying on a summary of a contract or legal document without reading the relevant clauses yourself - Treating it as a general assistant — it answers about your files, nothing else - Extremely long documents where an important detail could sit outside what it weighted Starter: Ask what a document is before asking what it says Orientation questions produce better follow-ups than summary requests, and the answer tells you whether it understood the document at all. If it misidentifies the document type, treat everything after that with suspicion. Prompt: What kind of document is this, who produced it, and who is it addressed to? Then tell me what it's trying to get the reader to do. Starter: Ask a question you already know the answer to Asks the assistant for one fact you can already see in the document, then compares its answer and citation against the page. The fastest calibration available. You learn whether to trust it on this document in about thirty seconds. Choose something specific and checkable — a date, a number, a named party. Starter: Follow one citation to its source Clicks the citation attached to an answer to jump to the passage it was drawn from, and reads that passage in full rather than the excerpt. Citations are the feature that makes this checkable rather than merely convenient. Following one turns the habit on. Follow the citation behind the claim you found most useful. Useful and wrong is the expensive combination. Starter: Compare two documents on the same question Answers across a set of files rather than one. The multi-document case is where it saves the most time and where people least expect it to work. Ask specifically where the documents disagree. Agreement is easy; contradictions are what you needed to find. Pitfalls: - It is a separate paid add-on. Neither the free Reader nor a standard Acrobat subscription includes it. Check what your plan actually covers before assuming you have access. - Scan quality caps everything. Text recognition runs before the AI does, so a poor scan produces poor answers with no warning that the input was the problem. - Do not accept a summary of a contract. Summarisation is a starting point for legal or financial documents, never a substitute for reading the clauses that bind you. - Not used for training is not the same as never leaves your device. Documents are still transmitted for processing. For genuinely regulated material, that needs a deliberate decision. - Tables and figures are read inconsistently. Numbers pulled from complex layouts are the most likely thing to be wrong. Verify any figure you intend to act on. Where to next: - Other document and PDF tools: https://www.launchpadhq.io/discover?category=document-pdf-processing - Research and academic tools: https://www.launchpadhq.io/discover?category=research-academic-tools ### Getting Started with Activepieces Canonical: https://www.launchpadhq.io/platform/activepieces/guide Data: https://www.launchpadhq.io/data/v1/guides/activepieces.json Last reviewed: 2026-10-14 Tool: Activepieces (https://www.launchpadhq.io/platform/activepieces): starting price Free, from $20/mo flat; free tier yes; data as of 2026-10-25 Automation built from public code — around 760 integrations anyone can read, and a free cloud allowance that resets every morning. You work on a canvas in your browser. You pick a trigger — the "when," the event that starts an automation, like a form being filled in — then add steps under it, choosing an app and what it should do. You run the whole thing once with real data and read what each step received before switching it on. What it is: - An automation tool: you build flows that connect apps you already use — Gmail, Slack, Sheets, Notion and around 760 others. Activepieces calls each of those integrations a piece, and the code for every one of them is published for anyone to read. - The thing that separates it from Zapier or Make is its licence. The main body of Activepieces is released under the MIT licence, about as permissive as software gets: you may run it, change it and build on it. One "enterprise" folder inside the same code is carved out of that and governed by a commercial licence instead. So self-hosting it (running it on your own computer or server rather than the company's) is genuinely free, but not every part of the product comes with you. - Cloud billing counts runs, not steps: one flow firing is one credit — the platform's unit of spend — however many steps it contains. AI steps are the exception and cost extra on top. - It has grown past plumbing. The same workspace now carries AI agents, spreadsheet-style tables, and a chat box that drafts a flow from a plain description, which the docs still mark Beta. Before you start: - The free cloud plan is a plan, not a trial. No card, unlimited flows, one user, and a daily credit allowance — a fresh batch each morning rather than a monthly bucket, so running out today never eats into tomorrow. - Run out and nothing breaks. The pricing page says your flows stay switched on and runs past the quota simply don't go through until credits refresh the next day. A mistake on this platform costs you a day, not a month. - You would start paying when the free daily allowance stops being enough, or when a second person needs to be in the workspace. Plus and Team are flat monthly plans with much larger credit allowances, and runs beyond the allowance are billed per credit. AI steps cost more credits the more capable the model is, and a paid plan lets you plug in your own AI account to bring every AI step back down to the cost of an ordinary run. - Worth checking before you commit anything real: the Activepieces cloud currently runs in one data region, Germany. If your work has to sit somewhere else, running it yourself is the answer rather than the cloud. - First step: sign up at activepieces.com, open the template library rather than a blank canvas, and run somebody else's finished flow once before building your own. Security: - Where your connected logins physically sit is a decision you actually get to make here, and it is worth making on purpose. Its security docs say credentials are held under 256-bit encryption with no interface to read them back out, and that sensitive values are stripped out of run logs. Choose to run it on your own machine instead and every one of those protections becomes something you are responsible for. Security: Worth knowing before you switch a flow on: - The free plan is one person. There is no way to let a colleague in without handing over your own login — adding people is what the paid plans are for. - Templates and pieces are written in the open by a lot of different contributors. That is exactly why the library is so wide, and exactly why you read what a template's steps do before you connect a real account to it. - If you self-host, a new version ships every Monday. Installing it is now your job, not theirs. Best at: - Flows with a lot of steps, since the credit cost does not move with the step count - Reaching an app through a piece somebody outside the company wrote and published - Keeping an automation running on hardware you own, indefinitely, without asking anyone's permission Okay at: - Working as a team on the free plan, which is one person with no separate projects - Cloud work under a data-location requirement, while Germany is the only region on offer - Staying entirely clear of technical vocabulary — the cloud side is click-based, but the self-hosting half of the story is written for people comfortable running a server Avoid: - Frontier-model AI steps on the free plan — each one costs many times an ordinary run, so a handful of runs uses up the whole day's allowance - Treating the free allowance as capacity for a backlog: credits refill daily rather than accumulating, so three hundred queued items cannot clear in one go - Hosting Activepieces for other people out of the enterprise folder — that part's licence forbids copying, distributing, sublicensing or selling it, and production use needs a paid seat count Starter: When a form is submitted → then file it and tell you about it Activepieces hosts the form itself, drops each response into a table or a spreadsheet, and sends you a message about it. Three steps, one credit. On Zapier that same flow spends a separate task per step out of a small monthly allowance; n8n prices by the run too, but has no permanent free cloud plan to do it on. Point it at a test spreadsheet and fire it by hand a few times first. The run view shows what each step received. Starter: Every weekday at 8am → then have an AI step summarise yesterday and post it A scheduled trigger gathers yesterday's messages or rows, an AI step condenses them, and the summary lands in your chat app or inbox. Activepieces posts what an AI step will cost before you run it, so you can weigh a daily habit against the daily allowance in your head. Zapier sells its AI features as separate products with their own plans, and n8n expects you to bring your own model account rather than quoting you a rate. Start on the fast model. If the summary is good enough there, you have just bought yourself ten times the runs. Starter: When the free daily allowance stops being enough → then move the same flow onto your own machine One install command brings up the free Community Edition on a server you control, where runs and users are unlimited and there is no credit system at all. Because the core is MIT-licensed, running it yourself is permitted even for paying clients or inside something you sell. n8n's self-hosted edition is free too, but its licence specifically forbids hosting it commercially for other people; Zapier has no self-hosted edition to move to at all. It wants a machine that stays on and Docker Compose. Do this after the cloud version has taught you what a flow is, not before. Starter: When a niche app you use does something → then start an automation from it Search the piece library for the app before assuming you are stuck; if a piece exists, its events work as a trigger like any other. Each piece is MIT-licensed and published as its own package, most of them written by people outside the company. On Zapier a missing integration is a request you file with a company; n8n is open too, but ships its own connectors under a licence that stops you reusing them freely. If the piece you want does not exist, the framework for writing one is public — a project rather than an afternoon, but not a closed door. Pitfalls: - What you self-host is the automation core, not the cloud product minus the bill. Community Edition leaves out projects, single sign-on, audit logs and Git Sync, and those last two sit in the paid part on cloud too. For one person automating their own work that is no loss; for a company it is the entire reason the paid tiers exist. - AI steps are the only thing that breaks the one-credit-per-run rule. Price a single run before you schedule it hourly. - The free daily figure has moved before. Community posts in mid-2026 report hitting the cap sooner than they used to, so treat today's allowance as a snapshot rather than a promise, and check the pricing page. - The chat box that builds a flow from a description is still marked Beta in the docs. Fine for a first draft; read every step it wrote before you switch it on. Where to next: - More ways to connect your apps together: https://www.launchpadhq.io/discover?category=workflow-automation - Other tools you can run on your own machine: https://www.launchpadhq.io/discover?category=local-open-source-ai ### Getting Started with Adobe Firefly Canonical: https://www.launchpadhq.io/platform/adobe-firefly/guide Data: https://www.launchpadhq.io/data/v1/guides/adobe-firefly.json Last reviewed: 2026-07-13 Tool: Adobe Firefly (https://www.launchpadhq.io/platform/adobe-firefly): starting price Free, from $199.99/mo; free tier yes; data as of not dated Adobe's image generator, built to be commercially safe — trained only on licensed and public-domain content, with legal backing for paying users. You describe the image you want and it generates several options in seconds. You pick one and refine it — change the style, the shape of the frame, ask for more variations — until you have something worth keeping. What it is: - Firefly is Adobe's generative AI for images (and increasingly video, audio, and vector art). You describe what you want and it generates it — as a standalone web app, and baked into Adobe tools like Photoshop (Generative Fill) and Illustrator. What sets it apart isn't image quality — Midjourney generally beats it there — it's legal safety: Firefly was trained only on content Adobe has the rights to, which is why Adobe can stand behind commercial use in a way most generators can't. It's aimed at marketers, designers, and businesses who need images they can actually use without a copyright headache. Every Firefly image also carries Content Credentials — invisible provenance metadata marking it as AI-generated, part of Adobe's transparency push. Before you start: - The free plan needs only a free Adobe ID (no card) and gives a small monthly batch of credits — enough to try it. Outputs have no watermark and are licensed for commercial use, but the free tier is not covered by Adobe's legal indemnity (see security). - Standalone paid plans add more monthly credits, and if you already pay for Creative Cloud, Firefly credits are usually bundled in. - The headline reason to choose Firefly over a flashier generator is the commercial-safety story below — understand exactly what it does and doesn't cover before you rely on it. Security: - Firefly's whole pitch is commercial safety: it's trained only on licensed content (Adobe Stock), openly-licensed, and public-domain material — not scraped from the open web — and Adobe does not train Firefly on your prompts or uploads. That clean training data is why Adobe can offer the legal backing below, which most generators can't. Security: The indemnity is real — and narrower than the marketing: - On paid plans, Adobe offers IP indemnification: if a Firefly-generated image is hit with a copyright claim, Adobe will defend you and cover it. Free-tier outputs are NOT covered. - It only backs clean, on-policy generations. The moment your prompt names a real brand, character, or person, you've stepped outside the safe zone — "trained on licensed data" does not make a trademark-infringing prompt safe. - It doesn't extend to third-party integrations/API wrappers or to reference images you upload — the coverage is for standard Firefly outputs on a paid plan. Security: - Separate but important: indemnity protects you from infringing someone else's copyright; it does not give you a copyright in your image. As with any generator, a pure prompt-to-image result may not be copyrightable — so you can use it commercially, but you may not be able to stop others from using something similar. Best at: - Commercially safe images for ads, marketing, and client work - Generative Fill / Expand inside Photoshop and Illustrator - Business use where legal defensibility matters more than raw artistry - Clean, on-brand product and marketing visuals Okay at: - Peak artistic quality and stylistic range (Midjourney leads here) - Highly stylized or experimental art (it's tuned for commercial-safe, not wild) - Photorealistic close-up faces and hands (improving, still imperfect) Avoid: - Relying on the free tier for indemnified commercial work (it isn't covered) - Prompts naming real brands, characters, or people (voids the safe zone) - Assuming 'commercially safe' means you own a copyright in the output Starter: A clean commercial image Marketing-style, no named IP — exactly the kind of prompt Firefly is built for and that stays inside the safe zone. Change the product and lighting; keep it generic, no brands or logos. Prompt: A bright, minimal flat-lay of a reusable water bottle on a pale wooden table with soft morning light, lots of empty space for text. Photorealistic. Starter: Try a few styles Shows style control and the kind of reusable brand asset Firefly does well. Ask for the same subject in "watercolor" or "3D render" to see the range. Prompt: A friendly cartoon fox mascot waving, simple vector style, bold flat colors, on a transparent background. Starter: Extend an image (in Photoshop) If you have Photoshop, open a photo and use Generative Fill/Expand to widen the frame or add a matching background. Firefly's integration into real design tools is its practical superpower, not just the standalone app. No Photoshop? The web app's expand feature does the same thing. Pitfalls: - The free tier isn't indemnified. The legal safety net that's Firefly's whole selling point only applies to paid plans. - Naming real IP breaks the safe zone. Clean prompts are protected; "make me a [famous character]" is not. - Commercial-safe is not the same as you-own-the-copyright. You can use it; you may not be able to stop others copying it. Where to next: - Image Generation & Editing: https://www.launchpadhq.io/discover?category=image-generation-editing - Video Creation & Editing: https://www.launchpadhq.io/discover?category=video-creation-editing ### Getting Started with AI21 Jamba Canonical: https://www.launchpadhq.io/platform/ai21-labs-jamba/guide Data: https://www.launchpadhq.io/data/v1/guides/ai21-labs-jamba.json Last reviewed: 2026-09-13 Tool: AI21 Labs (Jamba) (https://www.launchpadhq.io/platform/ai21-labs-jamba): starting price Free, from Custom; free tier yes; data as of not dated AI21's Jamba files are free to download and run yourself — which is just as well, because the company's own sign-up page no longer opens. You pick one of the published Jamba files by size, download it, and open it in an app on your computer that can load it. After that you type into that app's message box and read the reply there, and you swap in a different file to try a bigger or smaller one. What it is: - Jamba is a family of open-weight AI models from AI21 Labs — open-weight meaning the model file itself is published, so you can run it yourself. A model is the AI "brain" that does the actual thinking, and here you get the brain rather than a subscription to somebody else's copy of it. - Its distinguishing design is that Jamba is a hybrid: it mixes the usual transformer machinery with a different technique called Mamba, and AI21 built it that way to keep very long inputs cheap to process. - Where it sits for a beginner: this is the model you pick when you want a long-document reader running on your own hardware, under a licence you can read, rather than a chatbot you sign into. Before you start: - The front door is shut, and this is the first thing to know. AI21's sign-up page, its playground and its key console all sit on studio.ai21.com, and every one of those addresses redirects to AI21's marketing homepage as of our last check — including the "Start Now for FREE" button on AI21's own pricing page. Whether people who already hold an AI21 account still reach a console some other way is not something we can see from outside; AI21's own service address does still answer requests, so it may well be running for existing customers, but with the sign-up page closed there is currently no visible way for a new reader to get a key. AI21's homepage now advertises two enterprise products, neither of them Jamba, and asks you to contact sales. The way in is not to sign up; it is to download the files from AI21's Hugging Face account, which needs no AI21 account at all. - You will not be asked for money or a card, because you are downloading a file rather than buying access to a service. AI21's pricing page and its documentation currently describe its free-credit offer with different time limits; since the sign-up button behind both leads to the homepage, treat neither as something you can act on. The only route where a bill starts is going through a cloud provider's own account, and that is a detour rather than the normal path. - The machine you already have is the constraint here, rather than any plan or allowance. - You will be doing this inside a general-purpose local-AI app, not an AI21 one, because AI21 does not ship a desktop app. If you have never loaded a model onto your own computer before, that skill is the actual prerequisite here — pick it up on a tool built for it first and come back; Ollama and the rest of the Local & Open-Source AI category exist for exactly that step. - Which file you download decides your rights, and the two halves of the family differ. The newer Jamba2 files are Apache 2.0 — permissive, commercial use included. The older Jamba Mini and Large 1.7 files are under AI21's own Jamba Open Model License (last updated 25 August 2024), which is royalty-free and revocable, and which cuts off commercial use once you or your affiliates pass fifty million US dollars in annual revenue. Jamba Reasoning 3B is listed as Apache 2.0, though its model card also names AI21's own Jamba licence — worth reading the file's own licence text before you build anything commercial on it. Getting set up safely (Download only from `huggingface.co/ai21labs` — that is AI21's own account, and it is the only source where the file you get is the one AI21 published. There are dozens of Jamba re-uploads on Hugging Face under other people's usernames; some are legitimate community work and some are modified, and from the outside they look alike.): - Check the file size against the free space and spare memory on your computer before you start the download, not after. These are multi-gigabyte files and the sizes differ enormously between the small and large members of the family. - Understand who made the small-file version you are about to use. AI21's own local-inference page states that it "does not distribute or support GGUF builds and cannot verify the accuracy of third-party conversions" — GGUF being the repackaged format most desktop apps need. The one exception is Jamba Reasoning 3B, where AI21 publishes that repackaged build itself; for the Jamba2 files, every such build on Hugging Face today comes from someone other than AI21. - One reply back is your "it worked" checkpoint. If the app loads the file and answers a short question, the setup is fine and everything after that is a question of which file, not whether. Security: - Nothing you type reaches AI21 when you run a Jamba file on your own machine, so the advice about never pasting anything sensitive into a chatbot simply does not apply to the normal way of using this. What replaces it is a supply-chain question: you are installing a multi-gigabyte file that someone built, and for most Jamba builds that someone is not AI21. Security: If you do route anything through AI21's own service: - AI21's privacy policy, last updated 2 August 2026, states that it does not use your inputs or outputs to train its foundational or generally available AI models — with carve-outs for anything separately agreed in writing, or needed to operate the service. - AI21 Labs Ltd. names itself the data controller and gives its principal place of business as Tel Aviv, Israel. If where your data is handled matters to you, that is the answer. - AI21's website terms of service still carry a "Last Updated" date of 19 November 2024 and say plainly that they do not cover the purchase or licensing of AI21's actual products. So the document that governs your use of a model is the licence attached to the specific file you downloaded, not the terms on the website. Best at: - Reading one enormous input in a single pass — AI21 quotes a 256K context window (how much text the model can hold in mind at once) on every published size, including the smallest - Running on hardware you already own, without a subscription, a key or a monthly ceiling - Nine languages AI21 names outright — English, Spanish, French, Portuguese, Italian, Dutch, German, Arabic and Hebrew Okay at: - General chat quality, which is a function of which size you can actually run - Being served for you by somebody else — the managed options are thin and behind the current release Avoid: - Amazon Bedrock as your route to "current Jamba" — Bedrock's AI21 listing is Jamba 1.5 Large and Jamba 1.5 Mini, and AI21's own deprecation table gives 2025-05-06 as the date it retired 1.5 from its platform. You would be building on a version its maker stopped running over a year ago. - Betting a long-lived project on the line continuing to ship. Nothing on AI21's Hugging Face account has been updated since February 2026 — which is a reason to treat the line as a set of files you have already downloaded rather than a roadmap you are joining. - Reaching for it as a general chatbot replacement. You would be comparing a small file on your own laptop against a frontier cloud model, and it will lose that comparison; what you came for is what it does with very long inputs on modest hardware. Starter: Start with the 3B file, not Mini Downloads AI21-Jamba2-3B — the size AI21 describes as runnable "right on their own devices, including iPhones, Androids, Macs, and PCs" — and gets a first reply back on ordinary hardware before you commit to anything larger. AI21 publishes Jamba2 at exactly two sizes, 3B and a 52B-total Mini, with nothing in between, so the choice is genuinely binary rather than a ladder you can climb one rung at a time. Qwen publishes a graded run of sizes from under a billion parameters upward. Starter: Give it one document that is far too long for anything else Load a book-length file or a year of accumulated notes into a single conversation and ask a question that can only be answered by having read all of it. The hybrid Mamba design exists to stop memory use ballooning as the input grows, which is why AI21 quotes the same 256K figure on the 3B file as on the 398B one — the long-context claim is not reserved for the size you cannot run. Meta's Llama 3.2 3B is a pure transformer whose long-context behaviour is not the stated goal of its architecture. Starter: Work in Hebrew or Arabic and compare Ask the same question twice, once in English and once in Hebrew or Arabic, and see how far the quality actually diverges. AI21's documentation names nine officially supported languages and Hebrew and Arabic are both on that list, which follows from AI21 Labs being a Tel Aviv company rather than being an afterthought bolted on for coverage. Mistral's small open models are marketed on a European language set that does not name Hebrew. Starter: Check who repackaged the file before you trust the app that opens it On Hugging Face, look at the account name on the small-file build your desktop app is about to pull, and prefer ai21labs where it exists. Which Jamba you can open in a desktop app depends on who repackaged it, and AI21 has said in writing that it will not vouch for most of them. Meta makes no such disclaimer about Llama conversions. Pitfalls: - Following AI21's own quick-start. It tells you to sign in to AI21 Studio, open Settings and create a key, at addresses that no longer open. AI21's documentation also says you can try Jamba in a playground in your browser, but the address it gives redirects to AI21's homepage too, so we could not confirm that any such page is reachable today. - Assuming "Jamba" names one thing. Jamba 1.5, 1.6 and 1.7, Jamba Reasoning 3B and Jamba2 are separate releases with different licences and different retirement dates, and AI21's live pricing page still quotes rates for versions its own deprecation table has already retired. Match the version in any guide you read against what is actually published today. - Downloading Jamba2 Mini because it sits next to Jamba2 3B in the same collection and sounds small. "Mini" is 52 billion total parameters — a download and a memory footprint sized for a server, next to a sibling built for a phone. - Taking the first "jamba" hit in a local app's model browser. Ollama's library has no official AI21 entry; the Jamba results there are community uploads under individual usernames, with the download counts of a hobby project rather than an official release. Where to next: - Local & Open-Source AI: https://www.launchpadhq.io/discover?category=local-open-source-ai - AI APIs & Developer Services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Aider Canonical: https://www.launchpadhq.io/platform/aider/guide Data: https://www.launchpadhq.io/data/v1/guides/aider.json Last reviewed: 2026-09-08 Tool: Aider (https://www.launchpadhq.io/platform/aider): starting price Free, from See pricing; free tier yes; data as of not dated Free software with no AI inside it. Aider works in your terminal, edits the files you name, and commits each change — using an AI service you sign up and pay for separately. You open a terminal — a text window where you type instructions instead of clicking — in your project's folder, start Aider, and name the files you want changed. You describe the change in plain English; it edits those files and saves the result. You read what it did, then ask for the next change. What it is: - Aider installs on your own computer and edits the files of a coding project you already have. Its code is public for anyone to inspect. - It contains no AI of its own. You point it at somebody else's AI service — a dozen or so are documented, from OpenAI and Anthropic to a model running on your own machine — using an API key, a password that identifies your app and that spends your money. - Development has slowed considerably. The last packaged release was February 2026, the last change to the code was May 2026, and the homepage still recommends AI models from early 2025. It works and it is not abandoned, but it is not moving quickly either. Before you start: - The gate here is your starting position, not the install. Aider expects a project already on your computer, kept in Git — which tracks a folder of code so you can see and undo every change — and you drive it by typing. Not there yet? The AI coding tools category on this site has applications you click instead, Cursor and GitHub Copilot among them. - Aider is free; the AI is not. You pay a model provider by usage, and the cost follows how much code is in the conversation rather than how many questions you ask. - A genuinely free route exists: Aider's documentation points to OpenRouter's free models, capped by daily usage, and to running a model on your own computer through Ollama. The same documentation warns that Aider works poorly with weaker models, so treat that route as a way to try it rather than the finished setup. - Choose your model yourself. The release you install predates the current lineup, so pointing it at a very new one can print "Unknown context window size and costs, using sane defaults" — harmless, but it means Aider cannot tell you what you are spending. Getting set up safely (Install only from aider.chat's own instructions. Aider is a Python program published as the package `aider-chat`. Lookalike package names are a known trick — copy the command from the docs, not from a blog post or a video.): - Run it the first time on work you have already saved, or on a copy you can throw away. Security: - Aider has no servers of its own. It runs on your computer and sends your code to whichever AI service you supplied a key for, which means the privacy question here is one you answer rather than one the vendor answers for you. Pointed at a cloud provider, your code goes to that company under that company's terms. Pointed at a model running on your own machine, nothing leaves it. Security: Worth settling before you point it at anything real: - It sends more than the files you named — it also builds and sends a compact map of the whole project. An .aiderignore file and the --subtree-only option limit that. - Its own usage analytics are opt-in and, its documentation states, never include your code, prompts, chats or keys. aider --analytics-disable ends them permanently. Best at: - Changes to files you can name - Working inside a project that already exists, rather than starting one from nothing - Leaving a reviewable trail — each edit is its own commit, with a written message Okay at: - Large codebases, where its map of the project crowds out the code you want it reading - Telling you what it is spending, which it can only do for models it has cost data for Avoid: - Running it over work you have not saved — it commits as it goes, and that history is your undo button - Learning to code with it, because it writes and commits in one motion and the review it hands back assumes you can read a diff Starter: Ask before you let it edit Aider commits its edits as it makes them, so nothing sits in your editor waiting to be accepted the way Cursor's changes do. /ask is the one mode where nothing is written. Stay in /ask for the whole plan, then say "go ahead". Prompt: /ask What does this project do, and which files would need to change to add a "remember me" option to the login? Starter: Hand it one change, in files you chose It only edits files you have added to the chat, and its own tips say extra files "distract and confuse" the model and raise your bill. Claude Code goes and finds its own files; here the choice is yours, and it is your main lever on both quality and cost. /drop a file when it is done and /add the next one. Keeping the set small is the technique. Prompt: Add validation to this form so an empty name is rejected, and match the error style already used in this file. Starter: Undo something it did /undo reverses the last commit only if Aider made it, and the "(aider)" it writes into the commit author is what makes that distinction possible. Worth doing once on purpose, before you need it. It steps back one commit, not back to the start. Prompt: /undo Starter: Leave it an instruction inside your code Started with --watch-files, Aider watches your files and acts on any comment ending in AI!, then goes back to waiting — so you steer it from whatever editor you already have open, not only from ones a plugin supports, as GitHub Copilot does. End the comment with AI? instead and it answers rather than edits. Prompt: # handle the case where this list is empty AI! Pitfalls: - Many files and a long session is the expensive shape. - It commits as it goes — a session you were not watching leaves a run of commits you now have to read. - A weak free model is the fastest route to a bad first impression, and Aider's docs say as much. Suspect the model before the tool. Where to next: - More AI coding tools: https://www.launchpadhq.io/discover?category=ai-coding-development - Run AI on your own computer: https://www.launchpadhq.io/discover?category=local-open-source-ai - Model APIs and keys: https://www.launchpadhq.io/discover?category=ai-apis-developer-services ### Getting Started with AIVA Canonical: https://www.launchpadhq.io/platform/aiva/guide Data: https://www.launchpadhq.io/data/v1/guides/aiva.json Last reviewed: 2026-09-15 Tool: AIVA (https://www.launchpadhq.io/platform/aiva): starting price Free, from €15/mo; free tier yes; data as of not dated An AI composer you steer with styles and settings rather than by describing a song. You pick a musical style from a list, set a few things like key, tempo and how long the track should be, then press create. AIVA hands back finished instrumental tracks you can play straight away. You keep the ones you like, adjust them in a built-in editor, and download the file. What it is: - AIVA is an AI composer for instrumental background music — soundtrack-style pieces for videos, games and podcasts, rather than songs with singing. - It works differently from the AI music tools that have had the most attention. Instead of describing a song in a sentence, you choose from a library of more than 250 preset styles and adjust musical settings. The one text box it does have takes a prompt (the message you type) that shapes the chord progression — the sequence of chords underneath a piece. - What comes back is editable. Tracks open in a built-in editor where you can mute parts, swap instruments and change individual notes, and they download as MIDI (the note-by-note score rather than a recording, which you can open in music software and rebuild). Among AI music tools that is unusual, and it is the main reason someone who reads music would pick AIVA. - AIVA has announced a newer engine called Lyra that does take a plain-language description of a whole piece. At the time of writing it is still being offered as a private beta you register for, so treat it as a direction of travel rather than something you can use on signing up. Before you start: - Free to use after signing up at aiva.ai, and the free plan has no trial clock attached. What it limits is downloads — a small number each month, short tracks only, MP3 or MIDI files only. No plan currently allows a track longer than five and a half minutes. Whether every creation mode is open on the free plan is not spelled out in AIVA's plan comparison, so you may need to check that once you are signed in. - Who owns what you make depends on which plan you are on, and that is the first thing to settle. On the free plan AIVA holds the copyright, use is non-commercial only, and you must credit AIVA with a link to aiva.ai. Standard (a paid plan, plus VAT, cheaper billed annually than month to month) drops the credit requirement and allows monetised use on YouTube, Twitch, TikTok and Instagram only — but the copyright still sits with AIVA. Only Pro, the higher paid tier, transfers the copyright to you and lifts the platform restriction. Read that split as the current position rather than a settled one: AIVA's licence agreement carries no effective date of its own and never says which plan maps to which licence, and the helpdesk article explaining the split is dated 2021 — so the live comparison table on AIVA's pricing page is the most current statement of what each plan allows, and it is the one to check before you rely on any of this. - Whether you need to pay is decided by what the music is for, not by how much you earn from it. A school project or a video for friends sits inside the free plan. Anything that promotes a business, or anything you want to hold the rights to, does not. Security: - AIVA's privacy policy is dated March 11, 2019 and reads like a standard online-shop policy: cookies, analytics, Stripe for payments. It says nothing about the music you generate, and nothing about the audio or MIDI files you upload as an influence (a reference file you give AIVA so it can build something with similar characteristics). That silence, rather than anything the policy says, is what to plan around. Security: Think twice before uploading an influence: - If the reference file is someone else's copyrighted music, uploading it raises a rights question that none of AIVA's plans answers for you. - AIVA's licence bars you from using its compositions as training data for machine learning, and from using them to build a competing service. It makes no matching promise in the other direction about what happens to what you upload. - An unfinished melody of your own is the one thing worth keeping off the platform until the policy says what becomes of it. Security: - For ordinary use this is low-stakes: there is no personal information in a chord progression. The gap here is about your material, not about you. Best at: - Instrumental background music — the job it was built for - Classical, cinematic and orchestral pieces, where the style library goes deepest - Producing several versions of one idea in a single pass and picking the best Okay at: - Modern pop, electronic and hip-hop — present in the style library, but not where the tool's reputation sits - Landing the exact arrangement you pictured on the first attempt — expect several passes - Starting from a vague idea with no musical vocabulary — the controls assume you know roughly what a key signature and a tempo are Avoid: - Building a stock-music library on it. AIVA's licence excludes "large scale licensing of MIDI or Audio compositions", which stays true on the top plan. - Assuming a company qualifies for Pro's ownership terms. Starter: Make a track from a style Choose Create track then From a Style, pick a style from the library, and set the duration and the number of compositions you want. Press create and let it produce a batch rather than a single track. A style on AIVA already carries the instruments and the arrangement before you touch a single setting, so learning to shop the style list is the actual skill here — Suno has no style library to shop, only a sentence to write. Starter: Steer the chords with a sentence This box lives inside Create track then From a Chord Progression, where AIVA calls it Text-to-Harmony. AIVA's own guidance is blunt about its reach: the prompt "only affects the harmony that will be used to generate your composition. It does not alter other musical elements, like the instruments or the arrangement." Mistaking this box for a song description is the most common misunderstanding of the tool, and AIVA also recommends writing in English for best results. The same field takes a named piece as a reference — try A progression in the style of Beethoven's Moonlight Sonata. Prompt: A cinematic progression in D minor that builds tension slowly, using 7th chords and resolving on a IV chord. Starter: Upload something you like as an influence Open Styles then + Create to reach the Style Designer, upload an audio or MIDI file as an influence, and let it fill in that style's parameters. Then generate from the style it builds rather than from a preset. AIVA's manual describes an uploaded influence as auto-populating the style's parameters, so what carries over is a set of settings you can see and change afterwards — not the recording. Suno's cover feature works on the audio itself, which is why its results are finished and AIVA's are still adjustable. Upload a MIDI file rather than audio if you have one; MIDI carries the notes themselves rather than a recording of them. Starter: Change one instrument in the editor Open a generated track in AIVA's editor, find the instrument layers, and mute or swap one, then play it back. The manual lists per-note editing, tempo changes and effects such as reverb on the same screen. AIVA keeps the layers and the individual notes separate all the way through to the editor, so a change you make edits the score rather than re-rolling the whole track. Mute every layer but one and listen to it alone; it is the quickest way to hear what a style is actually made of. Starter: Rewrite a prompt that missed AIVA publishes preferences for this field that are unusual to find written down at all: name chords in Roman numerals rather than letters, avoid combining contradictory musical ideas in one prompt, and rephrase rather than retry when something misses. Udio publishes no conventions of this kind, so this is a platform you can learn to write for. Prompt: A simple, diatonic progression in dorian mode, slow and unhurried, with no key change. Pitfalls: - The licence is decided when you download, not when you create. AIVA's terms say the licence granted "will depend on the plan that User is currently subscribed to" at the point of download — so a track generated today and downloaded next month goes out under next month's plan. - Upgrading works backwards here, unlike most AI music tools. AIVA's helpdesk says a Pro subscriber owns the copyright of compositions created before the subscription started as well as during it, and keeps that ownership after cancelling. Worth knowing before you re-make something you already have. - A MIDI file is notes, not sound. Open one in other music software and it plays on whatever instruments that software assigns. The orchestral sound you heard on AIVA's site is AIVA's own playback, and it does not travel with the file. - You cannot get the same track back. AIVA says the chance of it generating the same composition twice is "practically impossible", so a version you liked and clicked away from is gone rather than regenerable. Where to next: - Music generators: https://www.launchpadhq.io/discover?category=music-generation - Video creation and editing: https://www.launchpadhq.io/discover?category=video-creation-editing ### Getting Started with Aleph Alpha Canonical: https://www.launchpadhq.io/platform/aleph-alpha-luminous/guide Data: https://www.launchpadhq.io/data/v1/guides/aleph-alpha-luminous.json Last reviewed: 2026-09-13 Tool: Aleph Alpha (Luminous) (https://www.launchpadhq.io/platform/aleph-alpha-luminous): starting price Custom; free tier no; data as of 2026-10-20 The German company behind Luminous — now sold to governments and large businesses through a salesperson, not a website you can sign up to. Aleph Alpha's software is installed inside an organisation's own systems rather than opened on a public website. Staff there sign in to an internal address, type a question into a chat box, and read the answer back. What it is: - Aleph Alpha is a Heidelberg company that builds language models (a model is the AI "brain" that does the actual thinking) for European governments and large regulated businesses, and sells them as an installed stack called PhariaAI. - Luminous — the model line that made its name, and what this listing was originally about — is no longer part of what it sells. The company stopped building general-purpose models in September 2024 and now co-builds domain-specific ones with individual customers. - Its documentation describes a chat app and a prompt-and-settings workspace inside each installation, but both live at an address belonging to the customer. Treat that as what the docs describe rather than something anyone can confirm without an account. - In April 2026 the Canadian AI company Cohere announced a merger with Aleph Alpha; that deal had not closed when this page was last checked, and nothing public says what happens to the PhariaAI product line afterwards. Before you start: - There is no sign-up. The site has one call to action — contact sales — and its pricing page no longer exists, so the honest answer to whether you will need to pay is yes, through a negotiated contract rather than a card. - The one exception is a Research Playground, and it is narrow: the request form asks for an academic or research institute email address and says access is granted at Aleph Alpha's discretion. It does not say how long a decision takes or how many applications are accepted, so treat approval as possible rather than likely. - One Aleph Alpha model is free to download — Pharia-1-LLM-7B, published on Hugging Face. Getting an answer out of it means installing the company's own Scaling library and writing a few lines of Python; there is no click-to-install version. - If what you actually wanted was an AI assistant you can open and type into today, the chat and text category on this site lists ones that are open to anyone. Getting set up safely (The Aleph-Alpha organisation page on Hugging Face (huggingface.co/Aleph-Alpha)): - Download only from the Aleph-Alpha organisation itself. Re-uploads of these files exist under other accounts; those accounts are not the publisher, and a model file is code your computer will run. - Check the size before you start. The published files come to around 14 GB, and they are the full-size originals rather than a shrunk-down version, so loading them needs roughly their own size again in memory. That is a rule of thumb worked out from the published file sizes rather than a hardware requirement Aleph Alpha states, so treat it as an estimate. - The instructions for loading them are Aleph Alpha's own. Follow the vendor's developer guide rather than a blog post. Security: - Aleph Alpha's whole pitch is that its software runs inside the buying organisation's own systems on European servers rather than on Aleph Alpha's, so the question of whether a vendor keeps and trains on what you type is settled by where the software sits rather than by a setting. That is a genuine structural difference from a chatbot you open in a browser tab. Security: - Two things temper it. The privacy notice on aleph-alpha.com is dated 4 September 2024 and covers the website, not the product. And the developer guide's promise that prompt data is not stored and user inputs are not logged points to Terms and Conditions that we could not find published anywhere on the public site — so the binding version is presumably the one inside a customer's contract. If you are the person evaluating this for an organisation, read the contract, not the website. Security: If you download the free model instead: - It runs on your own machine, so what you type never leaves it. Best at: - German-language work — the models were trained on a multilingual corpus and the vendor states they are culturally and linguistically optimised for German, French and Spanish - Organisations that must keep data inside Europe and be able to prove where it runs - Deep, document-heavy work inside one organisation — the published customer results are government administration, contract compliance, technical document search and engineering requirements Okay at: - Looking good on benchmark tables — Aleph Alpha states outright that its models were not tailored to the standard leaderboards and would be expected to underperform on them - Serving one person. Every part of the buying process assumes an organisation behind you Avoid: - Anything commercial done with the free download. The Open Aleph License 1.0 permits use only for non-commercial and non-administrative purposes — and "non-administrative" rules out public-sector use, which is precisely what the company sells to its paying customers - Committing to a long roadmap on the strength of what is available today — this company has already retired an entire model line once Starter: Read the licence before you read anything else Open the LICENSE file on the Pharia-1-LLM-7B page on Hugging Face and read the short paragraph beginning "Subject to the terms and conditions." It is a few lines long. Downloadable models normally arrive under one of a handful of standard permissive licences, where reading it is a formality; Aleph Alpha wrote its own, and its "non-administrative" restriction has no equivalent in those standard licences — so skimming this one the way you would skim a familiar licence gives you the wrong answer. Mistral's downloadable models carry no comparable clause. Starter: Find out whether the door is open to you Go to the Research Playground request form and read the eligibility line before filling anything in. Two minutes there saves you waiting on a reply that may never come. Elsewhere a playground is gated by a card, which at least tells you the price of entry; Aleph Alpha gates this one behind who your employer is plus a discretionary decision, so there is no amount you can pay to get in by this route — and the form is the only place that says so. OpenAI's playground is card-gated, not employer-gated. Starter: Read the customer results as the spec sheet Open the customers section of aleph-alpha.com and read the three deployments it publishes: an assistant rolled out to 80,000 users in a government agency, document retrieval at a global chip manufacturer, and requirements processing at an automotive supplier. Because the company sells models co-built per customer rather than a fixed product, there is no feature list, tier table or price sheet to compare against anyone — those three customer results are the only public description of what you would actually be buying. Anthropic publishes plans, prices and model specs, so its case studies are illustration rather than specification. Starter: Check whether it is still shipping Open the release notes index in Aleph Alpha's documentation and look at the top entry. When this page was written the most recent was PhariaAI v1.260700.0. Aleph Alpha's version numbers are date-encoded — 1.260700.0 is the July 2026 release — so the number alone tells you how current the stack is without opening anything, which is the only outside-visible sign of life for a product no member of the public can log into. Cohere's changelog does not encode the release date in the version string, and its products are publicly reachable anyway. Pitfalls: - Searching for "Luminous" and landing on pages that look live. The old Luminous blog posts and benchmark write-ups are still at their original addresses with no notice on them, and third-party review sites still quote Luminous pricing. None of it is current. - Expecting the free model to know recent things. Those published files were last updated in January 2025 and have not been refreshed since. - Reading the developer documentation as if you could follow along. It gives the workspace's address as pharia-studio followed by your own organisation's domain. - Assuming published weights means open source. The model files are public, but the licence is the company's own and does not grant the freedom to use it for anything you like, which an open-source licence (the code is public and anyone can inspect it) normally would. Where to next: - AI chat assistants you can open today: https://www.launchpadhq.io/discover?category=text-conversational-ai - Local and open-source AI: https://www.launchpadhq.io/discover?category=local-open-source-ai - AI APIs and developer services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services ### Getting Started with AlphaSense Canonical: https://www.launchpadhq.io/platform/alphasense/guide Data: https://www.launchpadhq.io/data/v1/guides/alphasense.json Last reviewed: 2026-09-12 Tool: AlphaSense (https://www.launchpadhq.io/platform/alphasense): starting price Custom; free tier no; data as of not dated AI search across company filings, analyst notes and expert interviews — sold to firms, not to individuals. You sign in, type a question into a search bar in plain English, and get back a written answer with each claim linked to the exact sentence in the document it came from. Click a citation to read the source, then ask a follow-up in the same box to narrow or redirect the answer. What it is: - AlphaSense is a search engine for business and financial documents — company filings, earnings-call transcripts, news, research notes written by banks' own analysts, and interviews with people who have worked inside an industry. The company counts around 500 million documents from more than 10,000 sources. The pitch is the library rather than the AI: most of what is in it is licensed and closed, so a general AI search tool pointed at the open web cannot see it. - It is built and priced for organisations. AlphaSense sells to investment banks, hedge funds, private equity and venture firms, asset managers, consultancies, law firms and corporate strategy teams; its terms of use (last updated 5 June 2026) license the service to named "Authorized Users" for "internal business purposes." - One note about everything below. The whole product sits behind a sign-in, so what follows about screens, buttons and settings is drawn from AlphaSense's own help centre rather than from us using it. Before you start: - You cannot sign up for AlphaSense yourself, and it does not publish a price. Its pricing page lists two plans — Market Intelligence and Enterprise Intelligence — with no numbers next to either, saying only that it sells annual subscriptions "ranging from enterprise packages to per-seat options" and asking you to contact its sales team. Expert calls and Canalyst financial models sit beside those plans as separately paid add-ons. Treat any dollar figure you find on a third-party "AlphaSense pricing" page as somebody's guess, because the vendor publishes none. - In practice there are three ways in. Your employer already subscribes and an administrator adds you. Or you are a current student, staff or faculty member at a business school whose library licenses it — Harvard Business School's and Stanford GSB's libraries both list AlphaSense, in each case restricted to their own current members, and Harvard's listing notes expert transcripts cannot be downloaded and that an account includes 25 company-model downloads for its lifetime. Or you fill in the trial request form at alpha-sense.com and wait. We could not see that form's fields without submitting it, so we cannot tell you whether a personal email address is accepted, how long a trial runs, or whether it is granted automatically. - What you will actually be able to read depends on the contract, not on the product. Blocks of source material are switched on per account — AlphaSense calls them "content sets" — and its help centre says real-time research from a given bank needs what it calls an "entitlement," which is permission that follows from your firm having a trading, advisory or coverage relationship with that bank. Older "after-market" research from the same banks does not need one. Two people on the same platform can run the same search and get different results. - The closest honest substitute you can learn on this week is BamSEC — AlphaSense owns it, it is self-serve, it has a free tier, and a paid BamSEC Pro plan above that. It covers US company filings and earnings-call transcripts rather than the licensed bank research and expert interviews. Security: - The unusual thing about AlphaSense's privacy position is that it is written down, and it is favourable. Its privacy notice (last updated 19 June 2026) states that it does not use customer personal data, customer-provided content, or personal data processed on behalf of customers to train generally available generative artificial intelligence models — and says the same about data pulled in from a firm's own connected systems. Security: Where the real constraints are:: - The account is your firm's, not yours: the terms of use forbid sharing login credentials outright — on a per-seat contract that is also the term most likely to be noticed. - An administrator can see AI spending. The help centre describes a Credit Usage Dashboard visible to account administrators only, reporting how much of the organisation's allocation has gone and when the contract period ends. It does not say whether that breaks down per person, and we could not establish it — assume your usage is visible in aggregate at least. - What you find is not yours to pass on. A separate third-party terms page (last updated 10 January 2025) carries each content provider's own rules, and they are strict: no republishing or redistributing, no downloading or derivative works in several cases, and one provider, TD Securities, bars using its data to train any AI tool. Several state plainly that their material is not investment advice and is not directed at your circumstances. Best at: - Finding one specific statement buried in a very large pile of company documents. - Tracking how a single topic — pricing, tariffs, a competitor, a supply chain — has been discussed across many companies over time. - Private-company research through expert interview transcripts, the part you genuinely cannot get free elsewhere. AlphaSense's own pages put the library somewhere in the hundreds of thousands of transcripts across 25,000–29,000 companies, though its figures differ between pages. - Standing watch: dashboards and alerts that push new mentions to you instead of you going back to search. Okay at: - Anything on the open web — the help centre describes web search as something you switch on to reach beyond the library rather than the default. - Questions whose answer is a single number. It is a document-search product first, not a spreadsheet. - Being explored casually. There are public interactive product tours on alpha-sense.com that walk through the interface without an account; they appear to be open, though we could not confirm whether an email is asked for before one plays. They are walkthroughs of a fully-loaded configuration, not the account you would get. Avoid: - Litigation and regulatory work. The terms of use expressly prohibit using the service for litigation purposes or in government or regulatory filings, and separately prohibit benchmarking. - Treating a generated answer as the finding. What it summarises is largely opinion — analyst notes, expert interviews, management commentary. - Deciding on your own to point it at your firm's internal files. Connecting shared drives is what the Enterprise Intelligence plan is for — a contract-level decision made above you, not a setting you turn on. Starter: Ask the question you'd otherwise read thirty documents to answer AlphaSense's help centre publishes almost this exact wording as its example prompt (the message you type). Test the sentence against Perplexity and it fails — Perplexity would answer from the open web, while this answer can only be assembled from the content sets your particular account is licensed for. Change "earnings calls" to whichever document type you actually care about, and cut the window to six weeks when you want the current story rather than the trend. Prompt: Act as an equity analyst. Summarize how [company] has discussed supply chain risks in earnings calls over the last 6 months. Group findings by theme and present them in a bulleted list. Starter: Interrogate a single document instead of reading it The help centre describes an "Ask in Doc" control on an open document, which confines the AI to that one file. Swap in ChatGPT and the sentence stops being true: here nothing is uploaded, the file is already in the library, and the scope is a setting rather than an instruction. Asking it to quote the lines it used turns a summary you have to trust into one you can check. Prompt: What did management say about pricing, and did the tone change from the previous quarter? Quote the lines you are relying on. Starter: Search like a librarian, not like a chatbot There is a real operator language behind the keyword bar, which is rare in an AI product. AlphaSense's published quick-reference guide documents AND, OR, NOT, quoted phrases, a TITLE() filter and proximity operators, and says typing ? in the search bar lists the full set. Quotation marks matter — the guide says they stop the search quietly matching near-relatives of your words. For words near each other AlphaSense documents proximity operators, its own example being NEAR2(travel space), but warns they cannot be combined with extra required or excluded keywords. Run a proximity search on its own. Prompt: "pricing power" AND (tariff OR tariffs) NOT guidance Starter: Turn a question into something you'd otherwise spend a morning writing The help centre describes three reasoning settings — Auto, Think Longer and Deep Research — with Deep Research the one that goes away and runs a multi-step job on its own. Choose it deliberately, because these AI features draw on credits (the platform's unit of spend — each thing you make costs some) that come out of your organisation's allocation and reset at the contract end date, not monthly. Name the companies and the window explicitly. The vaguer the question, the more the run costs and the more generic the report that comes back. Prompt: Compare how [three named competitors] have described demand in [market] over the last four quarters. Note where they disagree, and flag anything only one of them mentions. Starter: Before anything else, find out what your account can actually see In the search bar, open the "All Sources" control the help centre describes and read which content sets are switched on for you — filings, news, bank research, expert transcripts, your firm's own documents — then check whether web search is on or off. The material here is per-contract, which makes this the difference between a search that comes back empty because nobody said the thing and one that comes back empty because you are not licensed for the source that said it. Pitfalls: - Reading the summary and never clicking through to the source sentence. That link is the product's entire argument for itself over a free chatbot; skip it and you have bought an expensive chatbot. - Reaching for the heaviest reasoning setting out of habit. Auto is the default the help centre describes and it is fine for a lookup — a shared allocation disappears one unnecessary deep run at a time. - Treating expert-interview transcripts as neutral. They are paid, arranged conversations with individuals, which is exactly why AlphaSense describes a compliance process around them. Read one as a single informed person's view, not as a source of record. - Stopping at the first answer. The help centre describes highlighting a passage inside a result to ask about that passage specifically — a narrowing move a plain follow-up does not give you. Where to next: - Research & Academic Tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Finance & Real Estate AI: https://www.launchpadhq.io/discover?category=finance-real-estate-ai - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Amazon Bedrock Canonical: https://www.launchpadhq.io/platform/amazon-bedrock/guide Data: https://www.launchpadhq.io/data/v1/guides/amazon-bedrock.json Last reviewed: 2026-09-08 Tool: Amazon Bedrock (https://www.launchpadhq.io/platform/amazon-bedrock): starting price See pricing; free tier yes; data as of not dated AWS's single door to AI models from Anthropic, OpenAI, Meta and a dozen more labs — billed by what you use. You sign in to AWS, open Bedrock, and pick a model (the AI "brain" that does the actual thinking) from a catalog. You type a question in the browser and read the reply. When you want it inside your own software, you generate a key and call the same models from code — swapping models means changing one line. What it is: - Amazon Bedrock is AWS's single door to AI models built by other companies. Instead of an account with Anthropic, another with OpenAI, another with Mistral, you reach models from around eighteen labs behind one login and one AWS bill. It's built for people making software — an API (a way for programs to talk to each other without a person clicking), not a chat app. Before you start: - The barrier here is AWS, not Bedrock. There's no Bedrock signup — you open an AWS account, which needs a real credit or debit card even on the free plan, and you meet AWS's permissions and regions before you meet a model. - Cost is per use, with no subscription. A new AWS account gets a starting credit at signup and can earn more by trying services, Bedrock among them — but the free plan ends six months in, or when the credit runs out. A trial, not an allowance. - Anthropic's models need a one-time use-case form per AWS account before your first call. Other providers' models subscribe themselves on first use. - Which models you can reach depends on the AWS region you pick, and newer ones are called through a geographic profile — the same name with us. or eu. in front. Security: - The reason companies pick Bedrock over calling each lab directly is in AWS's own FAQ: your content isn't used to improve the base models and isn't shared with any model provider. Claude runs here without Anthropic seeing your text. Security: Your key spends your money — so give it a lifespan: - You choose the expiry when you create a long-term key, and AWS's own docs recommend those keys only for exploring. - It's shown once — copy it somewhere private, never into a public code repository or a screenshot. Security: - One honest wrinkle: geographic profiles spread a request across regions inside a geography like US or EU. Your data stays in that geography, but AWS states inputs and outputs may leave the region you picked, and may be stored there for abuse detection. Best at: - Putting one question to rival labs' models without an account at each - Building on Claude, GPT or Llama with your text staying inside your own AWS account - Teams already on AWS, who inherit the billing, permissions and logging they have Okay at: - Being a first AI tool — there's no consumer app here - Predictable monthly cost, since the bill follows use - Launch-day access — new frontier models often reach their own lab's API first Avoid: - Assuming a model you read about is callable — it depends on your region - Long-term API keys in production, which AWS's own docs advise against - Following an older tutorial into Bedrock Agents — AWS closed it to new customers on 30 July 2026 and points new builders at AgentCore Starter: Compare models before you write any code Put one question to up to three models in the console and read the replies side by side. Bedrock's catalog names around eighteen providers, OpenAI and Anthropic among them, and the console runs one question across three at once. That's how you find out a cheap model answers as well as the expensive one. Starter: Give yourself a key with an expiry date Generate a long-term key in the console and set it to expire; AWS's quickstart uses 30 days. Bedrock lets you choose the lifespan as you create the key, and its short-term keys last about twelve hours. A forgotten Bedrock key stops working; a forgotten permanent key keeps spending. Starter: Set a budget, and know what it actually does Create an AWS Budget with an email threshold before your first call. On AWS the alarm and the brake are separate: a budget threshold sends a notification, while stopping spend means configuring a budget action that applies a deny policy. Starter: Clear the Anthropic form first Submit the one-time use-case form for Anthropic models, once per AWS account. Anthropic is the provider on Bedrock that gates first use this way; access opens the moment you submit, and a form filed at an organisation's management account carries down to the accounts beneath it. Pitfalls: - Most Bedrock confusion is AWS confusion. Permissions, regions and billing all belong to AWS, and a free-plan account closes itself six months after signup. - A first call to a provider you haven't used can fail while Bedrock subscribes you in the background. Wait a couple of minutes and try again before assuming you set something up wrong. Where to next: - AI APIs & Developer Services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - AI Coding & Development: https://www.launchpadhq.io/discover?category=ai-coding-development ### Getting Started with Amazon Q Developer Canonical: https://www.launchpadhq.io/platform/amazon-q-developer/guide Data: https://www.launchpadhq.io/data/v1/guides/amazon-q-developer.json Last reviewed: 2026-09-08 Tool: Amazon Q Developer (https://www.launchpadhq.io/platform/amazon-q-developer): starting price Free, from $19/user/mo; free tier yes; data as of not dated Amazon's assistant for AWS — still live in the console, but closed to new coding-tool signups. You sign in to the AWS website, click the hexagonal Q icon in the top corner, and type your question into the panel that slides out. It answers in plain language, you ask follow-ups in the same panel, and it keeps the thread of the conversation as you go. What it is: - Amazon Q Developer is the AI assistant Amazon built into Amazon Web Services (AWS) — the cloud platform a large share of the internet runs on. It answers questions about AWS itself: what a service does, why a setting won't save, what a charge on your bill is for. - It absorbed Amazon CodeWhisperer, Amazon's earlier code-suggestion tool, in April 2024. If you land on a CodeWhisperer tutorial, this is where that product went. - Read this part before you start: the half of Q Developer that lived inside code editors is being retired. AWS stopped taking new signups for it on May 15, 2026 and ends support on April 30, 2027; the replacement is a separate AWS product, Kiro, at kiro.dev. What is still open to a newcomer is the assistant inside the AWS console and documentation, which AWS says the shutdown does not affect. That is what this page covers. Before you start: - This is for people using AWS, not a general assistant. AWS builds and sells it to answer questions about AWS and about your own account; nothing about it is offered as a general chatbot, so don't judge it as one. - You need a free AWS account, and creating one means giving AWS a payment method it can validate and passing an automated phone verification call. The assistant itself costs nothing at the Free Tier level. - There is nothing to install — the assistant is the hexagonal Q icon on the AWS console and on AWS documentation pages. The paid tier (Pro) is shut to newcomers, so the free level is all you get. - Skip the code-editor add-ons: new accounts can no longer sign in to them. If an AI coding tool is what you actually want, try kiro.dev or one of the coding tools linked at the bottom of this page. Security: - Anything you type here is a conversation with Amazon about your account. On the Free Tier, AWS's own documentation says your questions and its answers may be stored and used for service improvement, including model training. The documented way to switch that off is an AWS Organizations opt-out policy — an administrator setting, not a checkbox in your profile. On a personal account, assume your conversations are used. Security: Keep out of the chat panel: - Access keys, secret keys and passwords — it never needs them to answer a question - Customer or employee records you keep in AWS Best at: - Explaining an AWS service in plain language while you are looking at it - Turning a console error message into what it means and what to click next - Turning a question it can't answer into an AWS support case Okay at: - Anything not about AWS — it will try, but that isn't what it is for - Questions about your own resources — what it can see there is bounded by your AWS permissions Avoid: - Building any habit on the code-editor add-ons — new accounts cannot sign in to them at all, and the whole add-on line is on a countdown to end of support. - Treating a free-tier conversation as private: Amazon's documentation says free-tier questions and answers may be used for model training, and the opt-out is an organisation-level policy, not a personal setting. Starter: The service explainer Ask it with the S3 console open and Q answers for the service and region you are in — AWS shipped that console-context awareness in November 2024. A chatbot in another tab explains S3 just as well but has no idea which of your screens is on fire. Swap S3 for whichever service the tutorial you are following just told you to open. Prompt: I'm new to AWS. Explain what S3 is, what problem it solves, and the three things beginners most often get wrong about it. Keep it under 200 words. Starter: The error translator The panel opens on top of the page that failed, so the service, the region and the error are already in front of it — you paste one line instead of rebuilding the situation in another tab. If the fix needs a setting you can't find, ask "where exactly in the console is that option?" Prompt: I got this error in the console: "[paste the exact message]". Explain what it means, what usually causes it, and how to fix it from where I am. Starter: The bill question Q is the front door to AWS Support inside the console: AWS documents opening a support case and reaching a live agent as things you can just ask for, so a question it can't answer becomes a ticket rather than a dead end. When an answer is going nowhere, ask "can you connect me with a person?" Prompt: I'm being charged for [service] and I don't know why. Explain what usually causes that charge and what I'd change to stop it. If you can't tell from here, open a support case for me. Pitfalls: - AWS's own marketing pages have not caught up. The product page still shows "Get started for free" with download buttons and says nothing about the May 15, 2026 signup cutoff, which appears only in an AWS blog post and the documentation. If a download leads you to a sign-in you can't complete, this is why. - "Amazon Q" is several products: Q Developer, this one, answers about AWS; Amazon Q Business answers about a company's internal documents; and another version lives in Slack and Microsoft Teams. Tutorials rarely say which they mean. Where to next: - AI coding tools: https://www.launchpadhq.io/discover?category=ai-coding-development - Developer APIs and services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - General AI chatbots: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Apollo.io Canonical: https://www.launchpadhq.io/platform/apollo-io/guide Data: https://www.launchpadhq.io/data/v1/guides/apollo-io.json Last reviewed: 2026-10-23 Tool: Apollo.io (https://www.launchpadhq.io/platform/apollo-io): starting price Free, from $49/month; free tier yes; data as of 2026-10-14 A searchable database of business contacts plus the tools to email and call them — powerful for sales outreach, and easy to waste credits in. You filter Apollo's database by things like job title, industry and company size to find people who match your ideal customer. Seeing someone's verified email or phone number costs credits (Apollo's unit of spend). You save them to lists, then contact them through sequences — scheduled series of emails, calls and tasks — sent from your own connected mailbox. What it is: - Apollo.io is a sales prospecting platform. Its homepage describes a database of more than 240 million contacts and 30 million companies, which you search with filters to build lists of people likely to buy what you sell. It is aimed at salespeople, founders and recruiters doing outbound — reaching out to people who haven't contacted you first. - Finding someone is free; getting their details is not. Apollo's knowledge base says clicking to access a verified email or a phone number for a new contact spends credits and saves them as a contact in your account. Many other actions also draw on credits, including enrichment (filling in missing details on contacts you upload), AI research and writing, and calls made through the built-in dialer. - Around the database sit the outreach tools: sequences, a dialer, a meeting scheduler, and a Chrome extension that shows Apollo data while you browse LinkedIn, Gmail, your CRM or a company's website. Apollo also has an AI assistant and AI email writing; its 2026 release notes describe the assistant as free on paid plans as an introductory offer, with free accounts limited to a few chats. Before you start: - Sign up with a work email if you have one. Apollo's pricing page says free accounts on a verified company email domain get a much higher monthly credit cap (an anti-abuse limit) than free accounts without one, such as a personal Gmail address, which get only a small allowance. Apollo recommends using an email with a valid company domain. - Credits don't roll over. Apollo's pricing page says unused credits expire at the end of each billing cycle, including add-on credits you bought. Its credits page says an email is the cheapest thing to reveal per contact; phone numbers cost more, and Apollo doesn't refund credits for mobile numbers. - Paid plans and trials. The pricing page lists Free, Basic, Professional and a custom tier, and offers a 14-day trial of a paid plan; after a trial you can stay on Free with no time limit. Plan prices load dynamically on that page and we couldn't confirm them from an official source on review day, so check apollo.io/pricing for current figures. Free plans can connect Gmail mailboxes only. - Read the contract terms before paying. Apollo's terms carry an automatic-renewal notice and say fees are non-cancellable and non-refundable; its pricing FAQ says cancellations take effect at the end of the term and downgrades take effect immediately with no refund. Getting set up safely (Sign up at apollo.io and install the Chrome extension only from the link inside Apollo or its official Chrome Web Store listing — the extension sees the pages you visit on LinkedIn, Gmail and your CRM.): - Link your mailbox before you send anything. In Settings → Outbound → Mailbox, link the mailbox you'll send from. Apollo's getting-started checklist says to read its deliverability guidance before your first send. - Set conservative sending limits. Apollo's deliverability articles suggest starting around 50 emails a day per mailbox, spacing them out, and setting up your domain's SPF, DKIM and DMARC records (settings that prove your emails are really from you). Apollo also recommends sending outreach from a separate subdomain, so a bad campaign can't damage your main email address. - Watch your credit usage from day one. Settings → Credit usage shows what's been spent and on what; admins can cap credits per user. Check it after your first search-and-reveal session. Security: - Apollo holds information about people who never signed up for it, and the legal duty to contact them properly is yours. Apollo's GDPR overview says you must give the people you email a way to opt out, and that you're responsible for setting Apollo up to meet your own privacy obligations. In the US, the FTC's CAN-SPAM guide says commercial email must include a working opt-out that you honour within 10 business days, a valid physical postal address, a subject line that matches the message, and accurate sender details; Apollo can add an unsubscribe link automatically. Many other countries are stricter, and some require consent before you email at all. This is general information, not legal advice — Apollo itself says to check your obligations with local counsel. Security: Compliance tools to know about:: - Apollo's GDPR settings (admin only, under Settings → Outbound → Prospecting configuration) can block prospecting, emailing and email tracking for people located in the EU. Apollo says they only work when it can determine a person's location, so check where someone is yourself before contacting them. - Do-Not-Call screening checks numbers only against certain national registries — Apollo names the UK and Germany. Its GDPR overview notes calling rules vary by country, and some places require consent before certain calls rather than just letting people opt out. - If you are in Apollo's database yourself, its privacy center lets you request a copy of your data or have it removed. Security: - Apollo states it holds SOC 2 and ISO 27001 certifications. Connecting your mailbox and CRM gives Apollo access to them, so connect only the accounts you mean to use for outreach. Best at: - Building targeted lists of people at companies that fit your ideal customer, fast. - Running a whole outreach routine — find, email, call, book a meeting — from one place. - Seeing contact and company details while you browse LinkedIn or your CRM, through the Chrome extension. Okay at: - Data accuracy. Apollo shows verification statuses as confidence levels, not guarantees, and only refunds bounced-email credits under specific conditions. - Use by people who aren't in sales. It can help with recruiting or partnerships, but the product is built around selling. Avoid: - Revealing phone numbers in bulk on a small credit budget. - Blasting a new email address with hundreds of messages on day one. - Treating a contact in the database as permission to email them. Starter: Describe your ideal customer Before searching, write down two or three job titles, an industry, a company size range and a location for the people you want to reach. Then open People search and apply exactly those filters. Apollo's own getting-started checklist starts with defining your ideal customer and buyer personas. A tight search means you spend credits only on people who fit. Starter: Save the search, not just the people Save your filtered search so you can rerun it later, and set an alert if you want to hear when new matches appear. Credits are spent when you access someone's details, not when you filter, so this builds your pipeline before you spend any. Starter: Reveal a small, deliberate batch Pick ten people from your search, access their emails only, and add them to a new list. Then check Settings → Credit usage to see what it cost. Emails are the cheapest reveal; phone numbers cost more and aren't refundable. Starting with ten shows you the real cost of your workflow before you scale it. Starter: Build a short first sequence Create a sequence with three steps — a short email, a follow-up a few days later, and a final check-in — and add your ten contacts. Make sure the unsubscribe link is on. Sequences are where Apollo saves you time. Keeping the first one small means you can read every reply and adjust your message before sending more. Starter: Try the extension on a profile you know Install the Chrome extension, open the LinkedIn page of a company you already work with, and see what Apollo shows about it. Checking data on people and companies you know is the quickest way to judge how accurate Apollo is for your market. Pitfalls: - Signing up with a personal email. The small monthly credit limit on personal-email accounts makes the free plan hard to judge fairly. - Letting credits expire. They don't carry over between billing cycles, so plan what you'll reveal each month. - Turning on every AI and enrichment feature. Waterfall enrichment, AI research and the dialer all draw from the same credit pool. - Ignoring spam reports. Apollo's deliverability guidance notes unsubscribes don't hurt your sending reputation, but spam reports do — write to fewer, better-matched people. Where to next: - Sales, Marketing & SEO AI: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai - Automate the steps between tools: https://www.launchpadhq.io/discover?category=workflow-automation - Record and summarise sales calls: https://www.launchpadhq.io/discover?category=meetings-notes ### Getting Started with AssemblyAI Canonical: https://www.launchpadhq.io/platform/assemblyai/guide Data: https://www.launchpadhq.io/data/v1/guides/assemblyai.json Last reviewed: 2026-09-08 Tool: AssemblyAI (https://www.launchpadhq.io/platform/assemblyai): starting price Universal-2 $0.15/hr; free tier yes; data as of 2026-10-14 Turn recordings into speaker-labelled text — a transcription service built for programs to call, with a browser playground for trying it on your own audio first. You sign in, drag an audio or video file into the browser playground, and a transcript comes back with each speaker's lines separated. You flick switches to change what comes back, run the file again, and compare. When it looks right, you copy the code the page has written for you into your own project. What it is: - AssemblyAI is a speech-to-text service that programs talk to through an API (a way for programs to talk to each other without a person clicking): your code sends a recording, and it sends back the words, with timestamps, speaker labels and — if you ask — a summary. There is no consumer app and no meeting bot. It is the transcription engine other products are built on, sold by the hour of audio. - So it is a component, not a destination. Speaker separation, translation, topic tagging and blanking out personal details are switches on one request rather than separate products, and the model (the AI "brain" that does the actual thinking) it routes to is chosen the same way. Before you start: - You do not need to write code for the first hour. Signing up gets you a browser playground where you upload a recording, flip features on and off, and read the transcript on screen. Anything past that — a batch of files, an app of your own — means code. If you only want your own meetings transcribed, a meeting-notes tool does that with no code at all. - New accounts get a free credit grant with no card asked for — a one-off grant rather than a monthly allowance, which AssemblyAI's FAQ says does not expire. Everything after that is billed by the hour of audio rather than by subscription, with live transcription as someone speaks costing roughly double the flagship model's rate. Current rates are on its Launchpad page. Playground runs bill the same way. - Free accounts cannot opt out of model training. The setting that stops AssemblyAI using your files to improve its models is, by its own docs, a paid-plan feature. Evaluate with audio you would not mind a stranger reading. Security: - AssemblyAI's security page says audio and transcripts are not stored after processing. Its own documentation is more specific and less absolute: with no special agreement in place, uploaded audio starts being deleted after about 24 hours and transcripts after around 30 days. Neither is instant, and the docs are the version to plan around. Security: Two things to settle before you upload anything real:: - Files sent to the API can be used to train AssemblyAI's models unless you have opted out, are on its European servers, or have signed a healthcare agreement. The opt-out lives on the dashboard's Data Controls page, for paid accounts only, and the docs say it cannot be applied retroactively. - Your API key (a password that identifies your app, and that spends your money) belongs wherever your project keeps its secrets, never in a file you share. Free accounts get two, so you can retire one without breaking everything. Best at: - Long recordings — one file can run up to 10 hours, so a day of interviews needs no chopping up - Sorting out who said what — speaker labels and timestamps arrive with the words - Doing more than transcribing in one pass — summaries, topic tags and redaction are switches on the same request Okay at: - Live transcription as someone speaks — it works, but costs roughly double, and free accounts can open only 5 new connections a minute - Languages outside the main set — 99 are supported, but the flagship model's full quality covers 18 Avoid: - Evaluating on confidential audio while on the free plan — the setting that stops your files training its models is for paying accounts only - Reading "not stored after processing" as "gone immediately" — the docs put default transcript deletion at around 30 days - Hard-coding a model name and forgetting it — AssemblyAI retired Universal-3 Pro on 2 September 2026, and requests naming it now return errors Starter: Run one of your own recordings through the playground Uploads a file in the browser and shows the transcript, speaker by speaker. Playground runs draw down the same credit as production and appear in the dashboard's cost tab, so what you spend evaluating is a number you can multiply — not a free demo priced differently from the real thing. Use a difficult recording: crosstalk, an accent, background noise. A clean file only confirms what you assumed. Starter: Re-run the same file with speaker labels and language detection on Uploads the same recording again in the playground with the speaker-label and language-detection switches turned on, and compares the two transcripts. Every capability here is a flag on one request rather than a second service to buy, so re-running one file with switches on and off is how you size the job — and the playground then writes the code for the exact combination you chose. Starter: Decide between the cheaper model and the accurate one Runs the same audio through Universal-2 and Universal-3.5 Pro and compares both. Universal-2 costs noticeably less per hour than Universal-3.5 Pro — invisible on one file, decisive at volume, and a quality gap only your own audio can settle. The free credit covers both runs; rebuilding on the other one later will not be free. Starter: Turn on redaction of personal details before uploading anything sensitive Blanks names, phone numbers and card numbers from the transcript, and can bleep them from the audio in the same request. On a free account this is the only privacy control you actually have, because the training opt-out is paid-only. Redacting as you transcribe keeps identifying details out of a file you cannot opt out of. Pitfalls: - Treating the free credit as recurring. When the balance hits zero with no card on file, API access pauses until you top up. - Forgetting the playground costs money — every browser run bills against the same credit. - Judging speed from a free account: free runs about 5 jobs in parallel, paid about 200. Where to next: - Other developer APIs: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - Voice and speech tools: https://www.launchpadhq.io/discover?category=voice-speech - Tools that capture and summarise meetings: https://www.launchpadhq.io/discover?category=meetings-notes ### Getting Started with Autodesk Flow Studio (formerly Wonder Studio) Canonical: https://www.launchpadhq.io/platform/wonder-dynamics-wonder-studio/guide Data: https://www.launchpadhq.io/data/v1/guides/wonder-dynamics-wonder-studio.json Last reviewed: 2026-09-12 Tool: Autodesk Flow Studio (https://www.launchpadhq.io/platform/wonder-dynamics-wonder-studio): starting price Free, from See pricing; free tier yes; data as of 2026-10-20 Turn a video of a real person into a scene starring an animated character — no motion-capture suit, no 3D software on your own computer. You upload a video of someone moving, pick a 3D character — a digital figure built in 3D software — to stand in for them, and the site processes the clip. A while later the shot comes back with your character doing what the person did. Adjust the settings and run it again until it looks right. What it is: - Autodesk Flow Studio takes an ordinary video of a person and replaces them with a 3D character that copies their movement, their position in the frame and, to a point, their expression. It was called Wonder Studio until Autodesk — the company behind AutoCAD and Maya — bought its maker, Wonder Dynamics, and renamed it. The old name is still on the old website, the documentation site and nearly every tutorial video; it is the same product. - It is built for someone who has footage and an idea but no motion-capture stage and no visual-effects artist. What comes back is not only a finished clip but the raw material a 3D program expects. - That output shape is the real fork in the road. Runway's performance-transfer tools hand you a finished video and stop there. Flow Studio hands you parts — a gift if you are heading into Blender, Maya or Unreal afterwards, overkill if you only wanted something to post. Before you start: - You can still sign up, and there is still a free tier — but not where the old links send you. Since 12 August 2025, all new sign-ups and purchases go through Autodesk: start at autodesk.com/products/flow-studio/overview and use the Compare page's "Access now". The original wonderdynamics.com site is still up, still has a "Get Started" button, and carries no notice that anything has changed. Subscriptions bought on Wonder Dynamics before that date do not renew automatically. - You bring two things: a video, and a character. The video has to be an .mp4 or .mov, no more than two minutes long, and on the free tier under 100MB. The character is the harder half: Flow Studio expects a rigged 3D character file (.blend or .fbx) posed with the arms straight out in a T shape. Building one is a separate craft and this page does not teach it — if you do not have one, the built-in Templates are the way in. - The free tier is real but small: a modest monthly allowance of credits (the platform's unit of spend — each thing you make costs some), the finished video capped at 720p, a watermark (a mark on generated output identifying it as AI-made), one actor tracked per project, three custom characters and 1GB of storage. At the credit rate in Autodesk's own Flow Studio terms, a free month buys only a few seconds of finished full character replacement, or roughly five times as much if you take only the movement data and skip the render. Autodesk moved to "more granular pricing controls per model and configuration" in August 2026, and that rate is published for a paid entitlement rather than the free tier, so treat it as a ballpark and check your own credit cost before running a job. You would move to a paid plan the moment you want a second actor, a clean 1080p export, or more than a test's worth of footage; the entry Lite plan was cut in price when the free tier launched. - It runs in a browser, and only in Chrome or Safari — the documentation says it is "not yet optimized for mobile browsers". Your own computer's power is irrelevant; the work happens on Autodesk's servers. Security: - The upload here is video of real people, and that changes the question. Wonder Dynamics' terms of service say you keep ownership of what you upload and of what comes back: "you shall retain all rights, title and interest in and to the User Content and Processed Content." The same terms grant the company a "perpetual, irrevocable, nonexclusive, royalty-free and fully paid, worldwide license" to use processed content "for Wonder's internal development purposes, including, without limitation, improving and enhancing Wonder's AI models" — limited to "an anonymized and aggregated way." No opt-out switch is named in the public terms. Security: Three things worth settling before your first upload: - Get the person in the footage to agree first. Their face and the way they move are both in that file, and both go to a cloud service processed in the United States. - Know which terms you accepted, because there are two sets. Signing up now runs through Autodesk, whose General Terms were last updated 30 March 2026 — but you also end up with a Wonder Dynamics account governed by its own, undated terms of service, and it is those that carry the AI-improvement clause above. Autodesk's Flow Studio special terms (dated 8 December 2025) say they "take precedence over any terms You may be required to accept during Wonder Dynamics account creation", but that sentence sits inside a section about a paid Media & Entertainment Collection entitlement, so it may not cover a free sign-up. If your footage is sensitive, read both before you upload. - Assume uploaded footage stays. We could not find a retention period for uploaded video stated anywhere in the public terms or privacy policy, and nothing documents that deleting a project deletes the file. Best at: - Replacing a single actor in a single-camera shot with your own 3D character — no motion-capture suit, no stage, no markers on anyone's clothes. - Pulling reusable movement out of ordinary footage — body and hand tracking exported as an FBX file, which Blender, Maya and Unreal all read. - Producing the supporting pieces of a visual-effects shot automatically: the frame with the original actor erased, a record of how the camera moved, and the new character on its own layer. Okay at: - Faces. Transferring a facial performance needs a character built with the right facial controls, and the documentation says accessories and facial hair "will degrade the accuracy of the expression." - Anything long or complicated. On top of the two-minute cap there is a ceiling of 15 shots per upload, and a video cannot be trimmed once it is in Flow Studio — so you edit first, elsewhere, and upload a finished cut. - Close-ups. The docs say "extreme close-up shots, where very little of the actor's body is visible, might not process properly." Avoid: - Shots where the actor touches anything. Autodesk states it plainly: "character-object and character-character interactions are not supported" — a handshake, a prop, a doorknob, two people embracing. - Shots where something passes in front of the actor. Occlusion has only "partial support", and a foreground object is not put back in front of the new character, so your character walks through the table rather than behind it. - Footage you have already cut with fades, wipes or spins. Only hard cuts are supported; the docs say anything else produces artifacts. - Loose or dark costume. The documented advice is "tight-fitting, bright and colorful clothing" — a hoodie or a long coat is working against the tracker, not with it. Starter: Run a Template before you spend anything Opens one of the built-in example projects: it already contains footage and one of Autodesk's own characters, and finishes far faster than a normal job. Templates are pre-processed and run on the supplied characters, which the documentation says lets you "generate results without using up render time" — the only way to see what finished output looks like before the month's credits are gone. While you are in there, look at the list of export elements — knowing which are separate downloads changes what you would go out and shoot. Starter: Shoot ten seconds the way the documentation asks for One person, full body in frame, against a background they do not overlap with, in even light and with as little motion blur as you can manage. Flow Studio publishes a ranked list of which body parts its tracker most needs to see — head first, then legs, then hips and torso, then arms. Framing decided before you press record is worth more here than anything you can adjust afterwards. The docs' own words are "shoot wider, then crop in." Starter: Take the movement only, and skip the character Runs the clip through AI Motion Capture mode, which returns a file of body and hand movement instead of a finished video. You apply that movement to a character inside Blender, Maya or Unreal. By the credit rate quoted above, movement-only processing costs roughly a fifth of what full character replacement costs on the same footage. It is also the one mode that does not require a Flow Studio-ready character at all. Choose USD rather than FBX if you want every bone: the docs say FBX delivers "only retargeted bones" while USD delivers all of them. Starter: Put your own character through the validator before you upload it Installs Autodesk's free Blender or Maya add-on — from the Downloads page on the official help site, not a search result — and runs the same check the website runs, on your own machine and instantly. Flow Studio will not take an arbitrary 3D file: it wants a rig whose name ends in the tag BODY, at least one bone representing the hips, a T-pose, and under 1,500,000 polygons. Failing that check after an upload costs you a round trip; failing it locally costs nothing. The Downloads page also has example characters for Blender, Maya and FBX — copy one's bone naming rather than guessing. Starter: Use the Canvas to fix a frame, not to make a film Canvas is a separate workspace of connected boxes where you can send a rendered shot and then generate or edit still images and video with hosted generation tools. It is the one part of Flow Studio where you type a description of what you want instead of uploading it, and it sits on the same canvas as your rendered shot, so a background fix never leaves the project. The documentation's own advice for writing those descriptions: describe one main object or character per request, use "clear, literal descriptions instead of cinematic or narrative language," and for anything you intend to animate later, ask for a neutral pose with the limbs separated. We could not confirm from the public pages whether Canvas is included on the free tier or what each generation costs in credits — check the cost the interface shows you before you run one. Pitfalls: - Forgetting to tick an export before you process. The docs are blunt about it: enable everything you want "before starting processing, as they will not be accessible afterward." Missing one means running the whole job — and paying the credits — a second time. - Following a tutorial made under the old name. Almost every video tutorial you will find says "Wonder Studio", and most predate the dashboard redesign of August 2026 — the workflow is broadly the same, the screens are not. Where to next: - Video creation and editing: https://www.launchpadhq.io/discover?category=video-creation-editing - Gaming and creative AI: https://www.launchpadhq.io/discover?category=gaming-creative-ai - Image generation and editing: https://www.launchpadhq.io/discover?category=image-generation-editing ### Getting Started with Azure OpenAI Service Canonical: https://www.launchpadhq.io/platform/azure-openai-service/guide Data: https://www.launchpadhq.io/data/v1/guides/azure-openai-service.json Last reviewed: 2026-09-08 Tool: Azure OpenAI Service (https://www.launchpadhq.io/platform/azure-openai-service): starting price See pricing; free tier yes; data as of not dated OpenAI's GPT models, rented through a Microsoft Azure account — the enterprise route to the technology behind ChatGPT, with control over where your data gets processed. You sign in to Microsoft's Foundry portal with an Azure account, pick a model (the AI "brain" that does the actual thinking) from a catalogue, switch on your own copy of it, then type into a chat box and read the reply. Once it works there, your own program can send the same requests instead. What it is: - Azure OpenAI is how you use OpenAI's GPT models — the same family behind ChatGPT — as a service you rent through a Microsoft Azure account rather than buying from OpenAI directly. Same models, different landlord: the bill, the sign-in, the legal agreement and the support all run through Microsoft. - The name has moved twice, and old guides will confuse you. Microsoft folded this into its Foundry platform and now sells it as "Azure OpenAI in Foundry Models." The wider platform, called Azure AI Foundry until late 2025, is now Microsoft Foundry, and was Azure AI Studio before that. "Azure OpenAI Service," "Azure AI Foundry" and "Microsoft Foundry" in a tutorial all point at this. - Who it is really for: people whose organisation already runs on Azure, or who need to be able to say where in the world their prompts get processed. This is infrastructure, not a product you chat with. There is no consumer app — there is a portal, a test playground, and an API (a way for programs to talk to each other without a person clicking). Before you start: - There is no free tier, and that is the single most important thing to know before you sign up. A new Azure account currently comes with some starter credit to spend in its first 30 days plus twelve months of free monthly amounts on a list of other Azure services — but the AI models are not on that list. Microsoft's own support guidance is that Foundry has no free option and bills from the first request once you are on pay-as-you-go. Signing up needs a card; Microsoft says that is a small temporary authorisation and that you are not charged unless you choose to move to pay-as-you-go. - You need an Azure account, and to create a workspace inside it you need admin rights over that account. On a personal account you have them automatically. On an employer's account you often do not, and that — not any technical step — is where most people actually get stuck. - Cost is measured in tokens (a chunk of text, roughly a short word — how AI usage gets counted), billed per million, priced separately for what you send and what comes back. Playground poking around costs pennies. The reliable way to sleep at night is to set a budget alert in Azure's cost management before you switch anything on, not after. - Nothing to install — a browser and the portal at ai.azure.com. If all you want is to use OpenAI's models, ChatGPT or OpenAI's own developer platform will get you there in a fraction of the clicks. Security: - Microsoft's position here is unusually explicit, and it is the main reason organisations pick this route over going to OpenAI directly. Its documentation states that your prompts, your outputs, your embeddings and your training data are not available to OpenAI or other model providers, are not used by them to improve their models, and are not used to train foundation models without your instruction. The models themselves are described as stateless — nothing you send is stored inside the model. Security: Two things that stay true anyway:: - Prompts can still be stored and read by a person. Microsoft's abuse-monitoring system holds flagged data in a store kept separate per customer, which authorised Microsoft reviewers can open when the automated check spots a pattern. Only customers managed by a Microsoft account team can apply to have that storage and human review switched off. - Where your prompts are processed is a setting you choose, not a given. If data residency is the reason you are here, that one dropdown is most of your answer. Best at: - Putting OpenAI's models inside an Azure account you already have - Reaching models from other labs — Anthropic, Meta, Mistral, xAI, DeepSeek, Microsoft's own — out of the same catalogue Okay at: - Being anybody's first experience of AI. It is an enterprise console and behaves like one. Avoid: - Reserved capacity while you are still experimenting. It bills hourly for the capacity you reserved whether or not you send a single request, and it cannot be paused — the billing stops only when you delete the deployment. - Expecting to turn the safety filters off. They are applied by default to every deployment across hate, sexual, violence and self-harm categories. - Treating it as a ChatGPT substitute. No consumer app, no chat history to come back to tomorrow, no mobile client — the playground is a test surface, not a place to live. Starter: Switch on one small model and chat with it in the playground Creates a project, turns on your own named copy of a model, and opens a chat box against it — all in the browser, no code. On OpenAI's own platform every model is simply there the moment you hold a key. Here a model does nothing until you switch on a copy of it and give that copy a name — and it is *that* name, not the model's name, that your code has to ask for afterwards. Doing it once by hand is the quickest way to understand why so many Azure error messages amount to "no such model." Accept the default name the first time. Renaming it at this step is where most of the later confusion is born. Starter: Choose the deployment type on purpose rather than clicking past it Makes you read the one dropdown that decides where in the world your prompt is processed. This is the setting people come to Azure for. "Global" hands your request to whichever Azure region has room; "Data Zone" confines processing to the US, EU or Asia Pacific; "Standard" confines it to one Azure geography. OpenAI's own API gives you a single endpoint and no such choice at all. If residency is not a requirement for you, Global is the sensible default and the easiest to reason about — but choose it, don't inherit it. Starter: Find where the safety filter draws its line Shows you what a blocked request looks like, on your terms, before a real user finds it for you. Every deployment gets Microsoft's guardrails by default, and a blocked prompt comes back as an error rather than as a model politely declining. That is a different failure mode from a chatbot saying no, and it is the one that breaks applications written by people who never saw it in testing. You can tune the severity thresholds per category; removing the filters needs an approval only Microsoft-managed customers can request. Use clinical or news-report phrasing about violence or self-harm rather than anything genuinely unpleasant. The point is to see the shape of the error, not to test the limits. Starter: Look up the model's retirement date before you build anything on it Finds, in Microsoft's published schedule, the date the model you just switched on stops answering. Microsoft sets a retirement date automatically when a model launches — currently about eighteen months out — and publishes it, so the end of your model's life is knowable on day one. Standard deployments upgrade themselves to a replacement when that date arrives, unless you have set the version option to never auto-upgrade, in which case they simply stop working. OpenAI retires models too, but it does not hand you a per-deployment switch that decides whether your app quietly upgrades or quietly dies. Leave auto-upgrade on while you are learning. A model that improves under you is a far better surprise than one that stops answering. Pitfalls: - Two portals wearing almost the same name. ai.azure.com carries a "New Foundry" toggle, and Microsoft maintains a whole parallel set of documentation labelled "Foundry (classic)." A tutorial written for one will not match the screen in front of you if you are on the other. Check the toggle before concluding that a step has gone missing. - Hunting for a model that is not offered where you are. Availability is per region and per model. Before deciding something is broken, check Microsoft's region availability list for the model you actually want. Where to next: - More AI APIs and developer services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - Text and conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Beatoven.ai Canonical: https://www.launchpadhq.io/platform/beatoven-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/beatoven-ai.json Last reviewed: 2026-09-17 Tool: Beatoven.ai (https://www.launchpadhq.io/platform/beatoven-ai): starting price Free, from See pricing; free tier yes; data as of not dated Describe the background music a scene needs and it composes one to order — though right now there is no working way in. You describe the music you want in a sentence — what the scene is, the mood, how fast, how long — and press generate. A finished instrumental track plays back a few moments later. You listen, change the wording or adjust parts of it, generate again, and when one fits you download the audio file. What it is: - Beatoven.ai is an AI music generator built specifically for background music — instrumental scoring that sits underneath a video, a podcast, a game or an advert, rather than songs with singing. - What sets it apart is where its training music came from: Beatoven licenses directly from musicians and holds a Fairly Trained certification — an outside badge given to AI companies that train only on material they have permission to use — and says contributing musicians are paid when their work feeds a generated track. If the ethics of AI music are what is stopping you, this is the corner of the category built to answer that. - There are two generators inside it, not one. The newer one, which Beatoven calls maestro, is the kind you write to: you give it a prompt (the message you type) describing the music and it composes to that — we could not confirm what that box is labelled on screen, because the product is unreachable. An older one, listed as Composer in the plan table, works the other way round: you pick settings such as sections, moods, tempo and genre. Beatoven's terms of use still describe that settings-driven flow as how the service works, which measures how recently the text box arrived. Before you start: - As of September 2026 there is no working way in. Every route into the product — the sign-up button, "Try now", the subscribe buttons, the developer dashboard — points at an address called sync.beatoven.ai, and that name no longer exists on the internet's address book: not a slow server, nothing to connect to. The pricing page and the blog return "Page Not Found", and the homepage no longer shows a sign-up button. Beatoven has announced nothing about it, so open beatoven.ai and look for a way to sign up before assuming either that the product is gone or that it is coming back. If there still isn't one, the other music generators on this site are working today and one of those is where to start instead. - When it was last purchasable, generating was not what you paid for — downloading was. Beatoven meters downloads in minutes of finished audio, not tracks and not credits. The last published plan table sold monthly allowances of download minutes, with unlimited generating, and occasional users could buy minutes outright. That table is offline and cannot be re-checked. - The free Trial plan looked like a preview rather than a usable free tier: a small monthly allowance of generations — the table contradicted itself on the number, so call it a handful a month — and no download minutes against it. Beatoven's own help pages describe downloading a licensed track as something paid users do. The table is offline, so this could not be confirmed a second time, but the honest expectation is that you could hear what you made and not keep it. Nothing on Beatoven's public pages mentions marking free downloads either way; on the evidence available the limit is the download meter rather than a mark on the audio, though you may need to check that once you are signed in. - maestro launched without vocals, Beatoven said at the time that singing was still to come, and nothing on the site claims it has arrived — so assume instrumental until the product says otherwise. If you want a song with a singer, this is the wrong tool in the category. Security: - Beatoven's privacy policy is dated 26 August 2022: three years before the text generator the site now leads with, and before the product accepted uploads at all. It covers ordinary things — account details, payments through Stripe, sign-in through Auth0, Google Analytics and Microsoft Clarity for usage tracking — and says your data is stored and processed in India and the United States. The document that actually governs what you upload is not this one; it is the terms of use. Security: What the terms say about anything you upload: - Submitting a file grants Beatoven a worldwide, royalty-free, sub-licensable licence, in perpetuity, to use, reproduce, adapt and exploit it — including for promoting the service. - The same clause says Beatoven may keep what you submitted in its database after you close your account and stop using the service. - Survivable for a rough cut you are scoring yourself; not for a client's unreleased footage, and no setting changes it. Security: - There is nothing sensitive in a sentence describing a mood; the care belongs at the upload box, not the text box. Best at: - Scoring to a written brief — a named scene, one emotion, a pace and an exact running time - Short functional pieces: a 15-second podcast intro, a 45-second cue under a scene, a loop behind a game level Okay at: - Working from a reference instead of a description: Beatoven says you can hand it an audio file or a video, but its marketing pages describe this rather than showing it, so treat the fidelity as unproven until you try it - Fixing a near miss — Beatoven describes an editor where you swap instruments or change tempo, genre and mood by typing, though that is a claim on its own pages rather than something we could see - Anything long: the published examples top out around three minutes Avoid: - Releasing the track as a piece of music in its own right. The terms exclude selling copies of it or putting it on Spotify, Apple Music or any other streaming service, and the plan page carried the same bans in plainer words — adding that a Beatoven track may not be registered as your own work anywhere that records music ownership. - Assuming the licence outlives your subscription. Beatoven's FAQ calls it a "non-exclusive perpetual licence" and a product page calls it "lifetime", but the binding terms grant the right "during the term of this Agreement" and reserve every right not expressly given. Those are not the same promise; what happens after you cancel is genuinely unsettled, and in a dispute it is the terms that get read. - Feeding what it makes into another AI. The terms bar using the music inside any other artificial intelligence service, or to train any AI tool or system — worth knowing if there is an AI editor anywhere in your pipeline. Starter: Score a moment, not a mood This is Beatoven's own published example, and its shape is the lesson: situation, one named emotion, a tempo, an exact running time, then a single constraint on the arrangement. Suno's guidance points you at genre and vocal style because it is building a song; this one describes a moment in someone else's footage. Change only the last sentence and run it again, to hear how much "minimal and piano-driven" alone is controlling. Prompt: I need music for a sunset drive along a coastal road in a short film where the underlying emotion is nostalgia. Music should be slow, flowing tempo. The piece should run 45 seconds. Keep it minimal and piano-driven. Starter: Ask for a podcast intro by the second Beatoven publishes this as a template for podcasters and it is worth copying literally, down to the order of the clauses. Fifteen seconds is also the length where a stock library serves you worst, because nothing in it was cut to your runtime. Ask for the outro of the same show next, keeping every word but "intro" and the duration, so the two pieces share a family resemblance. Prompt: I need intro music for a tech and futurism podcast that feels thrilling. The pace of the music needs to be energetic, and the piece should run 15 seconds. My preferred genre for the music is hip-hop. Starter: Shortlist before you spend a minute Run the same brief three or four times with exactly one word changed each time — the emotion, then the pace, then the instrument — and pick by listening before you download anything at all. On Suno and Udio the meter sits on generating, so hesitation is what costs you. Beatoven puts the meter on the way out instead: here, being indecisive is free and being hasty is the expensive move. Starter: Get a sound effect from the same kind of box Beatoven ships a second generator just for sound effects, maestro Sound Effects, and describes it as taking a written description the same way the music one does. Most AI music tools stop at music and send you to a stock library for a door creak. Name the room as well as the object — half of what a sound effect is, is the space it happens in. Prompt: A heavy wooden door creaking open slowly in an empty stone hallway, with a faint echo. Starter: Keep the licence email with the project Every completed download sends you a licence by email for that specific track, carrying a track ID. Save it into the project folder rather than leaving it in your inbox, and put the ID in the audio file's name so the two never separate. Beatoven's documented procedure for a YouTube copyright claim is to report it to YouTube quoting the track ID from that licence, then file a form on beatoven.ai so the company clears it. That is a per-track escalation path with a reference number, and it only works if you still have the email. If you are scoring a series, keep one folder of licence emails per series rather than per episode. Pitfalls: - You get a licence, not the copyright. Beatoven's terms say that as between you and Beatoven, Beatoven is the copyright owner of the music you generate, and you control only the rights the licence gives you. Its own FAQ says the same thing in plain words. That is a different arrangement from most AI image tools. - There is a credit requirement almost nobody notices. The terms ask you to display "Music by Beatoven.ai" in connection with any use of your project, "wherever practicable". No plan was ever described as removing it. The wording is soft, but it sits in the binding document rather than on the pricing page, which is where people look. - The price has moved a lot, and every figure about this tool goes stale fast. Between early 2025 and mid-2026 the entry subscription became much cheaper while its download minutes doubled. - Anything the site calls "upcoming" should be read as absent. Beatoven's use-case pages still sell paid access to an "upcoming 'Audio to Audio'" feature, and still use an old "pro" plan name that the later pricing table replaced with Creator and Visionary. The marketing pages are the least current on the site. - Two people can be handed near-identical music. The terms say outright that your rights are non-exclusive, that Beatoven keeps using the catalogue its generator was trained on, and that it may licence music to other users that is the same as or similar to yours. The paid plan table advertised an "Exclusive music license" in the same breath, which cannot be true as written alongside the terms; where they disagree the terms are the binding document. That is the ordinary bargain of a generated library, but few vendors write it down this directly. Where to next: - Music generators: https://www.launchpadhq.io/discover?category=music-generation - Video creation and editing: https://www.launchpadhq.io/discover?category=video-creation-editing ### Getting Started with Bloomberg Terminal AI Canonical: https://www.launchpadhq.io/platform/bloomberg-gpt-terminal-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/bloomberg-gpt-terminal-ai.json Last reviewed: 2026-09-12 Tool: Bloomberg GPT / Terminal AI (https://www.launchpadhq.io/platform/bloomberg-gpt-terminal-ai): starting price ~$24,000/year; free tier no; data as of not dated The AI built into Bloomberg's financial data workstation — and why the "BloombergGPT" you read about isn't something you can sign up for. You sit down at a Bloomberg Terminal, type a short code into the box at the top of a panel, press the GO key, and a screen of financial data appears. Newer AI parts let you type a plain-English question instead and read back an answer with the Bloomberg documents it drew on, then ask a follow-up. What it is: - Two different things share one name here. BloombergGPT was a 50-billion-parameter model (the AI "brain" that does the actual thinking) that Bloomberg described in a research paper in March 2023. Bloomberg published the paper, not the model — there has never been a BloombergGPT website, app, sign-up page or download, and Bloomberg's own AI pages today do not mention it at all. - What you can actually use is the AI built into the Bloomberg Terminal, Bloomberg's subscription workstation for finance professionals. As of September 2026 Bloomberg's AI page leads with ASKB, which it calls "a new conversational AI interface" and says is "now in beta". Bloomberg also describes it as coordinating agents (AI that takes actions on its own rather than only answering) to work through multi-step requests, alongside older features: AI summaries on news stories, AI search across filings, transcripts and analyst research, and AI-assisted company financials. - Bloomberg says ASKB runs on "multiple commercial and open weight" models — a mix of outside models and publicly published ones — rather than on a single in-house Bloomberg model. Before you start: - The barrier here is the subscription, not the AI. The Bloomberg Terminal is sold to firms and institutions through a sales conversation: Bloomberg publishes no price anywhere on its site, and every Terminal product page ends in "Request a Demo". There is no free tier, no trial button and no credit-card checkout. Annual figures in the tens of thousands of dollars per user circulate widely in the press and in software-procurement write-ups, but Bloomberg does not publish them, so treat any number you see — including that one — as an outside estimate rather than a price. - There are three realistic ways to end up in front of one: your employer gives you a login; your university or school has a Bloomberg lab (Bloomberg runs a "Bloomberg Terminal on Campus" programme for universities and high schools, built around physical Terminals in a teaching facility); or you contact Bloomberg's sales team yourself. If none of those is true for you today, none of what follows is reachable yet — start somewhere free instead, like our Getting Started with ChatGPT page or the Finance & Real Estate category, and come back if a Terminal ever lands on your desk. - Bloomberg labels ASKB "now in beta" and has not published which subscribers have it, in which regions, or whether it costs anything on top of a Terminal licence. The older AI pieces are described as ordinary Terminal features — Bloomberg announced AI-powered earnings-call summaries in January 2024 and AI-powered news summaries in January 2025. So if the question box isn't in your login, the likely reason is the beta, not anything you did wrong; ask whoever administers your Terminal. - The AI sits on top of a system you still have to learn — the Terminal's own short-code navigation, confirmed with the GO key (Bloomberg's student guide teaches DES for a company description, FA for financial statements, WEI for world equity indices), which Bloomberg describes ASKB as complementing rather than replacing. If you want those basics before you sit down, Bloomberg publishes a self-paced course called Bloomberg Market Concepts; it doesn't state a price on that page. Security: - The privacy question here is the opposite of a consumer chatbot's, and it isn't about the AI. A Terminal licence belongs to one named person and to the firm that pays for it — Bloomberg says access "requires biometric (i.e., fingerprint) and multiple-method authentication" — so the account is identifiable, and it is not really yours. Security: Three things that follow from that:: - Assume your firm can read what you type in the Terminal. Bloomberg sells the product that does it: Bloomberg Vault captures "email, chat, collaboration and voice such as Instant Bloomberg (IB), Bloomberg Message (MSG)" and runs surveillance over the archive. If your employer licenses it, your Terminal messages are compliance records. - Don't share or borrow a login. Bloomberg does sell shared licences that several people log into, which is how most university labs run — if you're on one of those, treat the machine as a public one, and anything you leave on screen as visible to whoever sits down next. - We could not establish what Bloomberg does with what you type into ASKB. Its public AI pages don't say whether conversations are retained or used to improve the service, and its privacy notice is closed to automated readers, so we could not check it. Until you've seen it in writing, follow your firm's rules on putting client, deal or unpublished information into an AI tool — and if there are no rules yet, ask before you're the test case. Best at: - Questions about a specific public company — it is searching Bloomberg's own licensed store: filings, transcripts, Bloomberg News, sell-side and independent research from over 800 providers, plus in-house research from Bloomberg Intelligence, BloombergNEF and Bloomberg Economics. - Showing its work. Bloomberg says ASKB answers come with "transparent attribution to original research documents and news sources". - Reading a stack of long documents quickly, rather than one at a time. Okay at: - Being a general assistant. It is pointed at company and markets research; ordinary writing, coding and everyday questions are not what it's for. - Being finished. Bloomberg calls ASKB a beta — a reason not to rebuild your week around it yet. Avoid: - Justifying a Terminal with the AI. The AI sits on top of a data and news subscription; if you don't already need the data, nothing here is worth the entry price. - Paying anyone for "BloombergGPT access" — a course, a reseller, an API key, a "private endpoint". Whatever is being sold, it isn't the thing named on the tin. - Acting on a generated summary without opening what it cites. The attribution is there because a summary is a compression, and a compressed earnings call is exactly where the qualifier that mattered gets dropped. Starter: Ask a plain-English question about one company Swap test against AlphaSense, which also searches transcripts and broker research: the question is the same there. What differs is what the answer gets checked against — Bloomberg says its AI-assisted Company Financials lines reported numbers up against consensus estimates, and the consensus set on a Terminal is Bloomberg's own rather than a licensed third party's. You are paying for the estimates, not for the summarising. Run the identical question against a competitor's ticker; a like-for-like read is more useful than a single company's answer. Prompt: What were the main bullish and bearish takeaways from [Company]'s most recent earnings call, and how did guidance change from the prior quarter? Starter: Describe a piece of work, not just a question Bloomberg's AI page names "pre-earnings prep" and "post-earnings analysis" as multi-step activities you can describe in one request. That is the mechanism worth learning here, because it collapses a dozen one-code-one-screen lookups into a single instruction. Add the constraint that matters to you — "in bullet points", "only what changed", "assume I've read last quarter's release" — before you ask it to go longer. Prompt: Prepare me for [Company]'s earnings next week: what consensus expects, what has changed since last quarter, and the three things analysts pushed management on last time. Starter: Ask for a chart and keep the code underneath it Koyfin will draw you that chart too, and for free. What it can't hand back is the query that produced it: Bloomberg says ASKB surfaces the underlying Bloomberg Query Language (BQL) code when it pulls data, which turns the chart from a picture you have to trust into a query you can read, correct and run again. If a number looks wrong, read the code rather than re-asking the question — it shows which field it actually pulled. Prompt: Chart [Company]'s quarterly revenue and gross margin for the last five years, with the sector median alongside for comparison. Starter: Read the summary already sitting on a news story Bloomberg News stories inside the Terminal carry an AI-generated summary at the top. Bloomberg launched this in January 2025 as AI-Powered News Summaries — "three succinct bullet points at the top of Bloomberg News content" — and now describes AI summaries drawn from more than 30,000 sources. You don't ask for it and there's no box to type into: it's there when you open the story, on the desktop Terminal and in the Bloomberg Professional mobile app. It needs no new habit and no beta access, which makes it the honest place to start looking. Against AlphaSense, the difference is whose text is being compressed — this is Bloomberg's own newsroom copy summarised in the same window it was published in, not a third-party pass over other people's coverage. Starter: Put a stack of documents in front of it Document Search covers "hundreds of millions of documents, including company transcripts, sell-side research, news and proprietary Bloomberg content". Document Workspace is the companion for going deeper on a set you've chosen: Bloomberg describes it as extracting structured insights from multiple documents at once and laying them out for comparison. Doing this in the Terminal rather than uploading PDFs to a general chatbot changes who supplies the documents. Research from over 800 providers is already licensed and indexed here, so you are searching material you could not legally have assembled yourself. Start with two documents you already know well, so you can judge whether the comparison it produces is actually right before you rely on one you haven't read. Pitfalls: - Reading a blank answer as "there's no data". Bloomberg's AI features are named and scoped, while the Terminal itself has thousands of functions they don't reach. A question that draws a shrug may simply be outside what the AI covers rather than unanswerable on the Terminal. - Judging the whole thing on the newest part. ASKB is the piece Bloomberg is still building; the news summaries and document search have been ordinary Terminal features for longer. If the conversational answers disappoint you, that tells you about the beta, not about the rest. Where to next: - Finance & Real Estate AI: https://www.launchpadhq.io/discover?category=finance-real-estate-ai - Research & Academic Tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Bolt.new Canonical: https://www.launchpadhq.io/platform/bolt-new/guide Data: https://www.launchpadhq.io/data/v1/guides/bolt-new.json Last reviewed: 2026-09-08 Tool: Bolt.new (https://www.launchpadhq.io/platform/bolt-new): starting price Free, from $25/mo billed monthly; free tier yes; data as of 2026-10-25 Describe an app in plain English and watch it get built in a browser tab — on a free plan with rules. You open bolt.new in a browser and type what you want built. Bolt writes the app and a working version appears in a preview panel beside the chat a few minutes later. From there you keep talking to it — describe a change, or click the thing on screen you want different — and it rebuilds in place. What it is: - Bolt is an AI app builder from StackBlitz: you describe a website or web app in plain language and it writes the whole thing — the screens people see and the storage behind them — and puts it on a live web address. - Nothing is installed: it builds and previews your project inside your own browser tab rather than on a distant server, so a Chromebook is enough. - Your allowance is measured in tokens (a chunk of text, roughly a short word — how AI usage gets counted), and Bolt's documentation says most of yours goes on Bolt reading and syncing your project files. So the same request costs more in week three than on day one: there is more project to read. - It builds in JavaScript, the language of the web; back ends written in Python or PHP are not supported. Before you start: - Free, and no card: sign in with Google, GitHub or email. The free plan gives a daily token allowance under a monthly ceiling, and unused daily tokens never carry over. - That is enough for one small app and a few rounds of changes, not a full day of building. - Pro is a monthly plan with a much larger token allowance, no daily cap and unused tokens rolling over. Bolt states that subscriptions are non-refundable, so do your deciding on the free plan. - Publishing is included, but free hosting is capped at a monthly visit limit Bolt sets, and free sites carry a "Made in Bolt" badge. Past the cap the site goes offline until the month resets. Security: - Bolt's privacy policy (updated May 2026) says StackBlitz may use what you type and what the AI writes to improve the service, on anonymised data, and sends portions of any files you attach to outside AI providers. Its own instruction is the plainest summary: "Do not include secrets, API keys, passwords, or confidential information in prompts, source code." Security: Before anyone else uses what you built: - An app the AI wrote is not an app anyone has checked. Bolt ships a security audit for exactly this — the full project audit is on paid plans, the database check is on every plan. - Run it from Bolt's own button rather than asking for it in the chat: Bolt notes that a chat-requested audit decreases your token balance. Security: - Ownership is refreshingly plain: Bolt's documentation states that all code you create with Bolt and StackBlitz is your own, usable for any legal purpose including commercial ones. Best at: - One sentence to a clickable, working web app in a sitting - Small tools that store things — trackers, booking pages, internal lists - Changing wording and look by clicking the preview, not describing it - A shareable web address with no hosting settings to touch Okay at: - Phone apps — built through Expo, but a store listing needs paid Apple and Google developer accounts - Long-lived projects, where each message costs more as the project grows - Predicting spend: building and site traffic are metered separately Avoid: - Anything needing a Python or PHP back end — Bolt only builds JavaScript ones, so that project can't be done here at all - Real customer data before a security audit, whose full-project form means a paid plan - Subscribing in order to try it: Bolt states subscriptions are non-refundable Starter: Build one small, whole thing first Bolt spends tokens reading and syncing your project files, so the first build on an empty project is the cheapest one you will ever run. Something small but complete leaves most of the day's token allowance for changes. Keep version one to a single screen. Prompt: Build a simple reading log: one page where I can add a book, mark it reading or finished, rate it out of five, and see what I finished this month. Starter: Stack up the cosmetic changes, then save once Lets you point at things in the preview and change their text, colour, size and spacing. Bolt's documentation is specific: visual edits are free to make, and tokens are spent only when you save. Ten fixes made in one pass and saved together are charged once, not ten times. Finish the whole wording-and-colour pass before pressing save. Starter: Keep working after the daily limit Bolt's tokens page says that once the day's limit is gone you can still edit your project's existing code in Code view, which doesn't use tokens — so a typo or a wrong colour is fixable at 4pm. Code view is read-only in Safari; edit in a Chrome-family browser. Prompt: Show me the code for the page I just built. Pitfalls: - The daily and monthly limits are separate, and neither banks: a few heavy days can end your month early even if the rest of it was light. - Clicking "Attempt fix" over and over. Bolt warns against this by name, and each attempt spends tokens; rolling back through version history costs nothing. - Asking in the chat for something that has a button — a chat-requested security audit spends tokens, the button's version doesn't. Where to next: - AI Coding & Development: https://www.launchpadhq.io/discover?category=ai-coding-development - Workflow & Automation: https://www.launchpadhq.io/discover?category=workflow-automation - AI Chatbots & Assistants: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Canopy Canonical: https://www.launchpadhq.io/platform/canopy-tax/guide Data: https://www.launchpadhq.io/data/v1/guides/canopy-tax.json Last reviewed: 2026-09-12 Tool: Canopy (Tax) (https://www.launchpadhq.io/platform/canopy-tax): starting price $50/mo; free tier yes; data as of not dated Firm software for US accounting and tax practices — and the AI now doing work inside it. Canopy is a website your firm logs into. You click into a client's file and, in its AI thread, type what you want done — a prompt is just the message you type. What it is: - Canopy is practice-management software for US accounting and tax firms — one website holding the client records, documents, e-signatures, client portal, messages, task lists, time tracking and invoices that a firm otherwise scatters across five tools. It is a firm's filing cabinet and to-do list, not a chatbot you visit. - The AI is layered on top of that filing cabinet, and that is the whole point of it. Canopy Coworker — the AI you would actually use — is an agent (AI that takes actions on its own rather than only answering), and because it sits inside the system that already holds your clients, documents and deadlines, "chase everyone who hasn't sent their paperwork yet" is a sentence it has something real to act on. Before you start: - There is no price-and-buy button. Canopy's site offers a demo, a self-guided tour and a login — the buttons currently read "Get a Demo", "Take a Tour" and "Log In". There is also a multi-step "Get started" form that asks about your firm, but Canopy's own pricing FAQ says a trial is arranged by scheduling a demo and talking with a sales representative. The realistic first step is either working at a firm that already pays for it, or requesting a demo from getcanopy.com/pricing. - It costs money on day one and there is no free tier. Canopy publishes per-user monthly prices for its entry and middle plans, both billed annually; a third plan is marked "coming soon", and the largest firms get a custom quote. Canopy's terms of service, last updated 29 May 2026, say fees are non-refundable and that a plan auto-renews unless you cancel before the renewal date. Canopy names solo practitioners among its customers and publishes no minimum seat count, though we could not confirm one either way. - It is built for US firms. Canopy describes itself as built for U.S. accounting, CPA, tax and bookkeeping firms, its data is processed and stored in the United States, and its own feature table lists multi-currency and regional compliance as still to come. If your practice is outside the US, this is not your tool yet. - Some of the AI runs on an allowance, and two allowances that look alike behave differently. In Canopy's plan-comparison table, the AI meeting notetaker shows an hours allowance on each plan, larger on the middle one, footnoted "Hours per user/per month" — that one refills. Smart Intake shows a small number of credits on every plan, footnoted "Trial Credits. More can be added to meet the firms needs" — those are a one-off sample you buy more of, not a monthly refill. Credits are the platform's unit of spend: each thing you run costs some. - The AI you would sign up for is still being tested, and Canopy's own pages disagree about how plainly to say so. Its plan-comparison table marks both Canopy Coworker and Smart Tax Prep "Beta" on every tier, while its AI feature page presents them as part of the platform without using the word. Canopy's Coworker page says Coworker is free to customers during its open beta period — which also means today's price is not necessarily next year's, and Canopy has not published what it costs afterwards. If your firm is weighing Canopy for the AI, get that in writing from the salesperson. Security: - The sensitive data here isn't yours — it's your clients'. Tax returns, bank statements, Social Security numbers, things people hand an accountant and nobody else. That changes the question from "what should I avoid typing in" to "what did the firm agree to on every client's behalf when it signed?" Two Canopy documents answer that, and they do not obviously agree. Security: What each document says, and the date on it: - Canopy's AI governance page (last modified 10 July 2026) says: "We never use your data for training or testing AI models unless you explicitly opt in," and separately that "Your data is never used to train general-purpose models built by vendors like OpenAI or Anthropic." Opting in is a choice between testing-only and testing-and-training. - Canopy's terms of service (last updated 29 May 2026) say you "grant to Canopy permission to use, on an aggregated and de-identified basis, any Client Content or other information that Canopy learns, acquires, or obtains in connection with these Terms, for the purposes of providing and improving the Service and Canopy's products and services" — that is client material with the names stripped out and pooled with other firms', used to improve the product, and it is not described as needing a separate opt-in. - The terms are the binding document; the governance page is not. Both statements can be true at once if "training AI models" and "improving the Service" are meant as different things — but Canopy has not published where that line falls. It is a fair thing to make a salesperson answer before a firm signs. Security: Three more things worth knowing before a firm signs: - OpenAI and Anthropic sit behind Canopy's AI features. Canopy's privacy policy (last updated 29 May 2026) names OpenAI specifically as being used to generate emails and automate tasks. - Canopy says personal details are removed "using industry-standard techniques, unless it's essential to completing a specific task." In tax work it often is essential — a document request naming the wrong person is useless — so read that exception as doing real work. - Canopy says it holds an independent audit of how it handles customer data (it calls this SOC 2). Security: - None of this is unusual for firm software, and none of it means Canopy is careless. Best at: - Acting on the firm's own records — its AI can see the clients, documents, tasks and deadlines already inside Canopy. - Turning an inbox-shaped job into a tracked one: chasing missing paperwork, onboarding a new client, auditing a folder of files that nobody has looked at since March. - Meetings — it joins, records, transcribes and writes up a summary with action items in language that assumes an accounting context. - Naming and filing documents automatically. Okay at: - General AI help. Canopy describes its AI as an execution layer inside the firm's own system; it is not somewhere to ask questions, draft unrelated writing or make images, and no seat you buy includes any of that. - Firms that have only half-moved in. The AI works from what is already in Canopy, so a practice still keeping half its documents in email inboxes gets noticeably less out of it than any demo suggests. Avoid: - Relying on it for this season's return preparation. Canopy's own January 2026 announcement describes Smart Prep as a pilot built with a partner called Filed, initially covering basic and advanced individual US tax returns (Form 1040), with broader availability only after the 2026 tax season. - Letting anything the AI drafted reach a client unread. Canopy's terms state plainly that neither Canopy nor anyone associated with it warrants that the Service, or any results generated from it, will be accurate, reliable or error-free. Starter: Chase the clients who haven't sent anything in It leans on the checklist Smart Intake built for that client rather than on anything you paste in, and Canopy's Coworker page says it answers an instruction with a step-by-step plan you must approve or refine before it runs — which is what makes an instruction this sweeping safe to type at all. Change 14 days to 3 and you have a pre-deadline sweep; name a single client instead and you have a much smaller first thing to try. Prompt: Find every individual tax client for tax year 2025 who has uploaded nothing in the last 14 days. Draft a short, friendly reminder to each one naming the specific documents still missing from their intake checklist. Show me the full list to approve before anything sends. Starter: Ask where a job actually stands Canopy's Coworker page describes tracking live multi-step jobs and "Needs Attention" items inside the same conversation thread you typed into, so the answer and the follow-up work stay in one place. Swap "this client" for "my whole client list" to get a standing-position view, then ask it to turn any line you care about into an assigned task. Prompt: Summarise this client's file for me: what we're engaged to do, what's outstanding, who on my team touched it last, and anything that's been sitting waiting on someone for more than a week. Starter: Let the notetaker sit in your next client call It joins the meeting, records and transcribes it, and afterwards produces a written summary with action items, filed against that client's record. The write-up lands inside the same system holding that client's documents and tasks, so its action items become work you can assign to someone, rather than a transcript sitting in a separate notes app that nobody reopens. Try it on an internal meeting first. Recording a client without saying so is a conversation to have before the tool makes it easy. Starter: Send one client a Smart Intake questionnaire The client answers a short set of questions, Canopy builds a document checklist from their particular answers. The checklist comes from that client's own answers rather than one fixed form the whole practice gets, which is exactly what gives the chasing prompt above something specific to name when it writes the reminder. Run it on a client whose paperwork is always late — that is where an answer-driven checklist earns its keep. Starter: Drop a pile of client documents in and let it name them It reads uploaded files and renames and sorts them into the right client folders, instead of leaving "scan_0012.pdf" wherever the client dropped it. This is the one AI row Canopy's plan-comparison table marks simply "Included" on every tier, with nothing metered or labelled beta beside it — making it the piece a firm on the entry plan can actually lean on today. Feed it a messy backlog rather than one tidy file; the value shows up at volume. Pitfalls: - You can't check what changed from outside. Canopy's detailed product updates moved into Canopy Community, which needs a Canopy login to read — so "what happened to the AI last month" is not a question you can answer before you are a customer. - Buying the entry plan and finding the thing you wanted one tier up. Canopy's comparison table puts custom reporting, capacity planning, advanced task views and user roles and access controls above the entry plan. That last one is the one that bites, because it decides who at the firm can see which clients. - Mistaking the tour for your firm. The self-guided tour on Canopy's site asks for an email address and then plays a canned demo — it shows the product working on sample data, not your data and not your plan's feature set. - Landing on the wrong Canopy. The company renamed from Canopy Tax and its site moved from canopytax.com to getcanopy.com, and there are unrelated products called Canopy — check the address bar before you send anyone a link. - Reading "coming soon" as soon. Canopy's own table marks single sign-on, multi-office and entity management, and auto-renewals as coming soon, with no date attached to any of them. Where to next: - Meetings & Notes: https://www.launchpadhq.io/discover?category=meetings-notes - Workflow & Automation: https://www.launchpadhq.io/discover?category=workflow-automation - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Canva AI Canonical: https://www.launchpadhq.io/platform/canva-ai-magic-studio/guide Data: https://www.launchpadhq.io/data/v1/guides/canva-ai-magic-studio.json Last reviewed: 2026-07-13 Tool: Canva AI (Magic Studio) (https://www.launchpadhq.io/platform/canva-ai-magic-studio): starting price Free, from $12.99/mo; free tier yes; data as of 2026-10-26 Canva's built-in AI toolkit, Magic Studio — generate images, write copy, design layouts, and edit photos without leaving your design. You work in Canva as usual, and the AI tools sit inside the editor — type a description to get an image, describe a layout and get a template you can edit, or select part of a photo and describe the change. The result lands on your canvas, where you keep editing it by hand. What it is: - Canva AI is the AI layer inside Canva, the drag-and-drop design tool millions of non-designers already use. Branded Magic Studio, it's a bundle of tools woven into the editor: Magic Design (describe a design, get an editable layout), Dream Lab / Magic Media (text-to-image and video), Magic Write (copy), plus Magic Eraser, Magic Expand, background remover, and a newer conversational "just tell it what you want" mode. It's for anyone making social posts, presentations, or marketing graphics who wants AI help without learning a pro design tool. Before you start: - The free plan is a real design tool (huge template library, collaboration) with limited AI — enough to try Magic Studio, not to lean on it. Pro is a paid monthly plan with a much bigger monthly AI credit pool. - Credits are shared across all AI features and don't top up a la carte — heavy image + copy + editing use burns them fast, then you wait for the reset or upgrade. - One setting to check on day one: your AI training preference (see security). Security: - Good news on rights: you own both what you put in and what you create — Canva makes no copyright claim over your inputs or outputs — and AI-generated designs can be used commercially on every plan, including Free. Security: Two things to set and know: - AI training is on by default for Free and Pro users. Canva may use your content to help improve its AI unless you opt out in privacy settings. (Teams/Enterprise content is never used for training.) If your designs are sensitive, flip that off first. - Canva's IP indemnity (Canva Shield) is Enterprise-only — so unlike Adobe Firefly's paid indemnity, a Pro user carries the residual legal risk themselves. Keep commercial prompts clean and generic. - The usual generator caveats apply: AI images can occasionally resemble existing work, and a pure AI image may not be copyrightable — review outputs before high-stakes commercial use. Security: - There are no built-in approval controls on individual plans — anyone with access can generate and publish — so if you're producing for a brand or client, put your own review step in front of anything that goes public. Best at: - Turning a prompt into a finished, editable design fast - One-place workflow: generate, write, edit, lay out, export - Social posts, presentations, and marketing graphics for non-designers - Quick photo edits — background removal, object erase, expand Okay at: - Best-in-class image quality (Midjourney/Firefly lead) or long-form copy (dedicated writers lead) - High-resolution / print output (AI images default to lower res) - Brand consistency without setting up a Brand Kit first Avoid: - Assuming your designs aren't used for AI training (opt out first on Free/Pro) - High-stakes commercial work without a review step (no indemnity below Enterprise) - Prompts built on real brands, characters, or people Starter: Design from a description Magic Design turning a plain description into an editable layout is the fastest "wow" and the most useful everyday move. Add your brand colors in the prompt, or set up a Brand Kit first for on-brand results. Prompt: Make an Instagram post announcing a weekend coffee-shop sale: warm, cozy, hand-lettered feel, space for a short headline. Starter: Generate an image inside your design Shows Dream Lab dropping a generated image straight into your canvas — no downloading and re-uploading. Regenerate a few times; each attempt uses credits, so refine the prompt before spamming generate. Prompt: A friendly illustration of a steaming coffee cup with little hearts rising from it, flat modern style, warm colors. Starter: Fix a photo in seconds Upload a photo, then use Magic Eraser to remove an unwanted object or Magic Expand to widen the frame. These edits used to need Photoshop skills; doing them by describing the change is Canva AI's real time-saver. Works best on clean, well-lit photos with clear backgrounds. Pitfalls: - AI training is on by default (Free/Pro). Opt out in privacy settings if you don't want your designs improving Canva's AI. - Credits are shared and burn fast. Images, copy, and edits all draw from one pool, with no a la carte top-ups. - Indemnity is Enterprise-only. A Pro user owns both the design and the residual legal risk — keep prompts clean. Where to next: - Image Generation & Editing: https://www.launchpadhq.io/discover?category=image-generation-editing - Video Creation & Editing: https://www.launchpadhq.io/discover?category=video-creation-editing ### Getting Started with CapCut Canonical: https://www.launchpadhq.io/platform/capcut/guide Data: https://www.launchpadhq.io/data/v1/guides/capcut.json Last reviewed: 2026-10-15 Tool: CapCut (https://www.launchpadhq.io/platform/capcut): starting price Free, from $12.49; free tier yes; data as of 2026-10-14 A free-to-start video editor from the makers of TikTok, with AI captions and AI video built in — and terms worth reading before you upload your face. You add your clips, photos or a recording, and they line up in order along a strip at the bottom of the screen. You trim, rearrange and layer text or music by clicking and dragging, watching the result in a preview window above. When it looks right, you export a finished video file or post it straight to social media. What it is: - CapCut is a video editor made by ByteDance, the company behind TikTok, and it is built around the short, vertical videos people post to TikTok, Instagram and YouTube. It runs in a web browser, as a desktop app and on phones, and the three versions share the same basic layout, so what you learn on one carries over. - Most of CapCut is ordinary editing — cutting, adding text, music and transitions by clicking — with AI tools sitting alongside as buttons. The ones a beginner reaches for first are automatic captions (CapCut listens to the speech in your video and types it out as on-screen text, timed to the words), text-to-speech, background removal and noise clean-up. - It also makes video from scratch. Its AI video maker turns a script you paste in into a narrated video, and its AI video generator turns a written description into a short clip using ByteDance's Seedance models (the model is the AI "brain" that does the actual work). CapCut currently names two: Seedance 2.5 and Seedance 2.0 Mini. - Why a beginner picks it over a desktop editor such as DaVinci Resolve: CapCut is designed to get a phone-shaped video posted quickly, with templates you fill in and captions in a couple of clicks. It gives up fine control to get there, which is the right trade for social clips and the wrong one for a long film. Before you start: - You can start without paying: go to capcut.com and choose to edit online, or download the desktop or phone app. The free plan covers CapCut's free templates and basic editing tools. Paid plans are where AI credits (the platform's unit of spend — each AI thing you make costs some) come from: Pro is listed at $19.99/month and Ultra, with many more credits, at $79/month. There is also a Teams plan priced per seat. CapCut's own pricing page says its displayed prices are "for reference only" and that you see your final price after signing in. - Which individual AI tools work on the free plan is not spelled out in CapCut's plan list, and CapCut's own pages don't agree with each other about it. Expect some AI buttons — especially anything that generates video — to ask you to upgrade or spend credits. If you only want to trim clips and add captions to the odd social post, start free and see where you hit a wall before paying. - In the US, the company you are signing up with is no longer just ByteDance. Since January 2026, CapCut's US Terms of Service name TikTok USDS Joint Venture LLC — the US joint venture set up so TikTok could keep operating here — as the operator, and paid features in the US are provided by a sister company, TT Commerce & Global Services. CapCut was briefly switched off for US users in January 2025 and kept off the app stores for about a month; it is available normally today. CapCut is for ages 13 and up, and under-18s need a parent's consent under its terms. - A paid plan renews itself. CapCut's terms say a subscription renews for the same length as the last one unless you cancel before the current period ends, and that a full refund is available within 14 days of starting — but only if you haven't used the paid features. Security: - The thing to know about CapCut is in its Terms of Service, not its privacy settings. The US terms (last updated 15 April 2026) say CapCut doesn't own your videos — but by submitting them you give TikTok USDS Joint Venture, its affiliates and 'other connected third parties' an 'unconditional, non-exclusive, royalty-free, fully transferable, sub-licensable, perpetual and worldwide license to use your User Content.' A separate clause covers your username, image and likeness. The terms say this is subject to your platform settings, but they don't spell out what those settings change. Security: What that means in practice:: - Perpetual and sub-licensable are the two words to notice: the licence has no end date, and CapCut may pass it on to others. Uploading a clip is not the same as only storing it. - Think twice about uploading other people's faces, especially children's, or footage a client or employer owns. The licence you grant covers what you upload, and you can't grant rights to footage that isn't yours. - This wording drew wide criticism in June 2025. CapCut's response at the time was that the clause hadn't changed and applies only as far as needed to run the service — but the terms text itself carries no such limit. Security: - CapCut's US privacy policy (updated 15 April 2026) also says it may use your information to 'train, test, and improve technology, such as machine learning models,' including by scanning and reviewing the content you add — and it counts the prompts and files you submit to its AI tools as part of that content. Best at: - Captioning a talking-to-camera clip fast: captions appear in the same place — Text, then Auto Captions — on web, desktop and phone. - Short vertical videos for TikTok, Reels and Shorts, including from templates you just fill in. - Quick clean-ups a beginner would otherwise skip: removing a background, reducing noise, adding a computer voice. Okay at: - Making video from nothing. The AI tools will produce a clip or a narrated video, but they spend credits, and what comes out usually needs editing in the main editor afterwards. - Longer projects. CapCut Web caps how long an upload can be for captioning, and the whole design leans towards short clips. Avoid: - Using CapCut's music or stickers in anything that sells something without checking the mark first. Its Materials License Agreement (last updated 22 January 2026) limits ordinary library Sounds to personal, non-commercial use; only items carrying a 'commercial' mark are licensed for ads, and even then not on merchandise or as a logo. - Uploading footage you'd want to keep exclusive — a client's product launch, a paid shoot — given the perpetual, sub-licensable licence in the terms. - Treating a project stored only in CapCut's cloud as your only copy. In January 2026 a US data-centre power failure left exports failing for more than a day. Starter: Caption a talking clip, then fix what it misheard Film 30 seconds of yourself talking. In CapCut open Text, then Auto Captions, pick the language and generate. Then click any caption block to correct the words before you export. CapCut's own help page says its captions can mishear background noise, accents and fast speech — and that on the phone app you can't re-run recognition on just one bit; you delete that caption block and add captions again from the audio you want. Knowing that before you film saves a redo. Record somewhere quiet. CapCut says its web captioning works best in Chrome or Edge with clear audio. Starter: Find the "commercial" mark before you pick a song Before adding music or a sticker to a video that promotes a business — even a friend's bakery — look for CapCut's 'commercial use' mark on the item. Anything without it is for personal posts only. That split comes from the Materials License Agreement described above, and its licence is granted per exported video, so re-editing and re-exporting something can mean the material has to be licensed again. Starter: Turn a script into a narrated video CapCut's AI video maker takes a script pasted into its 'Enter script' box, lets you choose a style, a voiceover and the shape of the video, and builds scenes to match — then hands the result back to the normal editor, where you can add music and captions and export. You are editing its first draft, not accepting it. Keep the script short and in plain sentences — each sentence tends to become a scene. Prompt: Three ways to keep houseplants alive through winter. One: move them away from cold, draughty windows. Two: water less often — they grow more slowly when the days are short. Three: wipe the dust off their leaves so they catch more of the light there is. Starter: Describe a short clip and let Seedance make it CapCut's video generator runs on ByteDance's Seedance models, and CapCut's own guidance is to keep these clips around 5 to 12 seconds and to describe camera movement and lighting, which is why the prompt (the message you type) names both. Swap the coffee for anything you want to see, but keep 'slow push-in' or another camera move — CapCut's tips lean on them. If the generator offers a choice between Seedance 2.5 and Seedance 2.0 Mini, try the same prompt on both before settling on one. Prompt: A slow push-in on a steaming cup of coffee on a wooden table beside a rainy window, soft morning light, calm and cosy, vertical video, about 8 seconds. Pitfalls: - Don't take a CapCut tool page you found in search as the word on what's free. As noted under Before you start, CapCut's pages disagree; believe what the app tells you when you press the button. - The price you see before signing in may not be the price you pay. Check the final figure on the checkout screen before you confirm. - Exporting isn't the same as having a copy. CapCut keeps projects in its own cloud space. Download finished videos — and keep your original footage — on your own device, so an outage or a cancelled subscription doesn't take them with it. - Using a free template doesn't make its music yours. Templates often come with a sound attached; the commercial-use rule in starter action 2 applies to it too. Where to next: - Video creation & editing tools: https://www.launchpadhq.io/discover?category=video-creation-editing - Image generators: https://www.launchpadhq.io/discover?category=image-generation-editing ### Getting Started with Captions Canonical: https://www.launchpadhq.io/platform/captions/guide Data: https://www.launchpadhq.io/data/v1/guides/captions.json Last reviewed: 2026-10-15 Tool: Captions (https://www.launchpadhq.io/platform/captions): starting price Free, from $24.99/month; free tier yes; data as of 2026-10-14 Turn a phone video of yourself talking into a captioned, edited clip ready to post. You open the app on your phone or in a web browser, record or upload a clip of yourself talking, and Captions writes your words onto the screen as you speak them. You then tap a tool, or type an instruction such as "cut the pauses", watch the video change, fix anything it got wrong, and export the finished file. What it is: - Captions is a video editor built around one kind of footage: a person talking to the camera. It listens to the clip, puts the words on screen, and can cut out the pauses and "ums", tidy the sound, translate the captions and add music and extra footage around you. - It is made for the phone first. The iPhone app gets the full toolset, there is a lighter Android app called Captions Lite, and on a computer you use it in a web browser. - Its other half is AI-generated people. You can pick a ready-made avatar (a computer-generated presenter who speaks a script you type), describe one in words, or build an "AI twin" of yourself from a short recording, then have it deliver lines you never filmed. - The company behind the app is NOCAP, Inc. d/b/a Captions. Before you start: - You need to be 18 or over: Captions' terms bar anyone younger from using it beyond viewing the website. Sign up at captions.ai or in the iPhone app with Google, Apple, an email address or a phone number, and always sign back in the same way — the help centre warns that using a different sign-in method creates a separate, empty account. - There is a free plan, but it is narrow: one caption template and no AI credits (the platform's unit of spend — each thing you make costs some). That is enough to caption a clip. Anything the AI generates or edits for you — the chat editor, AI Edit, avatars, an AI twin, generated music or images — draws on credits, so in practice it means paying. The first paid plan, Max, is $24.99/month; Frontier is $69.99/month, and larger tiers cost more for bigger credit allowances. Captions says its listed prices reflect its iPhone plans. - Credits are used up when the AI generates something, not when you export, so a draft you throw away still costs what it cost. You can check what you have left under Account Settings, and on iPhone a credits history shows where each one went. - Some features — the built-in camera with a scrolling script to read from, for example — are documented as iPhone-only, and Captions' help centre says the web version is in English only. Security: - Captions' privacy section is mostly about your face and voice, because that is what you feed it. Its privacy policy (effective 10 November 2025) says it uses what it collects to "train our artificial intelligence/machine learning models", that it may generate information from your face or voice for some features, and that it may disclose information to third parties "for their own business and commercial purposes, including … training their algorithms." Its pricing page lists "Training data exclusion" as a feature of the Enterprise plan. The policy we read does not describe a way for an ordinary account to switch training off. Security: What the terms say about your face and voice (Captions Terms, effective 4 September 2025): - You own what you upload and what you make, but you give Captions a licence to store, change and use both "to improve the Services, and to develop new services and products". That licence continues after you leave. - The one specific promise is about your voice: Captions "will not commercialize your voice on a standalone basis without your written permission." The terms make no matching promise about your face. - You must have the written consent of every identifiable person in what you upload to use their name, likeness or voice. An AI twin of a friend or partner needs their written OK, not just a nod. - The usage policy (effective 4 September 2025) forbids presenting AI-made content as made "solely by a human", impersonating anyone, and making anything that relates to or appears to represent people under 18. Security: - Captions' privacy policy says it deletes face and voice measurements once their purpose is served, or three years after your last interaction with it, whichever comes first, and that your information may be transferred to and stored in the United States. Security: - One contradiction to know about if you'll use it for business: the help centre says you "retain full rights" to generated videos, but the terms (section 3.1) bar "making any commercial use of the Services including the Mirage Content" without written consent "except as authorized hereunder". The terms are the binding document. If you plan to run paid ads with it, read that section yourself first. Best at: - Captioning a talking-to-camera clip from your phone, with styling, and translating those captions into other languages. - Cleaning up a casual recording: cutting pauses and filler words, removing background noise, and making your eyes look at the lens. - Short vertical videos for TikTok, Reels and Shorts, where one person talks and the edit does the rest. Okay at: - Longer videos. On iPhone the manual editor stops at about 10 minutes and AI Edit at about 2, so it suits clips, not a full YouTube episode. - Footage with more than one speaker. AI Edit's own guidance asks for a single person talking. - Working on a computer. Captions' own feature table lists many tools as iPhone-only, so check that what you need is in the browser version before you subscribe for desktop work. Avoid: - Anything featuring your children, or anyone who looks under 18 — the usage-policy ban in Security, which the app's marketing doesn't mention. - Promoting a bar, a vape shop, a gun range, a dating service or anything involving gambling. The usage policy bars output containing alcohol, tobacco, nicotine, firearms, gambling, dating services and similar goods, so a local business in those trades can't use it for ads. Starter: Caption a clip you already have Import a 30-to-60-second phone video of you talking, choose the language being spoken, and let Captions put the words on screen. Then tap any word it misheard and correct it. With one caption template and no AI credits on the free plan, captioning and correcting a clip you already shot is the full free test before you spend anything. If the transcription is good on your voice and accent, the rest of the app is worth trying. Record in a quiet room and compare how much less you need to correct. Starter: Ask the chat editor for several changes at once Captions charges the chat editor by the message and by each action it takes, so one message asking for three changes uses fewer credits than three separate messages. Its help centre also says you can keep refining with follow-ups like "make it softer" — each of those is a new message. Once it's done, reply with one follow-up, such as "music a bit louder", and look at your credit balance before and after. Prompt: Cut out the long pauses and the ums, add calm background music quietly under my voice, and make the captions bigger and yellow. Starter: Let AI Edit cut a raw clip for you Upload an untouched clip of one person talking, pick an editing style, and Captions cuts it, adds captions, music and extra footage, and hands it back for you to adjust. Captions' instructions for AI Edit ask for footage under about three minutes that has not been edited — no trims, transitions or music. Tidying it up yourself first takes away what the tool works from. Run it twice on the same clip with two different styles and keep the one that fits how you talk. Starter: Describe an avatar before you make one of yourself Captions' avatar guide asks for "one flowing scene description" covering appearance, how the person faces the camera, and the setting and lighting, and says to match the avatar to the voice you'll use. Starting with a made-up person lets you see how avatars look and move before you decide whether to record your own face for an AI twin. Change one thing — the setting, or the age — and regenerate to see what moves. Prompt: A woman in her forties with short grey hair and a navy blazer, facing the camera and smiling slightly, standing in a bright kitchen with soft morning light, medium close-up. Starter: Read a script, then fix where you were looking On iPhone, record with Captions' camera while your script scrolls on screen, then run Eye Contact so your gaze looks as if it is on the lens rather than on the words. Captions puts a scrolling script and an eye-direction fix in the same app, so you can read your lines and still appear to be talking to the viewer. Record the same 20 seconds with and without the fix and watch them side by side. Pitfalls: - Logging out wipes your projects. Captions' help centre says signing out deletes every project and caption template stored on that device, and that clearing your browser cache does the same on the web. Its changelog reports project syncing since early 2026, but the warning is still up — export anything you want to keep before you log out or switch devices. - Signing in a different way from last time. The second, free account it creates can make it look as though your subscription has vanished. - Deleting the app, or even your account, does not cancel a subscription. Cancel where you paid — Apple's subscriptions page if you bought on iPhone, Captions' billing portal if you bought on the web — at least 24 hours before it renews. Monthly plans are non-refundable. - Cancelling also forfeits your credits. Whatever you have left disappears at the end of the billing period, so use them before the date, not after. Where to next: - Video creation & editing tools: https://www.launchpadhq.io/discover?category=video-creation-editing - Voice & speech tools: https://www.launchpadhq.io/discover?category=voice-speech - Image generators: https://www.launchpadhq.io/discover?category=image-generation-editing ### Getting Started with ChatGPT Canonical: https://www.launchpadhq.io/platform/chatgpt/guide Data: https://www.launchpadhq.io/data/v1/guides/chatgpt.json Last reviewed: 2026-06-22 Tool: ChatGPT (https://www.launchpadhq.io/platform/chatgpt): starting price Free, from $8/month; free tier yes; data as of 2026-10-24 A beginner's first 20 minutes with the most popular AI chatbot. You type a message (this is called a prompt — the instruction you give the AI), it types back in plain language, and you can keep the conversation going, including telling it to redo or improve an answer. What it is: - ChatGPT is OpenAI's general-purpose AI assistant — the one most people mean when they say "AI." - It's the best default starting point for a beginner because it does a little of everything competently — answering questions, writing and editing, making images from a description, and helping with code — before you graduate to specialized tools. - For coding specifically, OpenAI also offers a separate, more advanced tool called Codex that can take on a programming task and work through it for you. It's built for developers, not first-timers — but it's there if you grow into building software. Before you start: - Free to use with a sign-up (email, Google, or Apple) — no card required. First step: go to chat.openai.com, sign up free, and you're in. - Will you need to pay? The free tier is plenty for casual use — learning, drafting, brainstorming. You'd only pay (Plus, currently $20/month) if you want the best models reliably or hit the limits often. Paying buys better models, not more privacy. - The free tier has usage limits that reset on a rolling few-hour window. When you hit them, ChatGPT quietly swaps you to a smaller, faster model (the AI "brain" doing the work — there are smarter-but-slower and faster-but-simpler ones). Answers get noticeably shallower with no big warning; if a reply suddenly feels worse, that's usually why. Limits also tighten when OpenAI is busy. - Image generation has its own separate, tighter limit than text — often just a couple a day on the free tier, on a rolling window. You can run out of images while text still works fine, which catches people off guard. Security: - ChatGPT is mainstream and reasonably well-run, but it's a cloud service — and on the free tier, your conversations are used to train OpenAI's models by default. Good rule for any AI chatbot: treat everything you type as if a stranger might eventually read it. Security: Never paste in: - Passwords, bank or card numbers, Social Security numbers - Medical records tied to your name - Other people's private information - Confidential or proprietary work material — several companies have banned employees from doing exactly this Security: Two settings worth knowing day one: - Turn off training under Settings → Data Controls → "Improve the model for everyone" — paying for Plus does not do this for you - Use Temporary Chat for one-off sensitive questions — though "temporary" and "deleted" still aren't instant; OpenAI holds data around 30 days either way Security: - None of this makes normal use unsafe — the box just isn't a vault. Best at: - Brainstorming and getting unstuck - Rewriting and reshaping text - Explaining things at your level ("explain X like I'm new to it") - Drafting first versions you then refine - Making images just by describing them — easy and conversational Okay at: - Math and precise logic — check the numbers - Current events — it can search the web, but it's limited on free, so don't assume it knows this week's news - Coding — competent for small tasks and learning, but dedicated tools do better Avoid: - Anything where being wrong has real consequences — legal, medical, financial; it sounds equally confident when wrong - Facts you can't verify, especially sources and citations — it invents plausible ones - Truly current or niche information on the free tier Starter: The explainer (its best trick) ChatGPT is strongest when you give it topic + audience + length + focus all at once — this prompt does all four. Swap in any topic you want to learn. Prompt: Explain how compound interest works to someone who's never heard of it. Use a simple real-life example, keep it under 150 words, and end with one thing people most often get wrong about it. Starter: The rewriter (format control) "Give me options" works for almost any writing task and lets you pick the tone instead of settling for the first draft. Prompt: Here's a rough message to my landlord about a broken heater: '[paste].' Rewrite it three ways — polite, firm, urgent — each under four sentences. Starter: The thinking partner (back-and-forth) Turning it into a back-and-forth is where a chatbot beats a search engine. Prompt: I'm deciding between [two options]. Ask me five questions one at a time to help me think it through — don't give your opinion yet. Wait for each answer. Starter: The role-player (practice) The same pattern works for language practice or rehearsing any tricky conversation. Prompt: Act as a friendly interviewer for a [retail job]. Ask one common question, wait for my answer, give brief feedback, then continue until I say stop. Starter: The organizer (structure from mess) It's good at imposing structure on chaos — turning a messy list into something usable. Prompt: Turn this brain-dump into a clean weekly plan as a table, grouped by day, with a 'must-do' and 'nice-to-have' each: '[dump].' Pitfalls: - It sounds confident even when wrong — a made-up answer reads as smoothly as a correct one. Verify anything factual. - It invents sources, quotes, and citations — search for the real thing yourself. - The free tier downgrades you mid-session — quality quietly drops when you hit the cap; that's the cause, not you. - It doesn't truly remember — within one chat it follows along, but a new chat is a blank slate, and very long threads "forget" the top. One task, one conversation. - Taking the first answer — the whole point is the follow-up: "shorter," "more casual," "you missed X." Tries two and three are where it gets good. Where to next: - Image generators: https://www.launchpadhq.io/discover?category=image-generation-editing - AI coding tools: https://www.launchpadhq.io/discover?category=ai-coding-development - Research tools: https://www.launchpadhq.io/discover?category=research-academic-tools ### Getting Started with ChatPDF Canonical: https://www.launchpadhq.io/platform/chatpdf/guide Data: https://www.launchpadhq.io/data/v1/guides/chatpdf.json Last reviewed: 2026-09-09 Tool: ChatPDF (https://www.launchpadhq.io/platform/chatpdf): starting price Free, from $5/mo; free tier yes; data as of not dated Drop one long document in and ask it where things are — the least demanding tool in its category to actually try. You drag a PDF onto the page, and it opens next to a chat box. You type a question about the document and the answer comes back marked with the page numbers it drew on. You keep asking follow-ups in the same conversation. What it is: - ChatPDF is a single-purpose website from ChatPDF GmbH, a small German company: you give it one document, you ask it questions about that document, and it does nothing else. - That narrowness is the whole appeal — there is nothing to set up, no library to build first, and no developer step between you and an answer. - It is a reading tool, not a PDF editor. It never changes the file you gave it, and it cannot fill in a form, sign anything, or hand you back a new version. Before you start: - You can try it without signing up. ChatPDF's FAQ says the service "requires no account to get started," and the upload box is the first thing on the homepage. A free account adds saved history and the folders that let you chat with several files at once. - The free plan caps how many documents you can analyse each day. It counts documents, not questions, and nothing on the site states a limit on how many questions each document will take. - The paid plan is sold monthly or yearly. The website publishes no pricing page at all — the only published figure is on ChatPDF GmbH's own App Store listing for its mobile app — so what you meet at checkout on the web may differ from that one. - Files have an outer limit of 2,000 pages or 32 MB each, a figure that appears only in ChatPDF's developer documentation. The website states no separate limit for free uploads, so treat that number as a ceiling rather than a promise. Security: - ChatPDF's terms state that the service is hosted in the United States. The FAQ covers the technical side reasonably: documents are encrypted in transit and while stored, deletion is available at any time, and the storage provider holds SOC 2 Type II certification — an audit of how a company handles customer data. It is the storage provider's audit, not an audit of ChatPDF. Security: - The substance of this page is what the binding documents do not say. ChatPDF's privacy policy was last updated in January 2023 and is a general-purpose template about names, email addresses and billing details — it does not mention uploaded documents anywhere. It sets no retention period for your files, takes no position on whether they are used to improve the service, and names Stripe, its payment processor, as the only outside company involved. The FAQ, meanwhile, says questions are routed between OpenAI's GPT-4o and GPT-4o-mini models (the AI "brain" that does the actual thinking), which means passages from your document are being answered somewhere other than ChatPDF's own systems. None of that is evidence of bad handling. It is a policy that has not caught up with the product, and a thin policy is a fact worth knowing rather than a gap to fill in with optimism. Security: What that means in practice: - Delete a document when you are finished with it. That is the one control ChatPDF actually commits to, and its developer documentation shows a delete endpoint sitting behind it. - No stated retention period is not the same as a short one. Assume a file stays until you remove it. - The privacy policy's line "we do not process sensitive personal information" is describing the name and email address you give ChatPDF. It is not a statement about the medical letter or the contract you upload. - If you would not email the document to a stranger, the question you want answered has to be worth the upload. Best at: - Finding where a long document says something, and getting the page back with it - Getting through a paper or report you were going to have to read anyway, then reading only the parts that matter - Reading a document in a language you do not have — upload in one language, ask in another - Turning a document you were avoiding into a conversation you can start in about thirty seconds Okay at: - Scanned pages, where whether it can read the text at all is the whole question - Numbers lifted out of tables — get the page reference and look - Very long or very heavy files, which run into the page and size ceilings Avoid: - Coursework you intend to hand in. ChatPDF's terms contain an academic-dishonesty clause and reserve the right to report violations to your institution — while its homepage markets answering multiple-choice questions "faster than your classmates." The marketing and the contract disagree, and only one of them is binding. - Anything covered by US health-privacy law (HIPAA) or US financial-privacy law (GLBA). The terms say the service is not built for those rules and that you may not use it if your material falls under them. - Documents someone else made you promise to keep secret. The terms require you to warrant that what you upload does "not constitute confidential information," and separately prohibit uploading trade secrets or confidential documents you are not authorised to share. Starter: Ask where, and make it quote ChatPDF's developer documentation shows its answers carrying inline page markers in the form [P12], so a question shaped like this comes back as something you can go and look at rather than something you have to believe. Swap the three words for whatever you are actually worried about — a deadline, a fee, a named person. Prompt: Go through this document and find every place it mentions cancelling, ending, or renewing the agreement. For each one, give me the page number and quote the exact sentence. If the document says nothing about one of those, say so instead of guessing. Starter: Ask what the summary left out ChatPDF routes queries between the two models named under Security depending on the question, so two answers about the same file minutes apart can come from differently sized brains. A second pass aimed at the gaps is how you notice you got the smaller one. If the "left out" list contains something important, ask for it with page numbers and go read that page. Prompt: Summarise this document in ten bullet points. Then, separately, list three things in the document that your summary left out and that might matter to someone who has to act on it. Starter: Give it the meeting deck, not a PDF The accepted-formats list in ChatPDF's FAQ is longer than the name suggests — Word (.doc, .docx), PowerPoint (.ppt, .pptx), Markdown (.md) and plain text go into the same box as a PDF — so the deck you never opened is a legitimate thing to point it at. Ask it to flag anything that reads like an action assigned to a person. Prompt: This is a slide deck from a meeting I missed. Tell me what was decided, what was only proposed, and what was left unresolved. Give me the slide number for each one. Starter: Read something you cannot read ChatPDF's FAQ states the page references come back the same way when you upload in one language and ask in another — which makes this useful on a document you genuinely cannot decode, not just one you find slow. Ask for the original sentence alongside the translation when a date or an amount matters. Prompt: This document is not in English. Answer me in English: what kind of document is it, who sent it, what is it asking me to do, and what are the dates I need to care about? Give me the page number for each date. Starter: Put the old version and the new one in one folder In ChatPDF the multi-file case works by creating a folder and talking to the folder, and folders are the specific thing sitting behind its free account — so this is the moment where signing up either earns its keep or does not. If the two files are long, ask it to work section by section and tell you when it has finished each one. Prompt: These are two versions of the same agreement. List every substantive difference between them. For each one, tell me which version it appears in and on which page, and say plainly whether the change favours me or the other party. Pitfalls: - Ask a 90-page agreement "is there an auto-renewal clause?" and a confident "no" is not evidence there isn't one — ChatPDF answers from the passages it pulled out of the document, and a clause it never pulled produces exactly the same sentence as a clause that genuinely does not exist. The page markers are the tell: if the answer to a question about a long document cites two pages, it worked from two pages. For anything you would sign, use it to find the clause and then read the clause. - A badly chosen first upload costs you half of the day's allowance — and if there is also a cap on questions per document, you will find it only by hitting it. - A scanned page is a photograph, not text. ChatPDF's mobile app listing advertises reading text out of scans, but the website's own FAQ never mentions it — so on a scanned document, treat a thin or oddly empty answer as a sign it may not have read the page, rather than proof the document is silent on your question. - The reference you get back is a page, not a sentence. ChatPDF's developer documentation returns page numbers, while the marketing describes citations that scroll you to the exact source; on a dense page you may still be the one doing the finding. Where to next: - Other document and PDF tools: https://www.launchpadhq.io/discover?category=document-pdf-processing - Research and academic tools: https://www.launchpadhq.io/discover?category=research-academic-tools - General AI chatbots: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Clarifai Canonical: https://www.launchpadhq.io/platform/clarifai/guide Data: https://www.launchpadhq.io/data/v1/guides/clarifai.json Last reviewed: 2026-09-19 Tool: Clarifai (https://www.launchpadhq.io/platform/clarifai): starting price Free, from Custom; free tier yes; data as of not dated Unreachable since 2026 with no shutdown notice: a model catalogue that let you test AI models in a browser tab and then call the same one from your own code. You sign in, browse a library of AI models (the AI "brain" that does the actual thinking), open one in a browser chat box, type a question, and the reply streams back. When you want that same model inside your own software, the site hands you a ready-made code snippet for the exact model you just tried. What it is: - Clarifai is a catalogue of AI models with a control panel wrapped around it. You search the library, test a model in the browser, and — if you want — call that same model from your own software through an API (a way for programs to talk to each other without a person clicking). - It started in 2013 doing image recognition, and that is still the part it does best. The newer work is about *where* a model runs — Clarifai's servers, a cloud you rent, or your own hardware — rather than about building the models themselves. - One thing to know before you invest time in it. In May 2026 Nebius announced it had hired Clarifai's founder and core engineering team and licensed Clarifai's inference technology. Clarifai's older image models and its government work were explicitly excluded from that deal and stayed with the company. No shutdown notice was ever published, but as of 19 September 2026 the platform is not reachable: api.clarifai.com, docs.clarifai.com and status.clarifai.com have no DNS records, clarifai.com resolves to a registrar parking server with dead HTTPS, and www.clarifai.com points at an ISP address that also fails HTTPS. The company's email still routes through Google Workspace. The newest dated release note before the docs went dark was 12.4, from 7 May 2026. Before you start: - As of 19 September 2026 the API host does not resolve. api.clarifai.com has no DNS record, so nothing on this page — the browser Playground, the code snippet, a Local Runner's cloud address — can currently be reached. Treat what follows as a record of how the platform worked, and check whether the host answers again before you plan around it. - Before the hosts went dark, signup asked for no credit card and the Community plan included a small monthly allowance of operations — one that refilled, not a trial credit, but one shared meter that predictions, training and stored files all drew on. None of that can currently be reached, and whether it returns is unknown. - There is nothing to install to begin. Your first hour is a browser tab. A command-line tool only enters the picture if you decide to try Local Runners, which is the last step on this page and entirely optional. - Clarifai is a developer platform, not a chatbot. There is no assistant that remembers you, no conversation history, no mobile app worth having. If what you want is something to ask questions of, start in Text & Conversational AI and come back here when you want to see how models actually get served. Security: - Clarifai's documentation is unusually direct about the question everyone asks: it states that it does not use your private data to train its models or any other model on the platform unless you explicitly choose to share your inputs and annotations with the Community. So the thing that decides your privacy here is not what you type — it is one visibility setting, and whether you left it alone. Security: What that means in practice: - Sharing to the Community is a licence, not a preference. Clarifai's terms say that by making a repository public you grant every other user a worldwide licence to use and display that content, and grant Clarifai a perpetual, irrevocable right to it. Read that clause before you publish a model or a dataset, not after. - Clarifai calls its credential a PAT. It is an API key (a password that identifies your app, and that spends your money), and it draws on your monthly operations. Keep it out of screenshots, public code and chat messages, and delete and regenerate it the moment you think someone has seen it. - If you try a Local Runner, check whether it is private. Release 12.4 changed the default to private, but the command still accepts a public flag — and "public" here means a public web address pointing at software on your own computer. Best at: - Trying a lot of different models in one place, in a browser, without opening an account with each vendor separately - Image and video recognition - Keeping a model and its data on your own hardware while still reaching it through an ordinary cloud address Okay at: - Being someone's first AI tool. The browser Playground genuinely works without code, but everything surrounding it is built for engineers and reads that way. - Free-tier headroom. The Community plan's small monthly operations allowance, at one request per second, was enough to evaluate the platform and not enough to run anything on — and, as of September 2026, cannot be reached at all. Avoid: - Publishing anything to the Community before reading the licence you grant by doing it. Public content is licensed to every other user and, perpetually, to Clarifai. - Treating the Developer plan's promotional price as the standing rate. The pricing page labels it a promotion, and the plan has no published non-promotional price. Starter: Test a model in the browser before you do anything else Opens a model from the Community library in the Playground — a chat box in the browser — and runs it. No key, no install, no code. Clarifai's free meter counts "operations", and predictions, training and stored files all come out of the same monthly allowance. Testing models in the browser is the cheapest thing you can spend it on; uploading a dataset is the most expensive. Doing it in this order means you find out whether the models suit you while the meter is still full. The Playground shows a generated code snippet for whichever model you are looking at. Ignore it for now — you will want it in step four. Starter: Run a picture through one of the recognition models Uploads an image and returns labels for what is in it, using a model Clarifai trained itself rather than one it is reselling. Image recognition is what Clarifai has done since 2013, and it is the one part of the company the Nebius agreement explicitly left alone — the announcement named Clarifai's legacy computer vision models among the things not included in the deal. If you want to see this platform at its most settled rather than its most uncertain, look here. Use a photo the model should find hard — bad light, an odd angle, something half hidden. Clean stock images tell you nothing about where it breaks. Starter: Optional: point a Clarifai address at a model on your own computer Installs Clarifai's command-line tool and runs a single local-runner command, which gives a model running on your laptop a working Clarifai web address. This inverts the usual arrangement. The model and the data stay on your hardware and Clarifai supplies only the address and the authentication in front of them — hosted services like Replicate cannot do this at all, because running everything on their own machines is what they are. The free plan includes one runner with a short runtime limit, which is enough to watch it work; the paid Developer promotion is what removes the time limit. This is the only step that needs a terminal, and nothing else on the page depends on it, so skipping it costs you nothing. Starter: Make one call from your own code Takes the snippet the Playground generated, drops in a PAT created in your account's Security settings, and runs it once. Clarifai's authentication has a specific trap its own docs had to publish a notice about: to call Clarifai's own models you need a PAT rather than an app-scoped key, and you have to pass the literal values clarifai and main as the user and app IDs instead of your own. Almost every "my Clarifai key doesn't work" is this, and knowing it beforehand saves you the hour. Create the PAT only when you have code to put it in. Everything above this step works without one. Pitfalls: - The free plan allows one request per second. You will never notice this clicking around in the browser, and you will hit it instantly the first time you call the API in a loop. - Use a PAT, with clarifai and main as the user and app IDs — this changed in 2025 and a lot of older tutorials still show the old way. - Don't confuse Clarifai with Nebius Token Factory. Nebius licensed Clarifai's inference technology in May 2026, so the two now describe overlapping capabilities in similar words, but they are separate products with separate accounts and separate bills. Where to next: - More AI APIs and developer services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - Running open models on your own machine: https://www.launchpadhq.io/discover?category=local-open-source-ai ### Getting Started with Claude Canonical: https://www.launchpadhq.io/platform/claude/guide Data: https://www.launchpadhq.io/data/v1/guides/claude.json Last reviewed: 2026-07-27 Tool: Claude (https://www.launchpadhq.io/platform/claude): starting price Free, from $17/month; free tier yes; data as of 2026-10-26 A chat assistant with a genuinely capable free tier — and a training default worth knowing about before you start. You type in plain language and Claude responds, holding the thread of a long conversation rather than treating each message as a fresh start. Attach a file, ask a follow-up, or push back on an answer and it carries on from there. There are no commands to learn; the skill is giving it enough context up front. What it is: - A general-purpose chat assistant you use through a browser, a desktop app, or a phone app. - Strongest on long documents, careful writing, and problems where you want reasoning shown rather than a bare answer. - Free to use with no card required, with paid tiers that buy more usage rather than fundamentally different features. - The same brand covers several products — the chat assistant, a coding tool, and a developer API — which are governed by different terms and have different privacy defaults. Before you start: - The free plan covers web, iOS, Android, and desktop, and includes web search, file uploads, memory across conversations, and the ability to create files and run code. It is not a stripped demo. - Paid plans can be billed monthly or, for less per month, a year up front. The higher tiers buy usage headroom, not new capabilities. - Usage resets on a rolling five-hour window, and paid plans add a weekly ceiling on top. There is no published message count — how far you get depends on conversation length, model, and which features you use. - Set your training preference before your first real conversation. On free and paid personal plans, training is on by default. See the security section. Security: - The thing worth understanding here is not a setting, it's a contract. Anthropic sells access to the same models under two different agreements, and they have opposite privacy defaults. Personal chat plans — Free, Pro, and Max alike — are opt-out for model training: unless you turn it off, new and resumed conversations are used to improve Anthropic's models. Team, Enterprise, and direct API access run under commercial terms, where no training happens by default at all. Security: The numbers behind the toggle: - Training left on: conversations can be retained for up to five years. - Training turned off: the window drops to thirty days. - The toggle sits in your privacy settings and can be changed at any time, though changing it is not retroactive for data already processed. - Conversations you delete are not used for future training. - Incognito conversations are excluded regardless of the setting. - Conversations flagged by safety systems are the exception — they can be retained and reviewed whatever your preference. Security: - That five-year figure is the part to carry away. Retention on assistant products is usually measured in weeks, and leaving this one at its default stretches it to years — a difference of scale rather than degree. Because the same company offers the same models under stricter terms elsewhere, which door you walked through decides your default, not how careful you are once you are inside. Paying more does not change it; changing product does. Best at: - Working through long documents you upload and answering questions grounded in them - Drafting and revising prose where tone and structure matter - Explaining reasoning step by step instead of asserting a conclusion - Sustained back-and-forth on one problem across a long conversation Okay at: - Current events, where it depends on web search rather than its own knowledge - Arithmetic and data work, which is reliable when it writes code and shakier when it does not - Image understanding, which is competent but not its focus Avoid: - Anything requiring certainty about very recent events without checking sources - Legal, medical, or financial decisions taken on its word alone - Confidential material on a personal plan with training left on Starter: Give it a document and interrogate it The single highest-value thing a beginner can do, and the thing generic chatbot advice never mentions. Ask it to quote the exact passage behind each point, so you can check it rather than trust it. Prompt: I've attached a document. Before you summarise anything, tell me what kind of document it is and what it's for. Then give me the three things in it most likely to matter to someone who has to act on it. Starter: Ask it to argue against itself Assistants agree too readily. Forcing the opposing case is the cheapest way to get past that. Follow with "which of those objections is the one I'm most likely to dismiss unfairly?" Prompt: Here's a decision I'm leaning towards: [describe it]. Give me the strongest case against it. Don't soften it, and don't finish by agreeing with me. Starter: Make it ask you questions first Most disappointing answers come from insufficient context, not insufficient capability. If the questions are obvious, your brief was already clear — skip ahead. Prompt: I want help with [task]. Before you produce anything, ask me the three questions whose answers would most change what you write. Starter: Set up a project for recurring work Keeps a set of files and instructions attached to every conversation in that project. Stops you re-pasting the same background every time. Unlimited projects are a paid feature; the free plan can still use one conversation as a long-running thread. Put your standing preferences — tone, format, what to skip — in the project instructions rather than repeating them. Pitfalls: - The training default is on. The reputation for privacy applies to business plans. On a personal plan, your conversations train the model unless you go into privacy settings and turn it off — and doing so also shortens retention from five years to thirty days. - Confident and wrong looks exactly like confident and right. Fluent prose is not evidence. For anything that matters, ask for the source passage and check it yourself. - Long conversations drift. After many turns it may be working from an early version of your requirements. Restate the current goal periodically, or start fresh. - Usage is shared across everything. Chat, desktop, mobile, and the coding tool all draw from one pool. A heavy coding session eats the same allowance as your writing. - Free and paid run the same models. Paying buys usage and extra features, not a smarter assistant. If quality is your problem, better prompting fixes more than upgrading does. Where to next: - Other chat assistants: https://www.launchpadhq.io/discover?category=text-conversational-ai - Build with an API instead: https://www.launchpadhq.io/discover?category=ai-apis-developer-services ### Getting Started with Claude Code Canonical: https://www.launchpadhq.io/platform/claude-code/guide Data: https://www.launchpadhq.io/data/v1/guides/claude-code.json Last reviewed: 2026-09-07 Tool: Claude Code (https://www.launchpadhq.io/platform/claude-code): starting price Pro $17/mo with annual ($20 billed monthly); free tier no; data as of 2026-10-26 Anthropic's coding agent: point it at a codebase, describe the change, and it reads, edits and runs things. You open it inside a project folder on your own computer and describe the change you want in plain English. It reads the files itself, proposes edits, runs commands to check its own work, and asks permission as it goes. You read what it did and say what to fix. What it is: - Claude Code is Anthropic's coding agent — AI that takes actions on its own rather than only answering. Instead of handing you a snippet to paste, it reads, edits and runs a project's files on your machine, across a whole codebase rather than one file. - It runs in a terminal (a text window where you type instructions instead of clicking), in editors like VS Code and JetBrains, and in desktop, browser and phone apps. - Cursor is an editor you work inside; Claude Code is something you hand a task to and check on. GitHub Copilot finishes the line you are typing; this takes the whole change. Before you start: - This is a developer tool, and the entry requirement is a project of code on your own disk that you already work on. Never used a terminal or version control? A browser assistant like ChatGPT writes and explains code with nothing to install, and is the better place to begin. - You need a paid Claude plan — currently Pro, Max, Team or Enterprise — or an Anthropic Console account billed for what you use. Pro is the entry plan and Max costs more for more usage; current prices are on its Launchpad page. - Usage runs on rolling windows rather than a credit balance: a limit resetting roughly every five hours, plus a weekly ceiling. Anthropic describes Pro as suited to light work on small repositories. Hit the ceiling and you wait, or turn on paid usage credits. /usage shows where you are, and long sessions cost more, because each turn resends the whole conversation. Getting set up safely (Anthropic's official Claude Code documentation): - Install only from Anthropic's official installer or package listings — lookalikes exist, and the steps change often enough that blog-post instructions are a bad habit. - Start on a project tracked in version control with nothing unsaved: that tracking is how you see every line it changed, and how you undo it. - Permission prompts offer allow once and allow from now on. The second outlives the task, so grant narrowly and audit with /permissions. - On Pro, Max and Team, sessions currently start in auto mode, where a safety model reviews actions instead of you; Shift+Tab switches to manual mode, read-only until it asks. The flag that skips checks entirely is for throwaway containers. Security: - Your code is sent to Anthropic's models to be processed — how the tool works, not a side effect. On a client's private codebase, make that choice deliberately, and check your contract first. Security: Your account type is your data contract: - On Pro and Max, Anthropic currently trains future models on your data while that setting is on. Turn it off in account settings; retention follows the choice — currently around five years if you allow training, around thirty days if not. - Team, Enterprise and Console use fall under commercial terms, where Anthropic states it does not train on your code by default — the same tool on a different account type is a different deal. - It can change or delete files and run programs, so keep it to one project folder. And read what it proposes to run: a README or an issue it opens can carry instructions aimed at the agent, not you. Best at: - Explaining a codebase you have just inherited - Changes touching several files that must stay consistent - The maintenance you keep postponing: tests, tidying, dependency bumps Okay at: - Brand-new projects, where it builds more than you asked for - Predicting its own cost — usage climbs faster than the work suggests Avoid: - Codebases you are not allowed to send to a third-party service - Learning to code from zero — you need the judgement to spot wrong output Starter: Run it read-only on a codebase you do not understand It reads the actual files instead of reasoning from what you pasted, and manual mode lets you learn the permission prompts while it can change nothing. Prompt: Walk me through this codebase. What does it do, where does a request enter, and which three files should I read first? Starter: Give it one small, already-saved task It is judged on the loop, not the snippet — edit, run, see the failure, fix. A pass-or-fail task is the only way to see whether that loop closes on your project. Prompt: Add input validation to the login handler, then run the project's tests and fix anything that fails. Starter: Write a CLAUDE.md file for the project Creates a plain text file named CLAUDE.md in the project's top folder, holding the test command and the rules it should follow every session. Claude Code reads that file from the project root at the start of every session, so your test command and never-touch-this rules apply without retyping. Keep it short: a long one eats into every session, relevant or not. Pitfalls: - Trusting output without reading it. Fluent, fast, and confidently wrong often enough to matter. Unreviewed code carries your name on the commit. - Assuming a free Claude account includes it. It does not, and that catches people out at install time. Where to next: - More AI coding tools: https://www.launchpadhq.io/discover?category=ai-coding-development - Build with the API directly: https://www.launchpadhq.io/discover?category=ai-apis-developer-services ### Getting Started with Clay Canonical: https://www.launchpadhq.io/platform/clay/guide Data: https://www.launchpadhq.io/data/v1/guides/clay.json Last reviewed: 2026-10-24 Tool: Clay (https://www.launchpadhq.io/platform/clay): starting price $167/mo; free tier yes; data as of 2026-10-14 A spreadsheet that fills itself in: start with a list of companies or people, and Clay looks up details from many data providers and AI research — row by row, for credits. You build a table that looks like a spreadsheet. Rows come from a source — a company search, a CSV, your CRM. Then you add columns that enrich each row: look up a work email, a job title, company size, or ask Clay's AI agent to research something on the web. Each lookup spends Clay's usage units. What it is: - Clay is a data platform for sales and marketing teams. Its course for beginners teaches a four-step pattern it calls FETE: Find companies or people, Enrich them with more details, Transform the data with formulas or AI, and Export it to your CRM, email tool or a CSV. Clay's docs say every table begins with a source — a built-in company or people search, a CRM, a webhook or an uploaded file. - Its signature feature is the waterfall: Clay's docs describe it as several data providers run in a set order, so if the first one can't find an email, the next one tries. Clay's pricing page says it offers data from more than 150 providers through one account. - For anything a data provider can't answer, there's Claygent, which Clay's pricing page calls its AI agent with access to web research — for example, "does this company mention SOC 2 on its website?" — and AI columns that clean up or reshape data. Clay's AI docs say it uses models from OpenAI, Anthropic, Google and others, and that row data is only sent to a model when you run that action. Before you start: - Two meters, not one. Clay's credits guide splits usage into Actions (each enrichment, AI run or export) and Data Credits (paying for data from providers and AI models). A fully enriched record typically costs several Data Credits, per that guide. Exporting to CSV costs neither. - Free is for learning. As of 24 October 2026, Clay's pricing page lists a Free plan with small monthly allowances of Actions and Data Credits, a cap on rows per table, and no phone-number lookups. New accounts also get a 14-day trial with extra Data Credits, per Clay's plans guide; no credit card is needed to start. - Paid plans are expensive by beginner standards. The pricing page lists Launch at $167/mo (billed annually) and Growth at $446/mo (billed annually), more if you pay month to month, with higher tiers for more credits. Actions don't roll over; Data Credits partly do, with rules that differ for monthly and annual plans. - Using your own provider keys skips Data Credits for those lookups (the provider bills you instead), but each run still uses an Action. Getting set up safely (Sign up at app.clay.com/signup, linked from clay.com. Connect your CRM or email tools from inside Clay only once you know what you'll send there.): - Turn Auto-run off while you build. Clay's table settings guide says table-level Auto-run is on by default, so enrichments run whenever rows are added or edited. Clay recommends switching it off while building, testing on 5–10 rows, then turning it back on. - Test on 10 rows first. Clay's credit-saving guide suggests running a new column on about 10 rows before the whole table, and using "Only run if" conditions so you only enrich rows that qualify. - Confirm it works. Start from one of Clay's official templates, run it on a handful of rows, and check the credits it used in your workspace before scaling up. Security: - Clay is full of information about real people, and what you do with it is your legal responsibility. Clay's terms say you're responsible for having the right to upload and use your content. Its email best-practices guide — which says it isn't legal advice — notes you need a lawful basis for personal data in the EU and UK, every email needs an opt-out, and you shouldn't email people's personal addresses. Security: What Clay says about your data:: - Clay's AI docs say it doesn't use customer data to train models, apart from features that live in your own workspace. - Clay's pricing page lists SOC 2 Type II, GDPR, CCPA and ISO 27001 compliance, with reports available through its Trust Center. - Deleting a workspace removes its data after 30 days, per Clay's security doc. Privacy requests go to [email address on the web page]. Best at: - Building a detailed, accurate list of target companies and contacts by combining many data sources. - Answering custom questions about each row with AI web research instead of manual browsing. - Feeding clean data into a CRM or outreach tool automatically. Okay at: - Small budgets. The free plan is enough to learn; real prospecting at volume needs a paid plan. - Data accuracy. Clay's terms provide the service "as is"; spot-check results before relying on them. Avoid: - Running a new column on a huge table without testing it first. - Leaving Auto-run on while you're still building. Starter: Take the short course Work through the first lessons of Clay 101 on Clay University ("Get Started in Clay" and the FETE intro). Clay's interface packs a lot in. The official course explains the pattern every table follows, which makes every later step faster. Starter: Start from a template Open Clay's template "Find Key Decision Makers at any Company with just a Website URL" and run it on three company websites you know. Using companies you know lets you judge the results' accuracy immediately, and a template saves you building the columns yourself. Starter: Ask Claygent a question Add this as an AI research column on five rows. Asking it to say "unclear" when it can't tell keeps the column honest instead of filled with guesses. Prompt: Visit this company's website and tell me in one sentence what they sell and who their main customers are. If you can't tell, say "unclear". Starter: Export a CSV Export your finished table to CSV. Clay's credits guide says CSV export uses no Actions or Data Credits — a free way to take your results elsewhere while you learn. Pitfalls: - Unexpected deductions. Clay's credits guide lists the usual causes: Auto-run being on, AI columns charging for every row, and re-running columns. - Enriching every row of a big list. Filter or qualify rows first; per-workbook spend limits are an Enterprise feature, so on other plans the discipline is yours. - Assuming Free works for phone numbers. Phone enrichment needs a paid plan, and isn't included in the trial. - Cancelling instead of downgrading. Clay's credits guide says long-term retention isn't guaranteed after extended inactivity, and recommends downgrading rather than cancelling. Where to next: - Sales, Marketing & SEO AI: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai - Workflow & automation tools: https://www.launchpadhq.io/discover?category=workflow-automation - Research tools: https://www.launchpadhq.io/discover?category=research-academic-tools ### Getting Started with Cline Canonical: https://www.launchpadhq.io/platform/cline/guide Data: https://www.launchpadhq.io/data/v1/guides/cline.json Last reviewed: 2026-10-24 Tool: Cline (https://www.launchpadhq.io/platform/cline): starting price See pricing; free tier yes; data as of 2026-10-13 A free AI coding agent that works inside your code editor — it plans, edits files and runs commands, asking your permission at each step. You install Cline in your editor and connect it to an AI model. You describe a change in plain language; Cline reads your project, proposes a plan, then edits files and runs terminal commands — showing each change and waiting for your approval. You pay the model provider for the AI it uses, not Cline for the tool. What it is: - Cline is an AI coding agent — it doesn't just suggest code, it carries out a task across your project. Its docs describe an agent that reads and writes files, runs commands and uses a browser, with every action requiring your explicit approval. It runs as an extension in VS Code and VS Code-based editors such as Cursor and Windsurf, as a JetBrains plugin, and as a command-line tool. - It has two modes. In Plan mode it can read your code and talk the task through, but its docs say it cannot modify files or run commands. In Act mode it makes the changes. A good habit is to agree a plan first, then switch to Act. Cline also saves checkpoints — snapshots after each step, kept separately from your own Git history — so you can roll files back if a change goes wrong. - You can teach it your project's habits with rules files (a .clinerules folder, or an existing AGENTS.md), and extend it with MCP servers, a standard way to plug extra tools into AI apps. The VS Code extension and CLI are open source under the Apache 2.0 licence; Cline's README says its JetBrains plugins are not open source. Before you start: - The tool is free; the AI isn't. Cline's pricing page says it is completely free for individual developers, and you pay only for AI usage. You choose how: Cline's own provider (sign in, buy credits, no API key needed), ClinePass — an optional monthly subscription for extra usage on selected open coding models, per Cline's docs as of 24 October 2026 — or your own key from a provider such as Anthropic, OpenAI or OpenRouter, billed at that provider's rates. - Some free models exist, with conditions. Cline's docs describe rotating, time-limited free models for anyone with a Cline account, up to a quota, and note that free-model usage may be used to help improve those models. Don't use them for private code. - Running a model on your own computer is possible but demanding. Cline works with Ollama and LM Studio; its docs suggest 16–32 GB of memory for small models and 64 GB or more for larger ones, and warn local models are much slower. - You should be comfortable reading code. Cline asks you to approve each change and command. If you can't tell whether a change is safe, start with a chat assistant instead. Getting set up safely (Install from your editor's marketplace and check the publisher. In VS Code the genuine extension's ID is `saoudrizwan.claude-dev`, published as "Cline" with a verified cline.bot badge; JetBrains users install the "Cline" plugin from the JetBrains Marketplace.): - Keep auto-approve narrow. Cline can be set to approve some actions automatically. Its docs recommend auto-approving only "Read project files" at first, and call YOLO mode — which approves everything — dangerous, because Cline could delete files or push commits. - Know that .clineignore isn't a lock. Cline's docs say this file is being deprecated and is not a security boundary: ignored files can still be reached through mentions or shell commands. Keep secrets like .env files out of projects you open with an agent, or use the blocking hook its docs describe. - Decide on telemetry. Cline's usage reporting is on by default. Its docs say it doesn't include your code or prompts when you use your own key, but can include prompts when you use Cline's own provider. Turn it off under Settings → General → "Allow error and usage reporting". - Confirm it works. Open an empty folder, switch to Act mode and ask for something tiny, like a one-page HTML to-do list. Approving the first file it writes is your setup checkpoint. Security: - Cline acts on your real files and terminal, so the safety model is your approvals. Read each diff and each command before you accept it. Cline's docs say the model itself decides which commands count as "safe" — there is no fixed allowlist — which is why approving commands automatically is a bigger step than approving file reads. Security: Where your code goes:: - With your own API key, Cline's privacy notice says your content goes straight to that provider and Cline doesn't collect it. - With Cline's own provider or ClinePass, Cline receives your prompts and passes them to the model provider, some of which may train on it unless you opt out. - Only install MCP servers you trust; Cline's docs say so plainly, because an MCP server is code with access to what Cline can reach. Best at: - Multi-file changes you can describe in a sentence, with a plan you approve before anything is edited. - Choosing your own AI model and provider, including running locally. - Undoing mistakes, thanks to checkpoints separate from your Git history. Okay at: - Keeping costs predictable. You pay per token, and Cline's docs note its running cost estimate can differ from your actual bill. - Very long tasks. When the context fills up, Cline compresses it, and its docs warn earlier details may be lost. Avoid: - YOLO mode or broad auto-approval on a project you care about. - Opening folders that contain secrets or credentials. Starter: Build something tiny from scratch Cline's own walkthrough starts with a to-do app in an empty folder. Approving each file shows you the whole loop — plan, edit, approve — with nothing at risk. Then ask it to add one feature, and watch how it edits rather than rewrites. Prompt: Create a simple to-do list app in a single HTML file. It should let me add items, tick them off, and delete them. No frameworks. Starter: Plan before you act Run this in Plan mode, which can't modify files. Correcting a plan costs a few messages; correcting a botched edit costs more. Prompt: Read this project and explain what each top-level folder does. Then suggest a plan to add input validation to the signup form. Don't change any files yet. Starter: Write down your house rules Use /newrule with this. Rules save you repeating yourself, though Cline's docs note they use up context on every request — keep them short. Prompt: Create a project rule that says: use TypeScript, keep functions small, and never edit files in the generated/ folder. Starter: Practise rolling back After Cline makes a change, open the checkpoint for that step and choose Restore Files. Knowing you can undo makes it easier to let the agent work — try it once on something unimportant. Pitfalls: - One giant task. Long conversations cost more and lose detail; start a fresh task with /newtask for a new job. - Approving commands you don't understand. If a command looks unfamiliar, ask Cline what it does before approving. - Different defaults after an update. Cline's changelog warns default models can change between versions; pick your model explicitly. - Assuming the cheapest model is enough. Cline lets you use a stronger model for Plan mode and a cheaper one for Act. Where to next: - More AI coding tools: https://www.launchpadhq.io/discover?category=ai-coding-development - Run models locally: https://www.launchpadhq.io/discover?category=local-open-source-ai - AI APIs for your own key: https://www.launchpadhq.io/discover?category=ai-apis-developer-services ### Getting Started with Clio Manage AI Canonical: https://www.launchpadhq.io/platform/clio-manage/guide Data: https://www.launchpadhq.io/data/v1/guides/clio-manage.json Last reviewed: 2026-09-13 Tool: Clio Manage (https://www.launchpadhq.io/platform/clio-manage): starting price $49/user/mo; free tier yes; data as of not dated The assistant built into your firm's case files — it drafts, bills and schedules from work already logged there. You're inside your firm's case-management system, looking at one client's case file. A gem icon at the top of the screen opens a panel where you type a question in plain English. It answers using the files, notes and time entries already in that file, and you edit or discard what comes back before anything is sent. What it is: - Clio Manage is the software a law firm runs its day on — every client's case file, or matter (the law firm's word for one client's case and everything filed under it), lives in it, along with the time entries, documents and calendar dates attached to that matter. - Manage AI is the assistant Clio built on top of that pile. Because it can only see what your firm has already put into Clio, it is less a chatbot you consult and more a colleague who has read the file: you ask it where a matter stands, or hand it a court notice, and it works from your own records rather than from the open internet. - Clio launched it as Clio Duo and renamed it Manage AI in late 2025, so both names still appear across Clio's help pages and in older write-ups. - Clio sells Manage AI in every region it operates in, and it publishes no list of supported languages. Before you start: - This is not a tool you try for free and keep. Clio Manage is a paid, per-user subscription with no free plan. Clio publishes only the entry price for each region, including the US and the UK, and quotes the rest on request, so check Clio's own pricing page for your region. You can see it without paying: Clio offers a 7-day trial at clio.com/signup with, in its words, no credit card required, and Clio's support pages say the trial cancels itself if you never add one. Realistically, though, you will pay — this is a firm's system of record, not a weekend experiment. - Check which plan you are on before you assume you have the AI, and check it for your own country. Clio's plan tables differ by region. On the US, UK and Australian pricing pages the AI is bundled and Clio's own FAQ says it comes "on Core and above … as part of the plan rather than as a separate add-on" — the cheapest plan (Starter in the US, EasyStart in Australia) is marked No on every AI row. On the Canadian pricing page there is no AI section at all and Manage AI is listed as a paid add-on quoted by a sales rep, with no public price for it anywhere. Clio's help centre adds a third set of plan names again (Standard, Premium, Pro and Suite). Clio's feature pages, meanwhile, describe Manage AI as simply built into Clio Manage. Those sources genuinely disagree, so treat the comparison table for your own country as the answer and confirm against your account rather than against anything marketing says. - Someone with administrator rights has to turn it on. Clio's help centre is explicit that an administrator must activate the feature and accept its terms, so if you are not the person who administers your firm's Clio, your first step is a conversation, not a click. Firm-performance reporting is gated higher again — Clio's current table shows it only from Signature upward. - It is worth exactly as much as your firm's data hygiene. If your firm keeps half its time in a notebook, the assistant has half a file to read. - Where this page describes a screen, it is describing what Clio's own help documentation says the screen does — no account was opened to write it. The exact wording you meet once you are signed in may have moved. Security: - Clio's AI reads a store your firm has already filled: privileged client material, indexed and searchable, years of it. Clio's terms of service (effective 4 August 2026, under its legal name Themis Solutions) state that using the AI services "does not grant Themis the right to use any of Subscriber's Confidential Information for the purpose of training generalized LLMs" — no general-purpose model (the AI "brain" that does the actual thinking) is being trained on your clients. The same section reserves a narrower right you should read: Clio "may de-identify and aggregate the Content submitted to, and Output received from, AI Services" to improve those services. That is a meaningfully different promise from "your data is never touched," and it is the sentence to put in front of whoever signs off on this at your firm. Security: Three things worth confirming before a client's file goes through it:: - Where the query is processed. Clio's help centre states that "depending on where your firm is located, AI features in Clio may process your queries on servers located outside your home jurisdiction," and that afterwards "any resulting data remains securely stored within your region." Storage location and processing location are two answers, and only one of them is your region. - Who the third parties are. Clio's published subprocessor list (last updated 20 August 2026) names OpenAI, Anthropic, Microsoft (Azure Foundry), Google (Vertex AI) and Amazon (AWS Bedrock) as AI processors, with processing regions listed per vendor — Anthropic showing US only. If your engagement letters or your regulator require you to name who touches client material, that list is the document you need. - Who inside the firm sees the output. Clio states that the AI features follow your existing Clio Manage permissions, and drafted client communications are visible and sendable only to the matter's responsible staff member or responsible attorney. That is a genuine control, but it inherits whatever your matter permissions already are — which for many small firms is everyone. Best at: - Reconstructing where a matter stands from the file itself — the documents, notes and activity already logged — with citations you can click back to the source document. - Turning work you have already recorded into finished admin: draft bills with cleaned-up narratives, calendar entries pulled out of a court notice, tasks and notes off the back of a stage change. - Getting a first draft of a routine client update written from matter activity, rather than from you remembering what happened. Okay at: - Reading at volume. A single document analysis currently tops out at 25 documents, 50MB in total and 50MB per file, in DOCX, TXT or PDF; the date-extraction flow takes one document at a time, currently up to about 20 pages or 10MB. - Firm-performance questions. The free-text reporting works, but it sits a tier above the rest of the AI. - Sustained back-and-forth. The document analyser is a one-shot: Clio's own documentation says it "cannot preserve chat context across sessions or respond to follow-up questions based on past interactions." Avoid: - Calculating a filing deadline and trusting the number. Clio's help centre says it plainly and its marketing does not: "AI Scheduling does not currently apply local court rules for counting days when calculating dates. Always check calculated deadlines against official court rules." Extracted dates marked "Calculated" were derived from other dates, not read off the page. - Legal research. It reads your firm's records. It is not a source for what the law says, and asking it to be one confuses a file clerk with a library. - Sending anything it wrote without reading it. Clio's terms say outright that it "does not review Output for accuracy or completeness" and puts the review obligation on you — including on the billing narratives it rewrote, which are highlighted precisely because they changed. Starter: Find out where a matter stands before the client calls Because the assistant is reading the matter record itself rather than a document you pasted in, the summary comes back with in-text citations that open the source document in the file. MyCase will summarise text too; what it does not have is Clio's matter-scoped document store returning a clickable citation trail inside the same matter. Prompt: Summarise everything that has happened on this matter in the last 60 days — documents filed, tasks completed, money billed and anything still outstanding. Starter: Ask one long document a specific question The analysis flow will not let you skip the question — after you pick the matter and the documents it moves you to a step literally labelled "Next: Ask your question," and you choose Quick or Thorough before it runs. The answer can then be saved back into the matter as a note or a .docx rather than living in a chat log. PracticePanther has no equivalent forced question step scoped to a matter's own attachments, with the result written back into that matter. Prompt: In this agreement, what are the termination provisions, what notice period applies to each, and which party bears the cost on early termination? Starter: Let it build the week's draft bills, then check the blue On a weekly or monthly cadence you set, a scheduled job converts the time and expense entries already logged against hourly, flat-fee and contingency matters into draft invoices, rewrites the entry descriptions for clarity, highlights every line it touched in blue, and routes the draft to your named approvers rather than to the client. The mechanism is polish-and-route, not capture: Clio is not watching you work to invent the time entry, it is taking what is already recorded. Smokeball's pitch is automatic time capture, which is a different stage of the billing pipeline altogether. Starter: Turn a court notice into calendar entries you approve one by one You pick the matter, confirm the document type and upload a single document — a court notice, a scheduling order, even handwritten notes. It pulls out the actionable dates, labels any it worked out from another date as "Calculated," and presents the whole set for you to edit, delete or add to before you press Create events. The propose-then-confirm step is the mechanism, and it exists because of a limit Clio publishes rather than hides: a screen that makes you approve each date individually is doing real work. Nothing in MyCase pairs a one-document extraction with a "Calculated" label on derived dates and a confirmation gate. Starter: Deal with the client update it noticed you owe Without being asked, it watches matter activity — completed tasks, uploaded documents, a change of stage — and drops a suggested client update onto the matter dashboard as an AI Actions card. You open it, adjust the tone or the content, and send. This one is on a clock in both directions: the card appears off the back of a specific logged event, and the draft currently expires about a day later, so an unopened suggestion disappears rather than queuing up — an event-triggered draft with an expiry that PracticePanther does not offer. Pitfalls: - The analyser can only see documents attached to a matter. A PDF sitting in your inbox or on your desktop is invisible to it until it is uploaded into the file — which for a beginner reads as the AI failing when it is really the filing that hasn't happened yet. - Automated billing bills what is logged when it runs, not what you meant to log. Clio's own guidance is that everyone finalises their time and expense entries before the end of the previous business day; time entered the morning after simply isn't in that bill. - Two products, one gem. The document-analysis flow is also reachable as "Analyze with Vincent," which is Clio's separate research AI — so the same button can put you in front of a differently-scoped tool than the assistant you were using a moment ago. Note which one answered you. Where to next: - Legal AI tools: https://www.launchpadhq.io/discover?category=legal-ai - Workflow & Automation: https://www.launchpadhq.io/discover?category=workflow-automation - Meetings & Notes: https://www.launchpadhq.io/discover?category=meetings-notes ### Getting Started with CoCounsel Legal (formerly Casetext CoCounsel) Canonical: https://www.launchpadhq.io/platform/casetext-cocounsel/guide Data: https://www.launchpadhq.io/data/v1/guides/casetext-cocounsel.json Last reviewed: 2026-09-13 Tool: CoCounsel (https://www.launchpadhq.io/platform/casetext-cocounsel): starting price See pricing; free tier no; data as of 2026-10-20 The legal AI that Casetext became — grounded in Westlaw, and sold only through a sales call. You open a browser window, type what you need in plain English, and attach whichever documents it should work from. It thinks in steps you can watch, then hands back a draft or an answer with every citation linked, so you can click each one and check it before you rely on any of it. What it is: - CoCounsel Legal is Thomson Reuters' AI assistant for legal work — the product that used to be sold as Casetext CoCounsel, before Thomson Reuters bought Casetext in 2023 and folded it into the Westlaw family. - You describe a matter in plain language and it researches, drafts and reviews documents, grounding what it says in Westlaw case law and Practical Law's attorney-written guidance rather than in whatever the AI absorbed during training. - The August 2026 rebuild runs on Anthropic's Claude Agent SDK and behaves as an agent (AI that takes actions on its own rather than only answering) — planning several steps and showing its work as it goes, instead of picking a task off a menu the way the original Casetext version did. - It is a professional subscription tool: it assumes you are already doing legal work and have someone to bill it to. Before you start: - Casetext is not a product you can buy any more. - You will pay from day one, and you cannot find out how much on your own. Thomson Reuters publishes no price for any CoCounsel plan — its pricing page asks for your sector, number of attorneys, jurisdiction and preferred plan length before it will quote, and both the "free demo" and "free trial" buttons lead to a form that ends with a sales representative phoning you. There is no free tier advertised and no self-serve signup, so budget for a conversation before you budget for the software. Thomson Reuters advertises a free trial of CoCounsel Essentials but does not publish how long it runs, so ask what you are being given. - The rebuilt CoCounsel Legal launched on 20 August 2026 and is currently United States only — Thomson Reuters says Canada, the United Kingdom and Australia are expected later in 2026. If you practise outside the US, the honest answer today is "not yet." - There are four plans and they are not small variations on each other. Thomson Reuters' own comparison table shows Westlaw content and KeyCite flags — the red and yellow markers that say a case has been overruled or criticised — in only two of the four, and Practical Law in a different two. Get the plan named in writing before you agree to anything. - If what you actually want is to feel out what AI is like on legal-adjacent work — summarising a lease, putting a clause into plain English — a general assistant such as ChatGPT will let you do that today for nothing and without a phone call. Start there, and come back to CoCounsel Legal when a firm is paying for it. Security: - The vendor-side answer is unusually solid for this category, and worth taking as read: Thomson Reuters states that "CoCounsel Legal is not trained on any customer data," and lists independent audits — ISO/IEC 42001:2023 and SOC 2 Type 2 — plus a US federal authorisation in progress. That is not the part that will bite you. Security: The part that stays yours: - Attorney-client privilege is a duty you owe, not a feature the software carries. ABA Formal Opinion 512, issued 29 July 2024 and still the American Bar Association's standing guidance on generative AI, is explicit that a lawyer's confidentiality obligation can require informed client consent before client information goes into a tool like this. A vendor's security certificate does not discharge that, and no setting inside CoCounsel Legal does either. - Your prompt (the message you type) and everything you attach to it pass through outside AI providers under Thomson Reuters' contracts, not through Thomson Reuters alone — the company's own security notes for the Outlook integration say user content is not used to train third-party models and name OpenAI and Google as the providers in that chain. The protection is contractual, which is a different kind of protection from the data never leaving. - Accounts are administered by the firm, not by you. "Who else can see this matter" is answered by whoever runs your organisation's subscription, so ask them before the first client document goes in. Best at: - Turning a plain-language question into research where every case comes back linked and flagged for whether it is still good law. - First drafts that have to carry citations — briefs, memos, motions — with a separate pass that checks whether the authority cited actually supports each assertion. - Reading a large set of your own documents in one go and returning the answers as a sortable table rather than a wall of prose. - Marking up and rewriting inside Microsoft Word, where the drafting actually happens, rather than in a separate browser tab you copy out of. Okay at: - Novel or genuinely unsettled questions. It is strongest where authority already exists to cite, which is the opposite of where you most want a second brain. - Material outside Thomson Reuters' own libraries. It will read whatever you attach, but the grounding that makes it trustworthy is Westlaw and Practical Law; point it at a foreign statute or a trade standard and it is an ordinary assistant again. - Short everyday tasks. It will summarise an email thread perfectly well, and you will be paying research-platform rates for the privilege. Avoid: - Filing anything on the strength of a cheaper plan's output. CoCounsel Essentials and the Practical Law bundle carry no KeyCite flags at all, so nothing in those plans will tell you the case you are about to cite has been overruled. - Reading the visible reasoning trail as a check on the answer. Thomson Reuters ships a separate feature, Deep Research Verify, whose entire job is confirming that cited authority supports the assertion made — which is a fair indication the reasoning display was never doing that job. Starter: Ask the question, not for the cases Thomson Reuters owns Westlaw, so the cases this returns arrive already carrying KeyCite flags — Westlaw's own markers for whether a decision is still good law. Harvey's authority comes from LexisNexis under the alliance the two announced in June 2025, and there is no KeyCite flag anywhere in that stack. Prompt: Is a non-compete signed in one state enforceable against an employee who has since moved to Texas? Give me the controlling authority, tell me where courts disagree, and flag any case you cite that has been criticised or overruled. Starter: Draft from the house library, not from scratch The clauses come out of Practical Law, Thomson Reuters' own library of attorney-written model documents and practice notes. Neither Harvey nor Lexis+ AI can reach Practical Law at all; it is Thomson Reuters property, which is exactly why it is a plan you buy rather than a feature you enable. Prompt: Draft a mutual confidentiality agreement for a US software vendor and a hospital, using standard market clauses. Mark anywhere you have departed from the standard position and say why. Starter: Put a document set into a table The mechanism is that the questions travel across the set rather than the document — Thomson Reuters specifies up to 10,000 documents and up to 100 questions in a single pass, returned as a filterable table. Westlaw's own research tools cannot do this at all, because they search published authority rather than the files sitting on your matter. Prompt: Across the attached leases, answer for each one: who is the landlord, when does the term end, is there an assignment clause, and does it require landlord consent? Return a table with one row per lease and one column per question. Starter: Reach it from Claude instead of logging in Thomson Reuters and Anthropic announced a connection on 12 May 2026 that lets Claude talk to CoCounsel Legal directly, so a request you make in Claude can be routed into CoCounsel to research, draft and come back with validated citations. It is set up once by your firm as a connector rather than something you type into CoCounsel itself. This is a first-party bridge built by the content owner, which is what makes the citations that come back Westlaw-grounded rather than written by the AI. Harvey's content arrangement runs the other way — its LexisNexis alliance brings that material into Harvey's own interface, so there is nothing on that side for an outside tool to call. Thomson Reuters said general availability was expected during summer 2026 and has not published whether the connection needs anything on top of a CoCounsel subscription, so treat it as something to ask about rather than something you switch on. Pitfalls: - Searching for "Casetext" and buying from whatever comes back. Anything still offering you "Casetext" is a reseller, a review site, or simply out of date. - Expecting the Word and Outlook add-ins to be the same generation as the browser product. Thomson Reuters' own security notes for the Outlook integration still describe an earlier CoCounsel engine, so the surfaces may not all move at once. - Using a Workspace as personal scratch space. Thomson Reuters describes Workspaces as matter environments holding firm documents and institutional knowledge, which means the matter team, not just you — check the sharing rules with whoever administers the account. - Reading the descriptions here as a walkthrough. No account was opened to write this page, because there is no way in without a contract, so what the product does is Thomson Reuters' published account of its own software — verify the steps against what you see once your firm's login arrives. Where to next: - Legal AI tools: https://www.launchpadhq.io/discover?category=legal-ai - Research tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Cohere Canonical: https://www.launchpadhq.io/platform/cohere/guide Data: https://www.launchpadhq.io/data/v1/guides/cohere.json Last reviewed: 2026-09-13 Tool: Cohere (https://www.launchpadhq.io/platform/cohere): starting price Free, from See pricing; free tier yes; data as of not dated An enterprise AI company that sells its technology to other businesses — and how a beginner can still get in for free. Cohere's documentation describes the motion like this: you sign in, open the Playground, pick a model (the AI "brain" that does the actual thinking) from a panel on the right, type into a message box at the bottom, and read the reply. You adjust your wording or switch models and send again, then click through to the code. What it is: - Cohere is a Canadian AI company that sells its models to other businesses rather than running a chatbot of its own — the things it makes end up inside other companies' products, not on your phone. - It is still worth a beginner's twenty minutes, because Cohere publishes its current flagship, Command A+, as a file anyone can download from Hugging Face under the Apache 2.0 licence — the licence that lets a business, not just a hobbyist, run it — which almost no other frontier lab does with its best model. - Alongside writing text, Cohere sells narrower models that chat companies mostly don't: one that reorders search results by relevance, one that pulls text out of scanned documents, plus transcription and translation. - You'll find Cohere underneath somebody else's search box. Before you start: - Signing up is free and self-serve: go to dashboard.cohere.com, create an account with Google, GitHub, or an email and password, and Cohere's pricing FAQ says a trial API key (a password that identifies your app, and that spends your money) is created for you automatically and shown in the dashboard's "API Keys" section. You need to have reached the age of majority where you live to accept the terms. - You will almost certainly not need to pay while you are learning. Cohere's pricing table currently lists Command A+ as free for both "API key" and "Model download," and Cohere's model documentation says that "for both trial keys and production keys, Command A+ is free until rate limits are reached." The models that do cost money are the smaller and narrower ones — such as Command R, billed per million tokens (a chunk of text, roughly a short word — how AI usage gets counted) of input — and you have to choose them on purpose. - Free here has a contract attached, not just a ceiling. Cohere's pricing page states that trial keys "are rate limited and are not permitted to be used for production or commercial purposes," and its rate-limit docs currently put chat on a trial key under both a per-minute and a monthly call cap. So the free key covers learning and prototypes; the moment something you build takes money, it needs a production key, which bills pay-as-you-go and is requested from the Billing and Usage page by an account owner. - Cohere is built for people putting AI inside software, and most of it is reached from code. The part you can reach in a browser is the Playground, which Cohere's docs describe as a message box with a model picker — you may need to check what it actually offers once you are signed in. If you have never used an AI tool at all, an hour with a chatbot from Text & Conversational AI first will make everything here land better, and Cohere will still be waiting. Security: - Cohere draws its privacy line by account type, not by a setting you can flip. Its privacy policy addresses trial users directly: "we collect content you submit to the Cohere Products, and outputs generated by the Cohere Products. We may use this input and output data to conduct research and development." The free key you are handed at signup puts you on exactly that side of the line. Security: What Cohere itself asks of you: - Keep personal information out of it. The same policy says it in as many words: "Please do not submit personal information in inputs to the Cohere Products when accessing them on a research platform or as a Trial User." - Don't go looking for an opt-out switch. Controlling how Cohere trains on your data is described as something Enterprise Users negotiate, alongside a data processing agreement — a contract term, not a toggle in settings. - If you put a Cohere model in front of other people, Cohere's usage policy (the version published as of 13 September 2026 is dated 21 November 2024) requires your app to "disclose to the users that they are interacting with an AI system rather than a human." Security: - None of this is unusual for a developer platform. What is unusual is how plainly Cohere addresses the sentence to you. Best at: - Working in languages other than English — Command A+ is documented at 48 languages, including every official EU language - Answering from documents you hand it, with the reply marked up to show which document each claim came from - Being a model you could later run yourself, because the flagship's file is published rather than locked in the service Okay at: - Being your everyday assistant — the browser surface Cohere documents is a developer test bench - Anything involving pictures — Command A+ takes images as input, but Cohere sells no image generator - Telling you what things will cost: individual models carry public prices, but North and Compass both say only "Get in touch for custom enterprise pricing" Avoid: - Reaching for it as your daily chatbot. Cohere's own homepage lists North, Compass, Command, Transcribe, Embed, Rerank, Parse, North Small Translate, Aya and Model Vault — and not one of them is a consumer app you'd open in the morning - Following any Cohere tutorial written before this year. Cohere retired the command, command-light, command-r and command-r-plus names on 15 September 2025 and five more models on 4 April 2026, so older sample code names models the service no longer serves - Treating a citation as a fact-check. It tells you which document a sentence came from, not whether that document was right Starter: Put Command A+ in the model slot before you try anything else Open the Playground from the dashboard and choose Command A+ from the model list on the right, rather than accepting whatever is selected by default. Most vendor playgrounds sit you on a metered flagship whose weights you will never see, so what you learn about its behaviour is only ever rentable. With Command A+ the thing in the picker and the thing you could run yourself are one model, which is true of neither half of OpenAI's Playground. Command R and Command R7B are the smaller, metered ones; there is no reason to pick them while you are learning. Starter: Ask it in a language that isn't English Cohere sells two more models whose entire job is translation — command-a-translate across 23 languages and North Small Translate across more than 50. Anthropic, by contrast, ships no dedicated translation model and publishes no supported-language count to check against. Swap Polish for whichever language you actually need. The back-translation is the part that tells you whether to trust it. Prompt: Answer this in Polish, then give me an English back-translation underneath so I can check it: [your question] Starter: Hand it a document and make it show its working When you supply documents through Cohere's chat interface, its documentation says the reply comes back with a citations object naming which document supports a claim and marking "the start and end fields, which are spans that tell us the location of the supported claim inside the reply." Grounding is a return value here, not a habit you have to coax out of the model; OpenAI's chat completions return no per-span citation object for documents you supply. The docs describe that object in the API response, though — whether the Playground draws those spans on screen is not something they demonstrate, which is why this asks for the quotes in plain text instead. A rental agreement, a policy document, a long email thread — anything where "where did you get that?" is the question you'd ask a person. Prompt: Here is a document: [paste it]. Answer this question using only that document, and after each sentence of your answer, quote the exact line you took it from: [your question] Starter: Read the Rerank row on the pricing page and work out what it is for Look at the retrieval models on cohere.com/pricing — Rerank 4 Fast and Rerank 4 Pro are both priced per million tokens, with Pro costing a little more. Reranking is sold here as a model in its own right, with its own price, doing nothing else. Neither Anthropic nor OpenAI sells a standalone reranking model, so a team with a bad search box has to improvise one out of a chat model. Starter: Check which kind of key you are holding before you build on it Open the dashboard's "API Keys" section and confirm the key sitting there is the trial key created at signup, not a production key. Cohere is one of the few platforms where the free key's ceiling is contractual as well as numeric, and where a production key is a separate object rather than a card on file. There is no equivalent wrong-key state at OpenAI, where a key is one kind of object with a balance behind it. Pitfalls: - Landing on cohere.com, seeing "Request a demo" and "Sign in," and concluding there is no way in. The free self-serve signup lives at dashboard.cohere.com and is not advertised on the front page at all. - Reading "free" next to Command A+ as "unlimited." It is a price, not a quota — the rate limits sit on top of it and are what you will actually run into. - Confusing the two different frees. The Apache 2.0 licence on the downloaded Command A+ file permits commercial use; the free trial key explicitly does not. Same model, two different permissions, and it is the key that carries the restriction. - Expecting the Playground to cover the product. Cohere's docs describe it as supporting Chat and Embed — so Rerank, Parse and Transcribe, the models that most distinguish Cohere from a chat company, are the ones you cannot click your way into. - Trying to add a colleague and finding you can't. Cohere's FAQ states that "every account begins as a personal account," and that you "will not be able to add other members until you become part of an organization." Where to next: - AI APIs & Developer Services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai - Local & Open-Source AI: https://www.launchpadhq.io/discover?category=local-open-source-ai ### Getting Started with ComfyUI Canonical: https://www.launchpadhq.io/platform/comfyui/guide Data: https://www.launchpadhq.io/data/v1/guides/comfyui.json Last reviewed: 2026-10-23 Tool: ComfyUI (https://www.launchpadhq.io/platform/comfyui): starting price $20/month; free tier yes; data as of 2026-10-14 A free, open-source workbench for AI image, video and audio generation, where you wire the steps together like a flowchart — and run it on your own computer. You build a workflow: boxes called nodes, each doing one step — load a model, read your prompt, generate, save the image — joined by wires that pass the result along. You press Run, and your own graphics card does the work. Ready-made templates mean you rarely start from an empty canvas. What it is: - ComfyUI is a program for running generative AI models — the kind that make images, video and audio — with every step visible. Its docs describe it as a node-based interface and inference engine that is completely open source and can run on your own device. It is free, under the GNU GPL v3 licence. - Where most image apps give you one prompt box, ComfyUI shows the whole recipe. A basic text-to-image workflow is a short chain: a node that loads a model, nodes that turn your positive and negative prompts into something the model understands, a sampler node that does the generating, and a node that saves the result. Change any box — the model, the size, the number of steps — and run again. That openness is why so many new open models get ComfyUI support first. - Models are not included. You download them separately, often several gigabytes each, into ComfyUI's models folder. Built-in templates (sidebar → Templates) cover the models ComfyUI supports natively and tell you if a model they need is missing. - There are two optional paid extras. Partner Nodes let a workflow call closed models from other companies, such as OpenAI or Kling, using prepaid Comfy credits. Comfy Cloud runs ComfyUI on Comfy's own graphics cards in your browser if your computer isn't up to it. You don't need either to use ComfyUI locally. Before you start: - Check your hardware first. ComfyUI's docs list support for NVIDIA, AMD and Intel Arc graphics cards and Apple Silicon Macs; it can also run on a plain CPU, which the docs call slower. They don't publish a minimum amount of graphics memory (VRAM), but their beginner example says a small classic image model is about 4 GB and runs on a consumer card with around 6 GB of VRAM. Bigger, newer models need much more. - Pick an install. The Desktop app (Windows, Apple Silicon Macs and Linux) is the beginner route and needs at least 4.85 GB of disk before models. Windows also has a portable build you unzip with 7-Zip, and advanced users can install manually with Python. Plan on tens of gigabytes of space once you collect a few models. - Local use costs nothing. ComfyUI's credits page says ComfyUI is free for local use; credits only apply to Partner Nodes and Cloud. Partner Nodes show a price badge before you run them, and Comfy's docs say purchased credits are non-refundable. Comfy's pricing page says new Comfy Cloud users get a few free runs on real GPUs (plus free Comfy Agent credits), and paid Cloud plans start with Standard at $20/month, less if billed yearly. Getting set up safely (Download ComfyUI only from comfy.org/download or the official Comfy-Org GitHub repository. On a Mac, macOS may ask you to click "Open Anyway" in Privacy & Security on first launch — that's expected for the genuine app, and a reason not to install a copy from anywhere else.): - Treat custom nodes as software you're installing. ComfyUI's security policy says custom nodes are arbitrary Python code — they can do anything a program on your computer can. Its docs recommend installing only well-known nodes from trusted authors, through the built-in ComfyUI Manager, and avoiding obscure ones. - Keep it on your own machine. ComfyUI only listens on your own computer by default. The --listen option opens it to your network, and the security policy says you're then responsible for securing it. Don't turn that on unless you know why you need it. - Confirm it works. The default text-to-image workflow loads at launch; the Desktop app offers to download its starter model. Press Run (or Ctrl+Enter). An image appearing in the Save Image box means you're set up. Security: - Running locally is the privacy upside: prompts, images and models stay on your computer, and nothing is sent anywhere unless you use Partner Nodes or Comfy Cloud. Comfy's docs note you can turn Partner Nodes off entirely with a launch option, or run fully offline. Security: Where the risk actually is:: - Custom nodes. Most ComfyUI setups grow a collection of community add-ons. Each one is code with full access to your computer; stick to widely used ones and keep ComfyUI Manager's security setting at its default. - Shared workflows. A workflow file someone posts may need custom nodes you don't have, and ComfyUI Manager will offer to install them all. Check what they are before clicking Install All. - Models from the internet. Comfy's docs name Hugging Face, Civitai and GitHub as common sources. Download from the official model page where you can, and prefer the .safetensors files its examples use. - Your Comfy account. If you buy credits, Comfy's docs warn that signing in from an insecure, network-exposed setup could let someone steal your login and spend them. Best at: - Running open image, video and audio models on your own hardware, privately and with no per-image cost. - Seeing and controlling every step of a generation, then saving that exact recipe to reuse or share. - Trying new open models quickly — ComfyUI tends to support them early through templates. Okay at: - Being beginner-friendly. Templates help a lot, but error messages assume you know what a checkpoint or a latent is. - Running on modest hardware. Small models work on everyday graphics cards; the newest large models may not fit. Avoid: - Installing every custom node a downloaded workflow asks for without checking what it is. - Updating everything right before a project — Comfy's own docs warn updates can break custom nodes. Starter: Run the default workflow Open ComfyUI, let it fetch the starter model if asked, type a short description in the positive prompt box, and press Run. Right-click the result to save it. It proves your install and graphics card work, and every later workflow is a variation on this five-box chain. Change only the seed or the step count and run again, to see what each setting actually does. Starter: Open a template Click Templates in the sidebar and open one for a different task, such as image editing or upscaling. If it flags missing models, let it download them. Templates are maintained by Comfy and match the models ComfyUI supports, so they work without hunting for parts — the safest way to try something new. Starter: Recover a workflow from an image Drag a PNG that ComfyUI saved back onto the canvas. The whole workflow that made it, settings included, reappears. ComfyUI embeds the recipe in every image it saves. Knowing this means you never lose a setup you liked — and you can learn from images others share. Starter: Save your own version After changing a workflow, press Ctrl+S to save it. It appears in the Workflows panel (press W) for next time; use Workflow → Export to keep a JSON copy. Small, saved variations are how you build up a personal toolkit instead of rebuilding the same graph every session. Pitfalls: - "Out of memory" errors. Comfy's troubleshooting guide suggests lowering the image size or batch size and closing other programs that use the graphics card first. - Mixing models that don't belong together. Errors about tensor or channel sizes usually mean an add-on model was made for a different base model. - Red boxes after loading a workflow. They mark missing models or nodes; the Errors panel and ComfyUI Manager tell you what's needed. - Custom nodes behind most problems. Comfy's troubleshooting page says most reported issues involve custom nodes; starting with them disabled is the quickest test. Where to next: - More local & open-source AI: https://www.launchpadhq.io/discover?category=local-open-source-ai - Image generation & editing: https://www.launchpadhq.io/discover?category=image-generation-editing - Video creation & editing: https://www.launchpadhq.io/discover?category=video-creation-editing ### Getting Started with Connected Papers Canonical: https://www.launchpadhq.io/platform/connected-papers/guide Data: https://www.launchpadhq.io/data/v1/guides/connected-papers.json Last reviewed: 2026-09-15 Tool: Connected Papers (https://www.launchpadhq.io/platform/connected-papers): starting price Free, from $3/mo; free tier yes; data as of not dated Turn one paper you already have into a visual map of the field around it. You paste a paper's title or web link into the search box on the homepage and press enter. After a few seconds a page of circles appears, each one a related paper. You drag and zoom around it, click any circle to read that paper's summary, and start a new map from anything you find. What it is: - Connected Papers is a visual map-maker for academic literature. You give it one paper you already care about, and it draws you the cluster of papers around it. - It is not a chatbot and it does not summarise anything for you — there is no message box. The only thing you type is a paper's title, its web address, or its DOI (the permanent ID code printed on most published papers). - What it is genuinely good at is the problem of not knowing what you don't know: you have one paper, and no idea which forty others a reviewer would expect you to have read. Before you start: - You need one paper to start from. Connected Papers has no browse mode and no topic search — type a subject like machine learning and it will look for a paper with that phrase in its title, not a field to explore. Find one paper first, anywhere, then come here. - Log in before you start clicking. Without an account you get only a small monthly allowance of graphs; a free account raises it. The counter runs on the calendar month and resets on the 1st, and it counts distinct starting papers — reopening a map you already built shouldn't cost you another one, and the three example graphs on the homepage are free. - Paying is an unusually small decision here. The Academic plan — which covers students, university researchers, non-profits and personal use — is billed yearly or quarterly; there is no month-to-month option. Prices are set per country, so check the figure for yours on the pricing page. Most people never need it: the free account's monthly allowance is a lot of maps. - If you are on a university network, look at the top of the homepage before you pay anything. Connected Papers grants premium access by institutional IP address, and says it shows a line naming the institution providing it when that applies. There is also a scholarship program at [email address on the web page] for researchers who genuinely cannot afford a subscription. Security: - This is one of the lowest-risk tools on this site. You are not uploading documents or pasting private text — you type the name of a paper that is already published, which is public information. Nothing you write is read by a model. Security: What it does keep: - Your username, email address and password, if you make an account - The papers you add to your Saved Papers list - Your graph history — which papers you have started maps from Security: - Connected Papers says it stores this in Microsoft Azure, sends no marketing messages, does not share your details with third parties for marketing, and gives you the standard European data rights — access, correction, deletion, restriction, objection and portability. Its terms of service are dated 2023-02-21; the privacy policy shows no revision date at all, so read both as the version being served today rather than one that was recently reviewed. Best at: - Finding the papers around a paper you already have — the ones that draw on the same sources it does - Reading a field's shape at a glance - Checking a reading list you already have for holes, before you commit to it Okay at: - Handing you the paper itself — the map links out to arXiv, PubMed, DOI pages and PDFs where they exist, but Connected Papers hosts none of them - Fields where the underlying database has parsed fewer references; the map can only be as complete as what Semantic Scholar has indexed Avoid: - Using a graph as your evidence that you've covered the literature. The placement is a guess computed from overlapping citations, not a search of what has been written, and the vendor's own terms of service disclaim any warranty as to the accuracy or reliability of the material. - Starting from a paper with few references or few citations. Connected Papers will refuse outright with Not enough citations and references to build a graph — and a paper just over that line tends to produce a thin map that still looks as confident as a dense one. - Assuming a paper missing from the map isn't relevant — anything the database failed to parse never appears at all, and the picture looks equally full either way. Starter: Spend your first graph on a paper you know well Don't start with the paper you're unsure about. Paste in a paper you have already read and could argue about, and look hard at what comes back. You will be able to tell straight away whether the map is placing things sensibly. Circle size here is a citation count and circle colour is a publication year — two fixed encodings you have to learn to read, and the only way to learn them is on papers whose vintage and standing you already know. If the result looks wrong rather than unfamiliar, check you clicked the right paper in the autocomplete list — two papers can share a title. Starter: Read the Prior Works list before you read the graph Open the Prior Works tab. It lists the papers most often cited *by* the papers on your map, which in practice means the foundational work everyone in that corner of the field builds on. Click one and the map highlights every paper on it that cites that work. The list can be downloaded if you want it out of the browser. Prior Works is computed from the graph you just built rather than from a global citation count, so it surfaces what is foundational for this specific cluster — not what is famous overall, which is what a Google Scholar 'cited by' number gives you. Starter: Use Derivative Works to find the review you should have read first Open the Derivative Works tab. It lists papers that cite many of the papers on your map at once, which usually means either a review of the whole field or recent work at its edge. If you are new to the topic, one good review here saves you a week of reading. Because it ranks on how many of your graph's papers a later work cites, it reaches field-level reviews that a keyword search for 'review' or 'survey' will miss — the useful ones frequently don't say so in the title. Starter: Build a second graph from the best paper the first one found Pick the paper on your map that you most wish you had known about, and start a fresh graph from it. Where the two maps overlap, you have found the papers you can be most confident actually matter to your question. Each graph is anchored to its starting paper and recomputed from scratch rather than filtered, so moving the starting paper produces a genuinely different map — and the overlap between two of them is information no single graph contains. Connected Papers also builds multi-origin graphs from several papers at once, on every plan including free, if you would rather combine than compare. Starter: Check the build date before you trust a graph someone sent you Every graph carries the date it was built, with a Graph History panel listing earlier versions of it. If someone shares a Connected Papers link with you, look at that date first. If the site offers you a rebuilt version, take it — the database behind the map moves, and so do the citation counts that set circle sizes. Graphs here are saved objects with their own URLs and version histories rather than live queries. That is what makes them shareable, and it is also why a stale one looks exactly as authoritative as a fresh one. Pitfalls: - It looks like a citation tree and is not one. A strong line between two circles does not mean one cited the other — it means they draw on many of the same sources, or that many of the same later papers cite both. Connected Papers says this outright on its about page; the picture says otherwise, and the picture usually wins. - Very recent work barely shows up. A preprint (a paper posted publicly before peer review) from last month has almost no citations yet, so the arithmetic that places papers on the map has nothing to work with. The newest thing on your map is usually months or years old. - The citation counts come from Connected Papers' own source database and will not match the number Google Scholar shows for the same paper. Circle size is built on those numbers, so a paper can look smaller on the map than its reputation. - The picture is the easy part; the lists are where the work happens. List view sits alongside the graph, and people who only ever look at the picture get far less out of the tool than people who read the tables. Where to next: - Research tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Document and PDF tools: https://www.launchpadhq.io/discover?category=document-pdf-processing - AI chatbots: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Consensus Canonical: https://www.launchpadhq.io/platform/consensus/guide Data: https://www.launchpadhq.io/data/v1/guides/consensus.json Last reviewed: 2026-09-15 Tool: Consensus (https://www.launchpadhq.io/platform/consensus): starting price Free, from $8.99/mo; free tier yes; data as of not dated Ask a research question in plain English and get an answer built only from published papers — including the catch buried in the word "published." You type a research question into one search box in plain English, the way you'd ask a colleague. You get back a written answer with a citation attached to each claim and the papers listed underneath. Hover or click a citation to see the passage it came from, then ask a follow-up to narrow it. What it is: - Consensus is a search engine for published science. It answers a question by finding papers and quoting them, rather than by writing from whatever it absorbed in training, and most of what it draws on has been peer-reviewed (checked by other scientists before a journal publishes it). - It searches a collection of roughly 220 million papers, assembled from Semantic Scholar, OpenAlex, PubMed and its own web crawl, and refreshed weekly. Consensus's current documentation says "over 220 million" while its older pages still say "over 200 million", so read that as an approximate size rather than a count. Consensus describes it as the second-largest research index in the world after Google Scholar. - Ask a yes-or-no question and it adds the Consensus Meter: a tally of how the top papers it read came down — Yes, No, Possibly or Mixed — with the paper count behind each. - It is deliberately narrow. It has nothing useful to say about news, markets, company documents, or any question whose evidence has not been published in a journal. Before you start: - Free to use once you create an account — Consensus advertises no trial you have to remember to cancel, though no public page says whether a card is requested at signup, so expect to check that yourself. The free plan gives you unlimited ordinary paper searches (the ranked list of papers, no AI answer) and then meters the AI work on top, with small monthly allowances of Pro messages, Deep reviews and Study Snapshots. No vendor page states whether the Consensus Meter rides on the unlimited search underneath or spends one of those Pro messages, so you may find it counts against your allowance once you are signed in. - Consensus's own pages disagree about what the plans are. Its help centre subscription table, checked 15 September 2026, lists Free, Pro and a pricier Deep plan. A marketing pricing page still live on the same site advertises a cheaper "Premium" tier with a different feature list. The help centre and the product documentation agree with each other, so treat those as current — but check the price on the checkout screen before you pay. - Paying is for people running literature reviews week after week. A student settling an argument or anyone checking a health claim they read online will rarely reach the monthly ceiling. - It runs in a browser. There is an app in the iOS App Store using the Consensus name, but it appears to be published under an individual's name rather than the company's, so use the website unless you can confirm otherwise. Security: - Two Consensus documents say different things about what happens to material you upload. Its documentation says files you add to a Collection are "private, not shared with anyone, and not used to train models," and its privacy policy (effective 18 August 2026) says it does not use your data to train large language models and does not sell it. Its terms of service (dated 25 June 2025) separately grant Consensus an "irrevocable, non-exclusive, royalty-free, fully-paid, transferable, sub-licensable, perpetual" licence to use "User Materials." The terms are the binding document, so that licence is what you are agreeing to, whatever the friendlier page says. Security: - The privacy policy is unusually specific about where your text actually goes: to produce results, Consensus sends your queries and your uploaded documents to OpenAI, Anthropic, Google and Baseten under enterprise agreements. That is ordinary for a product of this kind, and it is worth knowing before you paste in anything you have not published. Best at: - Settling a yes-or-no question about the evidence — "does X help with Y" — and seeing which way the papers actually went - Finding the research behind a health or nutrition claim you saw somewhere else - Getting oriented in a literature you have never read, in language you can follow - Working across fields at once, since the index spans all scientific disciplines rather than one Okay at: - Very recent work — the index refreshes weekly, but journals and the aggregators it draws on lag publication themselves - Languages other than English — Consensus said in a 2024 changelog entry that it searches in around 100 of them, but its current documentation does not repeat the claim, so assume English is the best-supported until you have tested your own Avoid: - Treating the Consensus Meter as a verdict. Consensus's own help pages say the meter is "not a perfect reflection of all the science on a topic," that it "will occasionally incorrectly classify results," and that it "does not always account for details included in the original question." - Reading a blank or thin meter as "there is no evidence." The meter needs at least five relevant papers among the top twenty before it displays at all, so its absence tells you about your search, not about the science. Starter: Ask it the yes-or-no way The Consensus Meter only appears when your search is phrased as a yes-or-no question — the same topic typed as the keywords creatine cognition returns papers but no tally at all. Elicit, the closest comparison, treats both phrasings the same way and has no equivalent readout. Swap in any claim you have seen repeated online. The closer your wording gets to something a single study could actually test, the tighter the tally. Prompt: Does creatine supplementation improve cognitive performance in healthy adults? Starter: Read the answer by its citations, not its prose Hover or click each citation in the written answer to open the passage in the paper it came from, then check that the passage really says what the sentence claims. Consensus's own limitations page concedes that the AI "can misinterpret a paper and summarize it incorrectly" — and a misreading looks exactly like a correct summary until you open the passage behind the link, which is one click away. Starter: Spend one Deep review deliberately Switch the search mode from Pro to Deep, then ask a question with two or three parts to it. Pro and Deep are the same search box with a different budget behind them — up to twenty papers read in one pass, versus fifty to a hundred across as many as twenty searches run in parallel. Deep is the mode that finds the paper your first phrasing missed. Ask the same question in both modes back to back and compare which papers only Deep surfaced. Starter: Turn the preprints off and re-run it Switch on "Exclude preprints" in the filters and run your question again. Preprints (papers posted publicly before peer review) are in the index and labelled as such, so this one toggle shows you how much of an answer is resting on work no journal has vetted yet. The study-design filters are worth a look too — Consensus added 19 of them in July 2026, covering things like meta-analyses and randomised trials. Starter: Ask it what it cannot see Consensus can only answer out of papers it retrieved, so it reports a gap only where some paper says there is one — it has no way to tell you about a question nobody has published on. Asking directly is where that limit becomes visible. Follow it with "which of these findings failed to replicate?" Prompt: What are the main gaps and unanswered questions in the research on [your topic], and which studies argue the evidence is still unsettled? Pitfalls: - A tally of published papers is not a tally of the evidence. Studies that find nothing are far less likely to be published in the first place, so a meter reading "Yes — 9 of 11 papers" is counting what reached a journal, not measuring what is true. Consensus's limitations page admits its coverage is incomplete, but it never names this particular gap, so holding it in mind is your job rather than the product's. - Full text is uneven, and that shapes the answer. Consensus has complete text for open-access papers and for publishers it has struck deals with — Wiley, Oxford University Press, SAGE, BMJ and JAMA among them — and only the abstract for everything else. An answer can lean on an abstract's headline finding while the caveats that qualify it sit behind a paywall. - Retracted papers — work a journal has formally withdrawn — are badged, not hidden. Consensus marks them with a RETRACTED badge and never uses them in AI answers, but they still appear in the paper list. If you are browsing the list rather than reading the written answer, check the badges before you cite anything. - "Not enough evidence" is a real answer here, and easy to misread. Consensus says it will tell you when it cannot find sufficient relevant research rather than inventing a bridge, which is a genuine strength. But a beginner reads that as "science has not looked at this" when it usually means "this phrasing did not retrieve it." Try the question a second way before you conclude anything. - Medical mode is a different, much smaller collection. Switching to it narrows you from the full 220-million-paper index to roughly 8 million papers plus around 50,000 clinical guidelines. That is the right trade for a clinical question and the wrong one for everything else — and it is easy to leave switched on by accident. Where to next: - More research tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Working with documents and PDFs: https://www.launchpadhq.io/discover?category=document-pdf-processing - Learning and study tools: https://www.launchpadhq.io/discover?category=education-learning-ai ### Getting Started with Copy.ai Canonical: https://www.launchpadhq.io/platform/copy-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/copy-ai.json Last reviewed: 2026-10-14 Tool: Copy.ai (https://www.launchpadhq.io/platform/copy-ai): starting price $29/mo billed monthly; free tier no; data as of 2026-10-14 An AI writing assistant for marketing copy — ads, emails, social posts, and blog drafts from a prompt or a template. You either pick a template — fill in a product and an audience, get several variations back — or chat with it like a general assistant to draft and refine. Either way you read what comes back, ask for another pass or a different angle, and then edit the result yourself before it goes anywhere. What it is: - Copy.ai is an AI writing tool built for marketing content. At its core is a chat interface plus a big library of templates — ad copy, email subject lines, social captions, product descriptions, blog outlines — and a Brand Voice feature that learns your style. (It's since grown into a broader "go-to-market" platform with multi-step workflows, but for getting started, it's the chat-and-templates writing assistant.) It runs on multiple underlying AI models, which you can switch between depending on the task, and is aimed at marketers, small businesses, and creators who produce a lot of short-form copy. Before you start: - Copy.ai's public pricing page (copy.ai/prices, as of 2026-10-14) lists no free plan. Self-serve Chat is a monthly plan, cheaper billed annually; team Growth costs far more. Treat any leftover claims of a free word allowance elsewhere as stale until you see them on that page. - It writes text only — no images, no publishing/scheduling — so it's one piece of a content stack, not the whole thing. - Go in expecting to edit everything it produces; the getting-started skill is prompting well and then rewriting, not publishing raw output. Security: - On data, it's reassuring: Copy.ai states it does not use your content to train its models, it's SOC 2 compliant, and it offers data agreements for business accounts. (Free-tier data may be retained a bit longer than paid.) Security: The real caution is the output, not the data: - AI copy can be generic, off-tone, or factually wrong — and long-form drafts especially can read as obviously AI-written. Edit and fact-check everything before it goes out. - Generated text can sometimes echo published material closely enough to trip a plagiarism checker — run originality checks on anything you'll publish under your name. - As with any AI writing, a purely AI-generated piece may not be copyrightable — your edits and creative direction are what make it yours. Security: - Don't paste confidential client data, unreleased plans, or personal information into prompts you don't have the rights to use — good practice with any cloud AI tool. Best at: - Fast first drafts of short marketing copy — ads, emails, captions - Beating the blank page and generating lots of variations to pick from - Keeping a consistent brand voice across channels once it's trained - Non-writers producing usable marketing content quickly Okay at: - Long-form articles (drafts can feel generic; need real editing) - Factual accuracy (it can confidently state wrong things — verify) - Originality out of the box (check for plagiarism on published work) Avoid: - Publishing first-draft output without editing and fact-checking - Treating it as a research source (it isn't grounded — verify claims) - Pasting confidential or client-owned material into prompts Starter: Generate ad variations Short-form marketing copy in volume is exactly what Copy.ai is fastest and most useful at. Swap in your own product and audience; generate, then pick the one closest and rewrite it. Prompt: Write 5 short Facebook ad captions for a local yoga studio's beginner class. Warm, welcoming tone, each under 25 words, with a clear call to action. Starter: Train your brand voice Paste a few samples of your existing writing so Copy.ai learns your tone, then have it draft in that voice. Brand Voice is what stops the output sounding like generic AISpeak — it's the feature worth setting up early. Feed it your best-written examples, not your quickest ones. Starter: Draft, then edit hard Shows the real workflow — a fast draft you then cut, correct, and make human. After it drafts, ask it to "make it more specific and less generic," then do a human pass. Prompt: Write a 150-word intro for a blog post about saving time with meal prep, friendly and practical. Pitfalls: - It's a draft engine, not a publish button. Edit and fact-check everything; raw output reads as AI-written. - Watch for accidental plagiarism. Run originality checks on anything you'll publish under your name. - Pricing has been restructured and ranges widely. Confirm the current individual writing tier before you assume a number. Where to next: - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai - Workflow & Automation: https://www.launchpadhq.io/discover?category=workflow-automation ### Getting Started with Cursor Canonical: https://www.launchpadhq.io/platform/cursor/guide Data: https://www.launchpadhq.io/data/v1/guides/cursor.json Last reviewed: 2026-07-13 Tool: Cursor (https://www.launchpadhq.io/platform/cursor): starting price Free, from $8/mo; free tier yes; data as of 2026-10-25 An AI-first code editor that understands your whole project and edits across files as you describe what you want. You install Cursor as a desktop app, open a project, and work as you would in VS Code. Alongside the editor you can chat with the codebase, accept suggestions as you type, or describe a change and let the agent edit the files for you — then read what it changed and accept or reject it. What it is: - Cursor is a code editor built around AI — it's a fork of VS Code, so it looks and works like the editor many developers already know, with AI woven through it. Beyond autocomplete, you can describe a change in plain English and its agent edits across multiple files, and it can reason about your whole codebase, not just the file you're in. It's aimed at people who write code regularly; a total beginner can use it, but it assumes you're working in a real project. Before you start: - The free Hobby plan needs no card and includes a short trial of the paid features plus limited ongoing use — enough to decide whether an AI-first editor fits how you work. Pro is a flat monthly plan; current prices are on its Launchpad page. - Download Cursor only from cursor.com. Cursor has no authorized resellers; a "Cursor subscription" sold anywhere else may be fraudulent and can get the account suspended. - The free tier restricts context to the current file (no whole-codebase indexing or the multi-file agent) — enough to try the editor, not to feel its main advantage. - Ownership note: Cursor's maker, Anysphere, agreed in 2026 to be acquired by SpaceX (tying it into the xAI/Grok orbit). As of this review the deal is pending and hasn't changed the product, pricing, or how your data is handled — but if a change of control matters to you, watch for updated terms. Security: - To help you, Cursor sends the code it's working with to its servers and its AI model providers. On a personal project that's usually fine; on a work codebase, check your employer's policy first. Security: Privacy Mode is the control worth setting: - Cursor has a Privacy Mode (in settings, or set by a team admin). With it on, Cursor guarantees your code is not stored by its model providers or used for training. - If your code is sensitive, turn Privacy Mode on before you start, not after. - Never paste production secrets, customer data, or private credentials into any AI coding tool, Privacy Mode or not. Security: - The AI can be confidently wrong — it will suggest code that looks right and doesn't work, or that carries a subtle bug. You're still the author: read what the agent writes before you accept it, especially on multi-file edits. Best at: - Editing across many files from a single plain-English instruction - Understanding and navigating an unfamiliar codebase - Fast refactors, test-writing, and repetitive changes - Feeling like the VS Code you know, with AI built in Okay at: - Very large agentic tasks (workable, but review closely and mind the credits) - Predictable monthly cost (the credit pool can surprise heavy frontier-model users) - Absolute beginners with no coding context (it assumes a real project) Avoid: - Accepting multi-file agent edits without reading them - Pasting secrets or regulated data (turn on Privacy Mode first regardless) - Assuming Auto mode and frontier-model mode cost the same — they don't Starter: Download and open a project Install Cursor from cursor.com, open an existing project folder (or a fresh one), and sign in. It'll feel like VS Code. Everything depends on this, and the official site is the only safe source. Already use VS Code? You can import your settings and extensions. Starter: Turn on Privacy Mode In settings, enable Privacy Mode so your code isn't retained or used for training by the model providers. It's the one setting to decide deliberately before you point it at anything real. On a team plan, an admin can enforce this for everyone. Starter: Start on the free Hobby plan Use the free tier's trial and limited usage to try the agent and codebase chat before paying. You can judge the AI-first workflow at zero cost. Students had a discounted route historically — check current offers before paying. Starter: Ask it to explain, then change something Uses the two things Cursor does best at once — understanding your code and editing it from a description. Point it at a small real file you already have. Prompt: Explain what this file does in plain English. Then add input validation to the main function and write a quick test for it. Pitfalls: - Auto mode is unlimited; frontier models aren't. The credit pool only depletes when you manually pick the most powerful models — that's what causes surprise bills. - The free tier hides the main feature. Whole-codebase context and the multi-file agent are paid; the free tier is for evaluating the editor. - Buy only from cursor.com. There are no authorized resellers; anywhere else may be fraudulent. Where to next: - AI Coding & Development: https://www.launchpadhq.io/discover?category=ai-coding-development - Local & Open-Source AI: https://www.launchpadhq.io/discover?category=local-open-source-ai ### Getting Started with Deepgram Canonical: https://www.launchpadhq.io/platform/deepgram/guide Data: https://www.launchpadhq.io/data/v1/guides/deepgram.json Last reviewed: 2026-10-14 Tool: Deepgram (https://www.launchpadhq.io/platform/deepgram): starting price $0.0048/min promotional (strikethrough regular $0.0077/min). $200 credit on signup.; free tier yes; data as of 2026-10-14 Speech recognition priced by the minute of audio, built for developers putting it inside something else — plus one Mac app you can simply install. You open Deepgram's playground in a browser, point it at a piece of audio, and a transcript comes back within a few seconds with the words laid out on screen. Change a setting, run it again, and compare. When you want it inside your own software, that same request comes from a code file instead. What it is: - Deepgram does one narrow thing at industrial scale: it listens to audio and writes down the words, usually faster than the audio took to play. It runs the trick backwards too, reading text aloud in a synthetic voice. What it does not do is keep anything — no library of your recordings, no notes app, no Friday summary. - So the real decision is what you are shopping for. If the goal is your own meetings written down and searchable, tools built for that job — Otter and tl;dv on this site — will get you there sooner. Deepgram is sold to the people who build tools like those, which is why it is billed by the minute of audio rather than by the person using it. Before you start: - You can look before committing anything. Deepgram runs a playground page in the browser, and its own getting-started documentation sends beginners there to try the service without writing code or signing up. What that documentation does not say is whether the playground takes audio of your own or only runs Deepgram's sample — check that first, because the question that matters is whether it is accurate enough on your kind of audio. - Everything past the playground assumes code: using Deepgram properly means putting an API key (a password that identifies your app, and that spends your money) into a program. If you do not write code and have no plans to, that is not a shortcoming on your side — it is the shape of this product, and the meeting-notes tools under Where to next are the finished-article version of the same idea. - Signing up is free, Deepgram's documentation says it includes a starting balance of free credit, and the pricing page currently says no credit card is required. Transcription is billed by the minute of audio. It is a fixed pot that drains as you use it. Security: - Your audio travels to Deepgram's servers, which it says sit in the US, and comes back as text. Its compliance position is unusually strong for a tool a beginner might meet: SOC 2 Type 1 and Type 2 certification (an independent audit of how a company handles customer data), stated readiness for European privacy law, and healthcare paperwork on request. Security: Two things that are specific to Deepgram rather than to speech tools generally:: - Deepgram's Terms of Service currently say it may use your content "to improve our Services... including training and testing our Models," under a licence it calls perpetual and royalty-free. No account-level switch turns this off. The documented way out is adding mip_opt_out=true to each request — so it lives in the code, must be there from the first request, and cannot be applied backwards to audio already sent. - Treat the API key as a card number rather than a password. It authorises spending, not merely access: whoever copies it can burn through the free credit and then whatever payment method the account carries, and a stolen key looks like ordinary use from Deepgram's side. Best at: - Transcribing recordings in bulk — billed per minute of audio, currently fractions of a cent, with no per-person charge - Transcribing live, while somebody is still talking — which is why it sits inside voice assistants and phone systems - Sorting out who said what in a multi-speaker recording — speaker labelling is a documented setting on the same request, not a separate product Okay at: - Non-English audio — 50-plus languages transcribe, but Deepgram's docs reserve the full set of readability formatters for English, so a correct transcript in another language can still come back looking raw - Text-to-speech, priced per thousand characters — real, but not what Deepgram is known for Avoid: - Judging it on your very first transcript — smart_format is documented as defaulting to false, so untouched output comes back accurate but with no punctuation and no paragraphs - Budgeting from the headline streaming rates — the pricing page currently marks several of them limited-time promotional, and Flux text-to-speech is free only through 12 September 2026 Starter: Look at what Deepgram returns before anything formats it Runs audio through the playground — your own file if it accepts one, Deepgram's sample if not — with smart_format left at its default, and reads the raw transcript that comes back. smart_format is documented as defaulting to false, so Deepgram's untouched output arrives as a single unpunctuated block. Meeting that once, deliberately, is what stops you from concluding the recognition is poor when what you are looking at is a setting nobody switched on. If the playground lets you supply your own file, use audio containing numbers, dates or a phone number — the parts the formatter visibly rewrites, and only for English. Starter: Do the arithmetic before you write a line of code Estimates your monthly bill from the hours of audio you expect to process and the per-minute rate on the pricing page. Hours of audio per month × 60 × the per-minute rate. Billing is by the minute (current rates are on its Launchpad page), and the per-minute rate is low enough that the free credit covers hundreds of hours rather than a weekend trial. That multiplication tells you whether you are evaluating a purchase or something that stays effectively free for a year. Run it twice, at the pre-recorded rate and at the streaming rate. Live transcription costs more per minute, and that gap is frequently the whole design decision. Starter: On a Mac and not writing code? Install Saga instead Installs Deepgram's own Mac dictation app and dictates a paragraph into it, to hear how its recognition handles your voice with no code involved. Saga is Deepgram's own dictation app and the only route to its speech recognition with no program in between — currently free for early access, no card. It doubles as an accuracy test: if it mishears your accent, the paid service will too. Dictate something dense with names and technical terms. Generic prose flatters every speech tool; your own vocabulary is the real exam. Pitfalls: - Reading the free credit as a plan rather than a balance. How fast it disappears depends entirely on how many minutes of audio you push through. - Copying model names and settings out of an old tutorial. Deepgram ships changes most weeks and has already deprecated one speaker-labelling parameter; the changelog is the source of truth. Where to next: - The rest of the developer-services category: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - Speech tools that need no code: https://www.launchpadhq.io/discover?category=voice-speech - Apps that record and write up meetings: https://www.launchpadhq.io/discover?category=meetings-notes ### Getting Started with DeepL Canonical: https://www.launchpadhq.io/platform/deepl/guide Data: https://www.launchpadhq.io/data/v1/guides/deepl.json Last reviewed: 2026-10-15 Tool: DeepL (https://www.launchpadhq.io/platform/deepl): starting price €7.49; free tier yes; data as of 2026-10-14 A translator and a writing polisher in two boxes on one website. They're free to try, with one privacy rule you have to know before you paste anything. You paste or type your own text into the box on the left. DeepL detects the language and puts a translated, or polished, version in the box on the right. Click any word in the result to see alternatives, change it, and copy out what you need. What it is: - DeepL is a German company's translation service, and it's really two tools on one website. DeepL Translator turns your text from one language into another. DeepL Write keeps your text in the same language and suggests a cleaner version, fixing spelling, grammar and awkward phrasing. Both work the same way: your words go in on the left, and an edited version comes out on the right. - People choose it over a general translator because of how much control it gives over the result. You don't have to accept the first version. You can click any word in a translation to swap it for one of DeepL's listed alternatives, and in Write you can ask it to rephrase a whole sentence. The translator page says it handles more than 100 languages. Write covers a much shorter list: DeepL's developer documentation lists English, German, French, Spanish, Italian, Portuguese, Japanese, Korean and Chinese. - It's not a chatbot. You can't ask DeepL questions or give it instructions. It only ever works on the text you hand it. That makes it predictable, but it also means the box isn't a place to type "make this sound friendlier". Write gives you buttons and alternatives for that instead. - It can also translate a whole file. You drop in a Word, PowerPoint or PDF file and get a translated copy back with the original layout kept. Businesses also connect DeepL to their own software through its API (a way for programs to talk to each other without a person clicking). Paying for the API is separate from the plans described below, and a beginner won't need it. Before you start: - You can start right now at deepl.com, with no account and no card. Paste some text into the translator, or go to deepl.com/write for the writing tool. The free version has a monthly character allowance (letters, spaces and punctuation all count), a small monthly file-translation allowance with a size cap, and one glossary (your own list of how particular terms should be translated). DeepL also limits how much text you can paste at once, so very long text has to go in chunks. - Free DeepL is for text you'd be comfortable sharing. DeepL's free Terms of Use (last updated November 2024) say you "may not use the Free Services for processing content containing confidential or personal data of any kind." That rules out other people's names, addresses or details, and confidential work documents. The terms add that this kind of content "is only allowed within a DeepL Pro subscription." The privacy section below explains why. - You'll only need to pay if you translate a lot, or if what you translate is private or confidential. DeepL's own name for its paid plans is DeepL Pro. The cheapest is Individual, followed by Team and Business, which are priced per user. Each step up raises the character and file allowances, and Business has unlimited characters under a fair-use policy. All three are billed annually in euros with VAT included. Enterprise is custom-priced. DeepL doesn't publish dollar prices for these plans, so if you're outside Europe, check the price shown for your country on deepl.com/pro before you buy. - If you add any of DeepL's browser extensions (small add-ons to your web browser), there are versions for Chrome, Firefox and Edge. The extensions let you translate or polish text while you browse. You sign in with your DeepL account to use them. Security: - DeepL treats your text differently depending on whether you pay. DeepL's privacy policy, updated October 2026, says that on the free translator and free DeepL Write it processes the content you upload, and the results, "for a limited period of time to train and improve our neural networks and algorithms". Your corrections to its suggestions are used the same way. It doesn't say how long that period is. The policy offers no switch to opt out of this on the free version. The route DeepL offers is a paid subscription. Security: What paying changes, in DeepL's own documents:: - On the paid Translator, API and Write plans, the same policy says your texts "will not be permanently stored". They are deleted once the job is done and "will not be used to improve the quality of our services." - There is one narrow exception. DeepL's Pro Terms and Conditions (last updated August 2026) let it keep paid users' text, encrypted, for up to 72 hours when certain errors occur, so its engineers can fix them. Only selected staff can unlock that text, and only with management approval. After that it's deleted. - Translator and Write are paid for separately. DeepL's help article on DeepL Write says the paid version of Write, with its "maximum data security", is not included in a DeepL Pro translator plan. So text you paste into Write can still fall under the free-version rules above even after you've paid for the translator. Check which tool your plan covers before you paste private writing into Write. Security: - DeepL is based in Cologne, Germany, and its privacy policy is written under European privacy law. The policy says document translations, free or paid, are processed on DeepL's own infrastructure in Europe. You can delete your DeepL account in its account settings, or by emailing DeepL's support address, which the policy gives. Best at: - Getting the meaning of a message, email or web page in a language you don't read, then clicking individual words to choose the wording you'd actually use. - Writing in a language you know but aren't confident in. Translate your draft, then run it through Write to smooth it out, without leaving the site. - Translating a Word, PowerPoint or PDF file without rebuilding its layout by hand. - Tidying spelling, grammar and clunky sentences in English and a handful of other major languages, while seeing exactly what was changed. Okay at: - Languages beyond the big European and East Asian ones: because Write's list is short, in many languages you can translate but not get your writing polished. Avoid: - Pasting anyone's personal details, or anything confidential, into the free version, which DeepL's own free terms forbid. - Treating a translation as safe to send without anyone checking it when the stakes are high, such as a contract, a medical instruction or an official form. DeepL's free terms say it accepts no liability for the accuracy of its translations. Starter: Translate, then pick your own words Paste a short paragraph you actually need to understand or send, such as an email, a product review or a message from a friend, into the translator. Choose the language you want on the right, then click any word in the result you're unsure about and look at the list of alternatives DeepL offers. DeepL lets you click any single word in a finished translation and swap in one of its listed alternatives, without re-translating the whole thing, so you can stop at the one word that sounds wrong and fix just that. Try a sentence with an idiom in it, like "break a leg" or "it's raining cats and dogs", and see whether DeepL keeps the meaning rather than the literal words. Starter: Polish your writing in Write Paste something you've written into deepl.com/write: a cover letter paragraph, a work message, a post you're unsure about. Look at the edits it marks, turn on "Show changes" to see exactly what was altered, then click a sentence to pick a different rephrasing. Write sits on the same site as DeepL's translator, so a draft you've just translated can go straight into it in the same language without switching tools. Swap in Grammarly and that's no longer true, because Grammarly doesn't translate. Showing every change also means you learn what was wrong instead of silently accepting a rewrite. Write can turn American English into British English, or the other way round, by picking the variety in its language menu. That's useful if you're writing for readers on the other side of the Atlantic. Starter: Write in your language, send in theirs Write a short message in your own language, such as a note to a host, a question to a seller abroad or a reply to a colleague. Translate it into theirs, then paste the translation back into the translator going the other way to check it still says what you meant. Translating it back is a cheap check you can run without knowing the other language, and DeepL's word alternatives let you fix any line that comes back wrong before you send it. Remember that each run, there and back, counts against the free version's monthly character allowance. Starter: Translate one whole file Drop a Word, PowerPoint or PDF file onto the translator, something non-sensitive like a menu, a brochure or a public notice, and download the translated copy with its layout intact. DeepL's translator page invites you to drag and drop PDF, Word and PowerPoint files straight into the same box you type in. The free version's file allowance is small, so use it on something where keeping the layout actually matters. For a few paragraphs, pasting the text is cheaper. Pitfalls: - Signing up for an account doesn't make free DeepL private. DeepL's privacy policy covers "DeepL Translator / Write (free version)" in one section, whether or not you're logged in. What changes how your text is treated is a paid subscription, not an account. - DeepL's "Start free trial" buttons are for the paid plans, not for the free version. The free translator is the plain box on deepl.com/translator. A trial button leads you into setting up a paid subscription. Where to next: - International & Regional AI: https://www.launchpadhq.io/discover?category=international-regional-ai - AI chatbots and writing assistants: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with DeepSeek Canonical: https://www.launchpadhq.io/platform/deepseek/guide Data: https://www.launchpadhq.io/data/v1/guides/deepseek.json Last reviewed: 2026-10-15 Tool: DeepSeek (https://www.launchpadhq.io/platform/deepseek): starting price Free, from $0.30/$1.20 per 1M tokens in/out; free tier yes; data as of 2026-10-24 A capable chatbot from a Chinese AI lab — easy to start, with one decision to make first: where your words are stored. You type a prompt (the message you type) into the chat box on DeepSeek's website or app, and it writes back in plain language. On harder questions it shows its working first, folded away. Reply to steer it, correct it, or ask again. What it is: - DeepSeek is a general-purpose AI chatbot made by Hangzhou DeepSeek Artificial Intelligence Co., Ltd., a research company based in China. You can use it in a web browser, on your phone, or through a desktop app, and it handles the everyday jobs you would expect: answering questions, writing and rewriting, working through problems step by step, and reading photos and files you upload. - What sets it apart is less what it can do than who makes it and what they publish. DeepSeek releases its model (the AI "brain" that does the actual thinking) for anyone to download, under a permissive licence — so DeepSeek's AI can also be run on someone's own hardware, entirely outside DeepSeek's servers. That is unusual for a chatbot this capable, and it is why DeepSeek gets talked about far beyond its own app. - The desktop option, Harness Desktop, is a different kind of tool: an open-source assistant, currently in public preview, that DeepSeek describes as working with documents, spreadsheets and code and running recurring tasks for you. You do not need it to start; the chat is the beginner's entry point. - Developers can also connect their own software to DeepSeek through its API (a way for programs to talk to each other without a person clicking), paid for by use. That is a separate product at platform.deepseek.com, with its own documentation, and nothing on this page requires it. Current API prices are in the Quick facts panel beside this tutorial. Before you start: - Decide what you are willing to send to China before your first message. DeepSeek stores your data there, under Chinese law. That does not make casual use dangerous, but it should shape what you type — the security section below covers it. - Getting in is quick: go to chat.deepseek.com in a browser, or install the DeepSeek app, and register with an email address or a third-party sign-in, which is what DeepSeek's terms describe. The terms say the service is mainly intended for adults, and anyone under 18 should use it only with a parent or guardian's consent. - Check the Quick facts panel beside this page for current pricing before you rely on DeepSeek for anything regular. The only prices this site has verified are for the developer API, which charges per use — chat-app users do not need to touch it. - Expect the occasional "server busy" moment. DeepSeek runs a public status page at status.deepseek.com; if replies stall or error out, check there before assuming the problem is your connection or your account. Security: - DeepSeek's privacy policy (last updated 10 February 2026) states that it directly collects, processes and stores your personal data in the People's Republic of China, and its terms of use (last updated 27 March 2026) are governed by the laws of mainland China, with disputes heard by a court near the company's registered office in Hangzhou. The policy also says DeepSeek may share data to comply with legal process or government requests. You do not need to be alarmed to take that seriously: it simply means your chats sit under a different legal system from the one you probably live in, with fewer practical ways for you to challenge how they are used. Security: What the same policy says about your chats:: - DeepSeek's privacy policy says it uses your data to train and improve its models, and lists opting out of that as one of your rights. It describes no switch for this in the app, and it does not spell out how to make that request, so read the current policy before assuming you can opt out. - The policy says DeepSeek keeps your personal data for as long as you have an account, and longer where it needs to for legal or contractual reasons. Security: Before you type anything:: - Keep work material out of it unless your employer has said yes. Some organisations restrict it outright — Texas, for example, banned DeepSeek on state government devices in January 2025. - Leave out anything that identifies you or other people: full names with addresses, ID or account numbers, health details, photos of documents. - Deleting your account is permanent. DeepSeek's policy says a deleted account cannot be reactivated and its content cannot be retrieved — so export or copy anything you want to keep first. Security: - If you like DeepSeek but not where the chats are stored, its published models are the way round it: they can be run on hardware you control, where nothing leaves your machine. That is a heavier lift than signing in — the local and open-source AI tools on this site are the place to start. Best at: - Working through a problem step by step — maths, logic, planning — with its reasoning available for you to open and check rather than hidden. - Everyday writing, rewriting and explaining, at a level that holds up beside the best-known chatbots. - Moving between Chinese and English. Okay at: - Reading photos and files — DeepSeek's app notes say it has handled images since mid-2026 and that uploads were improved in July, but this is newer than its text work. - Being available when you want it — its own status page records outages of the chat and API, including one on the day this page was researched. Avoid: - Anything sensitive that you would not want held on servers in China under Chinese law — see the security section above for what that covers. - Questions on politically sensitive topics involving China. A September 2025 US government evaluation (NIST's CAISI) found earlier DeepSeek models repeated inaccurate or misleading Chinese Communist Party narratives about four times as often as US reference models. That study predates the current model, so treat it as a reason to cross-check rather than a settled verdict. - Posting its writing without saying where it came from — DeepSeek's own terms require you to label published output as AI-generated (more on that in the starter prompts). Starter: The worked problem (open the thinking) DeepSeek shows its working before the answer and, since its September 2026 app update, folds that working away automatically. Open it. When an answer is wrong, the mistake is usually visible in that section — and you can quote the exact step back ('your step 3 rounds too early') instead of just saying 'that's wrong.' Swap in any everyday calculation — a budget split, a travel time, a dosage chart you then double-check against the label. Prompt: A recipe serves 4 and needs 300 g of flour, 2 eggs and 250 ml of milk. I want to make it for 7 people, but I only have 3 eggs. Work out the new amounts, tell me whether my eggs are enough, and suggest a fix if they aren't. Show the steps. Starter: The post you are allowed to publish DeepSeek's terms (section 3.1) require anyone who publishes its output to label it as AI-generated, and say the app adds its own AI labels that you may not remove. ChatGPT's terms carry no general labelling requirement, so this is a DeepSeek-specific habit worth building from your first post. If the post includes names, dates or claims, check them yourself — the same terms make you responsible for the accuracy of what you publish. Prompt: Draft a short, friendly post (under 120 words) announcing that our neighbourhood book swap is moving to the library on Saturdays from 10am. Give me two versions — one casual, one more formal — and end each with a one-line note that it was drafted with AI help. Starter: The two-language check DeepSeek is built in Hangzhou, and its own Chinese App Store listing names translation as one of five headline uses alongside search, writing, reading and problem-solving. The round trip back into English is the part that protects you: it shows you what the other person will actually read. Try any language pair. Keep the pasted text free of personal details — see the security section. Prompt: Translate this into natural, polite Simplified Chinese, then translate your Chinese back into English so I can check nothing changed: '[paste a short message — for example, a note to a neighbour or a question for a shop]'. Point out any phrase that doesn't translate cleanly. Pitfalls: - Only download from DeepSeek's own pages. Search results include look-alike sites on other domains (for example deepseek.it.com) hosting copies of DeepSeek's documents. The real addresses are deepseek.com and chat.deepseek.com, and the official app is published by Hangzhou DeepSeek Artificial Intelligence Co., Ltd. If in doubt, search your phone's app store for that publisher name. - The thinking section is not the answer. It is the model working things out, and it can wander, contradict itself or abandon ideas before the final reply. Read the final answer as the answer; open the thinking only when you want to check how it got there. Where to next: - Run AI models on your own computer: https://www.launchpadhq.io/discover?category=local-open-source-ai - Other chatbots to compare: https://www.launchpadhq.io/discover?category=text-conversational-ai - AI from around the world: https://www.launchpadhq.io/discover?category=international-regional-ai - AI APIs for developers: https://www.launchpadhq.io/discover?category=ai-apis-developer-services ### Getting Started with Descript Canonical: https://www.launchpadhq.io/platform/descript/guide Data: https://www.launchpadhq.io/data/v1/guides/descript.json Last reviewed: 2026-10-14 Tool: Descript (https://www.launchpadhq.io/platform/descript): starting price Free, from $24/person/mo monthly or $16 annual; free tier yes; data as of 2026-10-14 Edit audio and video by editing the transcript — powerful, intuitive, and easier to misuse than it looks. Descript transcribes your recording and then lets you edit the media by editing the text. Delete a sentence from the transcript and it disappears from the audio and video; rearrange a paragraph and the footage follows. For anything dialogue-driven, that collapses a timeline-editing skill into a word-processing one. You play it back, cut more, play it again. What it is: - An audio and video editor built around transcript editing rather than a timeline. - Aimed at spoken-word content — podcasts, interviews, tutorials — rather than at cinematic video work. - Metered on two separate things at once: how much media you bring in, and how much AI you use on it. - Distinct from voice-generation tools: this edits recordings of real people rather than synthesising new speech, though it can do both. Before you start: - The free plan gives a small monthly allowance of media, a one-time grant of AI credits, watermarked exports at 720p, and a single seat. It is a real trial, not a demo. - Paid plans step up from Hobbyist through Creator to Business, all priced per person; see current plans on the tool page. Monthly billing costs more than annual. - Media minutes and AI credits are different meters. Manual transcript cuts barely touch credits; leaning on the AI assistant exhausts them first. Which limit you hit depends on how you work, not how much you work. - This is a desktop application. There is no meaningful mobile editing. - Cloning a voice requires a recorded consent statement and a substantial audio sample, by design. Security: - Voice cloning here is gated: creating a voice model requires recording a spoken consent statement plus a sizeable sample, which is a deliberate barrier against cloning someone who has not agreed. That gate is real and worth respecting rather than working around. Security: - The more distinctive risk is one no consent gate addresses. Transcript editing makes misrepresentation trivially easy. Deleting one word from a guest's sentence — a "not", a qualifier, a hedge — changes what they said, and the edit is seamless and invisible in the finished audio. Nothing warns you, because the tool cannot tell a tightening edit from a distorting one. Security: Working with other people's recordings: - Get consent to record before you record, and be explicit that the recording will be edited. - Draw your own line between removing filler and changing meaning, and hold it when a cut would be convenient. - Send the edit to the person for review when their words carry any weight — a claim, a commitment, an opinion attributed to them. - Treat voice cloning of anyone but yourself as needing express, specific permission, whatever the tool permits. Security: - Recordings are processed in the cloud, so interviews containing confidential or personal material are leaving your machine. Worth deciding before you upload, not after. Best at: - Cutting dialogue-driven audio and video quickly - Removing filler words and false starts in bulk - Producing transcripts, captions, and short clips from one recording - Making editing approachable for people who have never opened a timeline editor Okay at: - Multi-camera and heavily layered video, which it handles but is not built for - Music and sound design, which are outside its focus - Collaborative editing, which works but is not its strength Avoid: - Colour grading, compositing, and finishing work — use a real video editor - Editing anyone else's words without a clear line on what you will and will not change - Heavy AI use on a small plan, where credits disappear faster than you expect Starter: Edit something by deleting words Imports a short recording, waits for the transcript, then selects a few words in the text and deletes them — and plays the result back to hear them gone from the audio. The core idea sounds abstract and becomes obvious the moment you do it once. Use a recording of yourself. The first edits are always clumsy and you will not mind. Starter: Remove filler words in one pass Finds and strips hesitations across the whole recording at once. The single biggest time saving the tool offers, and the clearest demonstration of why transcript editing works. Listen back afterwards. Removing every hesitation can make speech sound unnaturally clipped — put a few back. Starter: Watch your two meters for one project Checks the media-minutes and AI-credits balances in your account before and after one project, to see which one moved more. Discovering which limit you hit first tells you which plan you actually need, and prevents the bill surprise this platform is known for. Do one project mostly manually and one mostly with the AI assistant. The difference in credit use is stark. Starter: Set your own editing line before you edit a guest Decides, before opening a guest's transcript, which cuts you will make to their words — fillers, false starts — and which you will not. Deciding what you will not change is much easier before you are staring at a cut that would make the episode better. Write it down. "Filler and false starts yes, meaning no" is a rule you can actually apply under time pressure. Pitfalls: - Two meters, not one. Media minutes and AI credits run out independently. Heavy AI users hit credits first; heavy importers hit minutes. Size your plan by which one you will exhaust. - Seamless edits can change meaning. Cutting one qualifier from someone's sentence is invisible in the output and misrepresents them. The tool will not warn you, so the judgement is entirely yours. - Bill shock is the most common complaint. Costs climb fast once a team exceeds included limits. Watch the meters during your first month rather than at the end of it. - It is not a full video editor. Colour work, compositing, and advanced effects belong elsewhere. Descript is for cutting dialogue, and it is excellent at that. - Everything uploads. Recordings are processed in the cloud. If an interview contains confidential material, that decision needs making before the upload, not after. Where to next: - Voice and speech tools: https://www.launchpadhq.io/discover?category=voice-speech - Video creation and editing: https://www.launchpadhq.io/discover?category=video-creation-editing ### Getting Started with Devin Desktop Canonical: https://www.launchpadhq.io/platform/windsurf/guide Data: https://www.launchpadhq.io/data/v1/guides/windsurf.json Last reviewed: 2026-09-07 Tool: Devin Desktop (https://www.launchpadhq.io/platform/windsurf): starting price Free, from $15-20/mo; free tier yes; data as of 2026-10-20 The code editor that used to be called Windsurf, with an AI agent built in. Devin Desktop is the app formerly called Windsurf. You install it and open a folder of code in it. Describe a job in a side panel and its agent — AI that takes actions rather than only answering — proposes edits across the files involved, for you to approve. What it is: - Windsurf, and before that Codeium. Cognition bought it in July 2025 and renamed it in June 2026 — same editor, same account, same plans, windsurf.com now redirecting to devin.ai. A tutorial saying Windsurf is still the right one. - It is honestly a developer tool, assuming files and folders. Cursor is the closest comparison; Claude Code runs in a text window with no editor around it. Its own bet is fleet management — one subscription covering the agent on your laptop, agents in the cloud, and a command-line version. Before you start: - Free to start: unlimited inline edits and tab completions, a light agent quota, a reduced model list. You need a machine you can install software on (macOS, Windows 10 or later, mainstream Linux) and a folder of code. - Usage is metered in tokens — chunks of text, roughly a short word each, how AI usage gets counted — not messages, so a big model on a big file costs far more than a quick question. Budgets refresh daily as well as weekly. - Living off autocomplete you probably will not pay; hand the agent multi-file jobs most days and you will want the entry paid plan — current prices are on its Launchpad page. This category reprices often. Getting set up safely (https://devin.ai/download — Cognition's own download page): - Download from devin.ai/download and nowhere else. The rename left stale "download Windsurf" links across blogs and aggregator sites, and a code editor runs with your permissions, on your files. - Pick the download that matches your computer. On a Mac, click the Apple menu then About This Mac — it will say either Apple silicon or Intel. On Windows, open Settings then System then About and read the system type, which will say 64-bit or ARM. Downloading the wrong one is the most common reason the app will not open. - Onboarding imports your settings and keyboard shortcuts from VS Code or Cursor. Check before switching for good: you cannot install extensions from a marketplace here, and other AI-completion extensions are incompatible by design. Security: - Your code leaves your machine: relevant files go to a cloud model on every request. Normal, not sinister — but think before pointing it at a client's private codebase, anything under a confidentiality agreement, or a project with passwords in it. No offline mode exists. Security: Before pointing it at real work: - By default Cognition may use data processed through the product to train models; the opt-out, which also switches on zero data retention with its providers, is a paid-plan setting. On free, assume your code contributes. - Cognition is independently audited on how it handles customer data (SOC 2 Type II). Enterprise agreements go further: no training without express prior written consent. - Autocomplete only adds text at your cursor; the agent creates, rewrites and deletes files and runs terminal commands project-wide. Keep it in version control and commit before setting it loose. Best at: - Multi-file changes — renaming a concept everywhere, wiring in a new page - Explaining a codebase you did not write, from the actual files - Everyday typing speed, via unlimited inline completion Okay at: - Building from nothing before you know what you want - Very large old codebases — more context costs more quota Avoid: - Work where source code cannot leave your machine - Learning to program from zero — it writes code you do not understand - Being your only copy of anything Starter: Tour a codebase you did not write A read-only tour barely touches the light agent quota the free plan gives you, and a specific answer confirms it is reading your disk rather than guessing — Cursor's free tier is shaped differently, so this ordering is specific here. Prompt: Read this project and explain what it does and where I would change the homepage. Starter: Your first agent edit, deliberately tiny Pre-2026 guides describe an approval flow around an agent called Cascade, since retired, so learn this from the current app. Reviewing before approving is the whole safety model, so keep the change small. Prompt: Change the main heading on the homepage to say Hello from my first edit. Change nothing else, then list every file you modified. Pitfalls: - Accepting large agent edits without reading them. The approve button is right there and reading nine files of changes feels like it defeats the purpose. It does not. Where to next: - Other AI coding tools: https://www.launchpadhq.io/discover?category=ai-coding-development - Run models on your own machine: https://www.launchpadhq.io/discover?category=local-open-source-ai ### Getting Started with Docsumo Canonical: https://www.launchpadhq.io/platform/docsumo/guide Data: https://www.launchpadhq.io/data/v1/guides/docsumo.json Last reviewed: 2026-09-09 Tool: Docsumo (https://www.launchpadhq.io/platform/docsumo): starting price Free, from See pricing; free tier yes; data as of not dated Pull the numbers out of stacks of business paperwork — and see exactly where each one came from. You sign in, pick the kind of document you are handling, and upload a file. Docsumo reads it and shows you the values it found next to the page it found them on. You click through each one, fix anything wrong, approve it, and the finished data goes out to a spreadsheet or another system. What it is: - Docsumo is a business tool for pulling structured data out of large numbers of documents — invoices, bank statements, utility bills, pay slips, ACORD insurance forms, W-9s, bills of lading — and pushing that data into whatever system needs it next. - It is aimed at teams handling the same document type over and over; lending and banking, insurance, healthcare, logistics and accounts payable are the sectors Docsumo names on its own solution pages. - What separates it from a chat-with-your-PDF reader is what it hands back: not an answer in a conversation, but a table of fields, each one carrying a confidence score (how sure the software is that it read that field correctly) and a link to the exact spot on the page it was read from. - That review-and-approve step is the actual product — the reading is the easy half. Before you start: - You can start without talking to anyone. Signing up at app.docsumo.com currently gets you a trial of about 1,000 pages inside a 14-day window, with ten user seats. The pricing page does not say whether a card is required at signup, so expect to find that out at the form. - After the trial, there is no published price. Both paid plans — Business and Enterprise — say "Talk to us", so continuing means a sales conversation and a quote based on your volume and document types. If you are one person with a stack of PDFs rather than a team with a repeating process, Acrobat AI Assistant on this site is the gentler starting point and ends in a price you can see. - Bring the documents you actually receive, not a tidy sample: a clean demo invoice tells you nothing about your scans, your layouts and your suppliers' formatting. - Extraction currently stops after the first 20 pages of a file on the documented upload routes — browser, email and programmatic. Longer files still upload and process; only the first 20 pages get read. Files are capped at around 25 MB. Docsumo's lending pages advertise 200-page financial documents, so higher limits may exist on paid plans — worth asking before you assume either number. - It expects a repeated shape. Around 50 common document types come pre-trained; for anything else you teach it yourself from a sample set, which Docsumo puts at about 20 examples. Security: - The documents you put into Docsumo are usually not yours. They are your borrowers', your patients', your policyholders', your suppliers' — which changes the question from "am I comfortable with this" to "am I allowed to do this." Security: What Docsumo says it holds: - SOC 2 (an audit of how a company handles customer data) — Docsumo says it certified to Type I in September 2021 and to Type II roughly six months later. - GDPR (European privacy law governing what companies may do with your data) — its privacy policy says data is stored in accordance with it, and its published list of outside companies it relies on shows hosting regions in the US, EU, UK, Australia and India. Security: - The passage worth reading yourself is in the privacy policy, last updated January 2024: uploaded documents are "retained and used for further research, development and training of the artificial intelligence." A 2025 company blog post about the SOC 2 audit says the opposite — that collected data "is not used by the platform in any way." A blog post does not override a privacy policy. If your files carry other people's financial details, get that contradiction resolved in writing before you upload anything real. Security: - The same public list of outside providers names Anthropic and OpenAI alongside AWS, Google Cloud, MongoDB and Cloudflare, so document content can leave Docsumo's own systems and reach third-party AI services. Deletion is by written request to [email address on the web page], which the policy says is actioned within 30 days, with account data removed at the latest two months after an account closes. Best at: - High-volume, repetitive paperwork where the same handful of fields matter every single time. - Producing data another system can swallow. Docsumo names business systems it syncs into — NetSuite, SAP, Encompass, Epic, Guidewire — and offers spreadsheet downloads and automatic hand-off to other software. - Work where a person has to be accountable for the numbers — the click-to-source check turns verifying a value into a glance instead of a re-read. - Document types it already knows, so common paperwork works on day one. Okay at: - Document types Docsumo has never seen, until you have taught it one from your own samples. - Poor scans. It straightens page orientation for you and says higher-resolution images extract more accurately, which is a polite way of saying a phone photo of a creased receipt is the hard case. - Spreadsheets and other already-digital files: .xlsx and .xls upload alongside PDFs, JPGs, PNGs and TIFFs, though the platform's centre of gravity is clearly scanned paperwork. Avoid: - Treating a high confidence score as a signature. It is the software's own estimate of its own reading — a prompt to look, never a substitute for looking. - One-off documents you only need to understand once. The output is a table of fields you defined in advance, not an answer to a question, so a document whose shape you will never meet again has nowhere to land. - Assuming a pre-trained document type fits your issuer. "Bank statement" covers a great many banks; Docsumo's own setup flow expects you to test it, correct it and retrain against your actual forms — the pre-trained version is where the work starts, not where it ends. Starter: Run your worst document first You upload the worst-quality document you genuinely receive rather than a clean sample, then check every extracted field against the page before approving anything. The trial's pages are capped, so what you spend them on decides what you learn — and because extraction quality tracks scan quality here, a clean sample answers a question you never had. Then run the same document type at its best quality and note the gap. That gap is your review workload, forever. Starter: Click a field and watch the page light up Opening a processed document and clicking each extracted field in turn highlights the region of the page it was read from, so you can confirm that region really holds the number you wanted. Docsumo's product pages are where the click-to-highlight behaviour is documented, and verification being one click rather than a hunt through the original is the specific thing a trial exists to test. Deliberately choose a document with two similar numbers on it — a subtotal and a total, a due date and an invoice date — and see which one it grabbed. Starter: Teach it one document type it does not know You pick one form Docsumo has no pre-trained type for, gather twenty real examples of it, and train a custom type on them before deciding whether the tool fits. Twenty examples is the cheapest test of whether your non-standard paperwork is reachable at all — and if you cannot find twenty, you have learned something about whether it is worth automating. Do this in week one. If it needs correcting, you want the correcting to happen while pages are still free. Starter: Send the output somewhere real You process ten documents and approve them, then export the results in the format your team would really use and check that the columns line up with what the receiving system expects. Extraction accuracy is the half everyone tests; the export shape is the half that breaks integrations later, and both export routes are on the trial plan, so this is checkable before any sales conversation happens. Try the email route too — Docsumo accepts documents as email attachments, with a total email size limit of around 25 MB. Starter: Get the quote inside the trial window In the first few days you send Docsumo your monthly page volume and document types, asking for a written quote alongside written answers on document retention and training. Because the price only exists as a quote (see Before you start), the 14-day window is also your negotiating window — and asking in week two means deciding with the trial already expiring. Pitfalls: - The accuracy headline is a vendor claim, and Docsumo's own pages do not agree with each other about it — its custom-model feature page says 90%+, its product and lending pages say 95%+. Treat either as marketing. The missing percent is not spread evenly, either: it clusters on your worst scans and your oddest layouts. - Unreviewed output is the expensive failure mode. The design routes low-confidence items to a person and lets the rest flow, which means a value the software was confident about and wrong about goes straight through — into a spreadsheet, an underwriting file, a payment run — with nobody having looked at it. - Setup is front-loaded and easy to under-budget in a 14-day window. You define a document type, choose or auto-detect its fields, run a test, correct the results, save, and only then process at volume. Teams routinely spend the whole trial building and never reach the accuracy question they started for. - The privacy policy is dated January 2024 while the product has visibly moved on — it is now described in terms of AI agents, and the model APIs on its provider list did not exist in that form when the policy was written. A document that old may not describe what happens to your files today; ask rather than assume. - The trial's ten seats invite you to hand accounts around. Docsumo describes the software adapting to reviewer corrections, so the people you invite are also the people shaping how it reads your documents. Decide who is allowed to approve before you invite anyone. Where to next: - Other document and PDF tools: https://www.launchpadhq.io/discover?category=document-pdf-processing - AI inside business software: https://www.launchpadhq.io/discover?category=ai-plugins-business-software - Workflow automation: https://www.launchpadhq.io/discover?category=workflow-automation ### Getting Started with Dragon Medical One Canonical: https://www.launchpadhq.io/platform/dragon-medical-one/guide Data: https://www.launchpadhq.io/data/v1/guides/dragon-medical-one.json Last reviewed: 2026-09-12 Tool: Dragon Medical One (https://www.launchpadhq.io/platform/dragon-medical-one): starting price See pricing; free tier no; data as of not dated Speak your clinical notes straight into the patient record — a Windows app your employer licenses for you, not one you can sign up for. You sit at your work computer with the patient record open, put the cursor where the text should go, switch the microphone on, and talk. The words appear as you speak. You say a short command to fix anything wrong, then switch the microphone off. What it is: - Dragon Medical One turns speech into typed text inside the software a hospital keeps its patient records in, and steers the screen with spoken commands. - The ownership moved and the name hasn't caught up. Nuance built it; Microsoft bought Nuance, so the product page now sits on microsoft.com under Microsoft's health solutions — old nuance.com links redirect there — while downloads and support stay on nuance.com addresses carrying a 2026 Nuance copyright. Microsoft also sells a newer, separate product called Dragon Copilot and points Dragon Medical One customers towards it; the two are not the same thing. - Nobody outside a subscribing organisation can open the app to check. Everything here about buttons, commands and screens comes from Microsoft's and Nuance's own user guides and admin documentation — the vendor's description, not a walkthrough by someone who used it. Before you start: - Your employer buys this; you can't. Microsoft's product page carries no price and no trial — the main button reads "Buy additional licenses", the other route is "contact us". An administrator creates your account in the console Microsoft calls Nuance Management Center and assigns your licence. So the first step is asking your IT team whether your organisation has a seat free. - Because it is quoted rather than listed, there is no honest sticker price — and normally none of it comes out of your pocket. - It is a Windows program: Microsoft's requirements page lists Windows 10, Windows 11 and Windows Server, and names no Mac or iPad version. Your phone can stand in as the wireless microphone through a companion app called PowerMic Mobile, but recognition still happens on a Windows PC. - It types into named systems, not anywhere: Epic (including Haiku, Canto and Rover), Oracle Cerner PowerChart Touch, MEDITECH Expanse and MConnect, plus athenaClinicals and Allscripts on the vendor's version-comparison page, alongside Word, Outlook and Edge. Ask about anything outside that list before your organisation signs. - Not a clinician whose employer will license this? The dictation tools in our voice and speech section (Otter is the usual start) do the same job, minus the clinical vocabulary and record-system integration. Getting set up safely (Your IT team installs it. If you are fetching it, the installers live on the vendor's resource centre at dragonmedicalone.nuance.com — still a Nuance-branded address in September 2026, though the product pages moved to microsoft.com. Take it from nowhere else.): - The installer expects an organisation token — a code the vendor says arrives "in your Welcome Kit email" when an organisation buys licences. Without it there is nothing to install into. - Check the version. The resource centre currently offers build 2026.3, and states that version 2021.4.5 or earlier "will be end of life and no longer function after December 31, 2026". - The vendor's download page warns that "depending on your security settings, Windows might mark downloaded files as unsafe and prevent the installer from running" — normal here. - Your "it worked" checkpoint: sign in, choose your microphone, put the cursor in a note and watch your words appear. A floating box instead means it didn't recognise the field — covered below. Security: - The privacy question isn't really about you: what you dictate is someone else's medical record, under an agreement between your employer and Microsoft. That binding paperwork is not currently public. Security: What the vendor publishes, and what it doesn't: - Microsoft's healthcare compliance page, updated 20 July 2026, says a HIPAA (America's patient-privacy law) business-associate agreement is included by default in its online services terms for covered customers. Its product list names neither Dragon Medical One nor Dragon Copilot — a general Microsoft position, not confirmation this product is covered. - The subscription agreement Nuance published for the marketplace edition of Dragon Medical One no longer loads: as of 11 September 2026 its address redirects to Microsoft's healthcare homepage, so the terms in force could not be read. - The published data figures belong to Dragon Copilot, not this product. Its FAQ, updated 27 July 2026, says audio, recognised text and transcripts are kept up to 90 days, encrypted in transit and at rest, and "deidentified or pseudonymized for AI model improvement". No equivalent page for Dragon Medical One could be found, so don't assume those numbers carry across. Security: - The one piece you control is the microphone. Microsoft's guidance says to switch it off as soon as you've finished so the app doesn't "capture extraneous dictation" — worth more than it sounds with the patient still in front of you. Best at: - Long stretches of free text, landing where you put them rather than following your mouse — the docs call that anchoring your speech focus - Words general dictation tools mangle: drug names, procedures, local abbreviations - Starting without homework — Microsoft's product page says no voice-profile training is required and that it calibrates audio and detects accents automatically Okay at: - Ordinary non-clinical writing. The comparison page shows it in Word, Outlook and Edge, so it will take a letter — an expensive way to write one - Any particular voice. The "no training required, accents detected automatically" claim is the vendor's; how it holds for your accent and your room is something only using it answers Avoid: - Expecting it to sit in on the appointment and write the note. Dragon Medical One types what you dictate; the listen-and-draft product is Dragon Copilot, sold separately - A cheap headset. The vendor specifies audio at 16 kHz, 16-bit mono and publishes a list of tested microphones; a bad one is the commonest reason recognition feels worse than advertised - Assuming any text box will take dictation — the vendor publishes a named list of supported editors and record-system clients Starter: Say the punctuation out loud Dictate the marks as part of the sentence — "period", "comma", "new paragraph", "new line". Microsoft's guidance: speak "clearly and naturally as if you were speaking to someone in person", in "complete thoughts or sentences", and "dictate with punctuation". A pause is what separates text from commands: the vendor says to "pause briefly before and after saying a command so Dragon Medical One knows that you are not dictating text". Pause briefly after switching the microphone on and before switching it off — the vendor says this stops the first and last few words being clipped. Starter: Fix a mis-heard word without touching the mouse Correct by voice: "select [the word]" and "select [word] through [word]" highlight, "scratch that" or "delete that" remove what you just said, "undo that" and "redo that" step back and forward, "cap that" and "all caps that" fix capitalisation. Reaching for the mouse to fix one word gives back the time dictation saved. The docs pair this with "Anchor Speech Focus", which holds your text in the field you chose even when the screen moves. Say "what can I say" — the vendor's help article describes that opening a searchable list of commands grouped by what they do. Starter: Teach it the words it keeps getting wrong Say "add that to vocabulary" with the word selected, or "manage vocabulary", to open the Add word window. The vendor's quick reference notes you don't need to press the talk button on a PowerMic to train a word. The words it fumbles are the ones your specialty uses daily and nobody else uses at all. Vocabulary also lives at organisation level: the admin documentation describes managing it centrally, so a word can be added once for a whole group. Starter: Turn the paragraph you retype daily into a spoken shortcut Highlight text you've just dictated and say "make that an auto-text" to save it as a reusable block; "manage auto-texts" shows what you've got. Blocks can hold blanks marked with square brackets, and "next field" moves between them. A keyboard macro drops a block and abandons you in it; here one spoken phrase drops the block and a second walks you through the gaps. The admin docs also describe defining blocks for a whole group. The bracket characters are a setting — the vendor lists square brackets as the default and says the delimiters can be changed. Starter: Know what to do when your words won't go in When the cursor isn't in a field it recognises, the vendor says a dictation box opens on its own. Dictate into it, correct it there, then say "transfer text" to push the text where your cursor is; "recall text" brings it back if it landed wrong, "discard text" throws it away, and "open dictation box" opens it deliberately. It is the documented escape hatch for every screen outside the supported list, and knowing it is a designed fallback rather than a fault beats any individual command on day one. The transfer runs on a keystroke: the vendor names Shift + Insert as the default and says the method can be changed. Pitfalls: - Losing text to another window. It goes wherever the cursor is, so a notification that moves focus mid-sentence sends the rest of a paragraph somewhere you'll have to find. - Chasing help on the wrong site. Support articles sit on support.microsoft.com, admin and deployment documentation on learn.microsoft.com, and installers and version notices on a nuance.com resource centre — knowing which of the three you want saves an annoying half-hour. Where to next: - Dictation tools you can sign up for yourself: https://www.launchpadhq.io/discover?category=voice-speech - Apps that record and write up meetings: https://www.launchpadhq.io/discover?category=meetings-notes - Healthcare AI: https://www.launchpadhq.io/discover?category=healthcare-ai ### Getting Started with Duolingo Max Canonical: https://www.launchpadhq.io/platform/duolingo-max/guide Data: https://www.launchpadhq.io/data/v1/guides/duolingo-max.json Last reviewed: 2026-09-09 Tool: Duolingo Max (https://www.launchpadhq.io/platform/duolingo-max): starting price Free, from $12.99/mo; free tier yes; data as of not dated The top Duolingo tier, what it actually adds, and how to check your language is covered. You learn in the phone app the same way as always, tapping through short lessons along a path. Once you subscribe, extra buttons appear inside those lessons and on the practice screen. Tap one and you either type a short conversation with a cartoon character, or speak out loud to one and hear it answer back. What it is: - Duolingo Max is the top of three tiers: the free app, then Super, then Max. - Max includes everything Super does — no ads, unlimited hearts (a benefit Duolingo's help centre says may be part of a test it is running, so what you see may differ), personalised review of your mistakes — and adds two AI conversation features on top: Roleplay, a typed run through a set scenario with one of the app's characters, and Video Call, where you speak to a character called Lily and she answers in the language you're learning. - It unlocks no extra lessons and no extra courses. Duolingo says it offers "the same lesson content to all users", so what the money buys is practice modes, not curriculum. - Some things that used to be Max-only are now free: Explain My Answer, the in-lesson grammar explanation, went free for every learner in January 2026, so older reviews and comparison articles overstate what the tier includes. Before you start: - The AI features cover a much shorter list of languages than the app does. Duolingo's help centre currently lists Video Call for learners of English, French, Spanish, Italian, German and Portuguese, and Roleplay for English speakers learning French, Spanish, German, Portuguese or Italian — plus English courses for Spanish, Portuguese, Japanese, German and French speakers. Duolingo states it plainly: "Some AI-powered features are not available across all courses." If you're learning Welsh, Hindi or Navajo, you can pay for Max and receive only the Super benefits. - Max is a phone purchase. Duolingo's help centre says Max "cannot be purchased on web" and that "the features are only available on the mobile app", and that a subscription bought on the web cannot currently be upgraded to Max at all. If you do your learning on a laptop, this tier does nothing for you. - Duolingo doesn't publish the price anywhere public — you see it in the app's Shop, billed through Apple or Google, and it varies by country and by how long a plan you buy. It sits above Super, which is the comparison that matters. It can also be bought as a family plan shared with five other people, and Duolingo's terms refer to promotional codes for extended free trials of Super or Max, so a trial may be offered to you; the length isn't published. - Beginner-level calls with a second character — a bear called Falstaff, who asks you questions and suggests phrases rather than free-forming a conversation — were announced in January 2026 as iOS-only, with the rollout said to continue through 2026. On Android your options may currently be narrower than the marketing suggests. Security: - Duolingo's privacy policy, last revised 26 May 2026, is unusually specific about the voice features, and it's worth reading the AI section before your first call. It says that when you use Video Call or other AI features, "the text and audio you submit may be shared with AI vendors such as OpenAI and Google", that those vendors "are not permitted to use any personal information for their own purposes", and that Duolingo itself "may generate, record, and store audio recordings or transcripts" and use them "for product improvement and personalization purposes, including training and running Duolingo's own artificial intelligence models" (the model is the AI "brain" doing the actual thinking). Security: Three lines from that policy worth acting on: - Its own instruction: "Please do not submit any personal, sensitive, or confidential information when using Video Call or other AI features." Speaking is looser than typing — it's easy to say your employer's name or your street mid-sentence. - There is an opt-out. The policy says you "may also choose not to share your audio with us for product improvement purposes within the app Settings", and taking it doesn't stop the calls working. - It commits that "we will not use your audio recordings to develop any voice cloning technology." Security: - Duolingo is rated for all ages and a large share of its users are children, so the policy runs a separate regime for Child Users — under 13 in the US, or the local age of digital consent elsewhere. They register with a username rather than an email, get non-personalised ads and no third-party analytics, and, the line that matters here, their "speech data is not shared with Duolingo for product improvement purposes." Child Users can join a paid family plan. The age on the account is the switch that turns all of that on, so it's worth getting right before a child starts making calls. Security: - One setting unrelated to Max but easy to miss: Duolingo profiles are public by default and can be found by name, username, email address or phone number. You can set yours to private in Settings. Best at: - Speaking practice with nobody else in the room — you can stall, restart and be bad at it at no social cost - Building the habit of assembling a sentence under mild time pressure, which tapping and matching exercises never ask of you - Rehearsing one narrow, predictable situation before you have to live it — Roleplay's scenarios are errands like ordering a coffee Okay at: - Measurable speaking gains — Duolingo's own research lab reported in March 2026 that learners who called at least twice a day improved more on speaking measures than those doing lessons alone. It's the company studying its own product, and it hasn't published sample sizes - Genuinely open conversation — Lily adjusts to your level and won't penalise mistakes, but she steers toward the vocabulary your course has already taught - Listening comprehension — the speech is clear and paced for a learner, which is the opposite of a real conversation in a noisy café Avoid: - Getting past intermediate. Duolingo said in April 2026 that nine of its most popular courses now reach B2 on the CEFR scale — the European yardstick for language level, where B2 means "independent user", not fluent. That is the ceiling of the courses, and nothing in Max raises it. - Fixing your pronunciation. The calls deliberately don't penalise mistakes, which is exactly what stops them functioning as a pronunciation drill. - Reading and writing. Both of the features Max adds are conversation, and the tier contributes nothing to either. Starter: Open with a real errand, not small talk Say or type this as your opening prompt — the message you give the AI. Video Call is unscripted — Lily follows whatever you open with — so your first sentence is the only steering the feature gives you, and spending it on "how are you" spends the whole call. Swap in the scene you're genuinely nervous about: a landlord, a doctor's receptionist, a border guard. Prompt: I need to explain to a pharmacist that I have a headache and ask what I can take without a prescription. Let's practise that. You be the pharmacist, and start. Starter: Use the escape hatch on purpose, in the language Duolingo's own description of the feature says you can ask Lily to repeat or slow down, so the recovery move is built into the design rather than something you improvise mid-call. "What does that word mean?" behaves the same way. Prompt: Can you say that again more slowly? Starter: Run the same Roleplay scenario twice in one sitting Play a scenario through, then immediately restart it and answer differently. Roleplay scenarios are fixed and written by Duolingo's curriculum staff, so a second run is a controlled comparison — you hear precisely what you fixed, which the unscripted calls can never give you. On the second pass, try to say one thing the scenario didn't ask for. Starter: Read the transcript for the sentence you gave up on Duolingo's product posts describe a transcript and a post-call review after a call ends; if yours shows one, read it looking for the moment you switched to English or trailed off. That transcript is the only written record the app makes of your own spoken mistakes — every other exercise in Duolingo only ever shows you Duolingo's sentences. Starter: Check your own course before you pay Open the practice screen — the dumbbell icon — in the app on your phone. Duolingo's help centre says that's where free learners find the "Unlock with Max" button and where the Roleplay and Video Call entry points live. Because the two features cover different course lists, the app on your account is the only reliable answer for your specific language pairing — Duolingo's public pages list languages, not course combinations. Pitfalls: - The streak is an engagement mechanic, not a measure of learning, and Max doesn't change that. You can hold a 400-day run of one-minute lessons and still not manage the conversation you're paying to practise. - Course content is increasingly AI-generated. Duolingo announced in April 2025 that it would become "AI-first" and phase out contractors for work AI can do; the announcement drew heavy public backlash, and the CEO later softened parts of it — Fortune reported in April 2026 that he had backed away from judging employees on their AI use. The practical consequence for a learner is small but real: an odd sentence or translation in newer content is worth checking rather than trusting. - Lily won't mishear your accent because she's tired, won't change the subject, and has no queue behind her — so the calls are the warm-up for a human conversation, not the substitute for one. - You manage the subscription wherever you bought it. Duolingo's help centre says the platform you first subscribe on is where the subscription lives, and its terms say payments are non-refundable and that Apple or Google handle refunds for store purchases — so cancelling means going into the App Store or Play Store, not the Duolingo website. Subscriptions renew automatically until you do. - The calls are short by design. Duolingo has said conversations run about a minute early in a course and stretch to a few minutes as your level rises. It's a daily drill, not a lesson you can settle into, so plan for it as five minutes a day rather than an evening's study. Where to next: - More education and learning tools: https://www.launchpadhq.io/discover?category=education-learning-ai - Voice and speech AI: https://www.launchpadhq.io/discover?category=voice-speech - AI chat assistants: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with ElevenLabs Canonical: https://www.launchpadhq.io/platform/elevenlabs/guide Data: https://www.launchpadhq.io/data/v1/guides/elevenlabs.json Last reviewed: 2026-10-14 Tool: ElevenLabs (https://www.launchpadhq.io/platform/elevenlabs): starting price Free, from $5 equivalent, Creator $18.33, Pro $82.50); free tier yes; data as of 2026-10-14 Turn text into remarkably natural speech — and, with permission, clone a specific voice — across dozens of languages. You paste text, choose a voice — a stock one, a community one, or one you've cloned from a sample — and it generates the audio. You listen, adjust the text or pick a different voice, and generate again until it sounds right, then keep the take you like. What it is: - ElevenLabs is an AI voice platform. Its core is text-to-speech that sounds convincingly human; it also does voice cloning (making a synthetic version of a specific voice from a sample), dubbing, and sound effects. It's used for audiobooks, video narration, podcasts, and voice agents. You mostly drive it by typing the text you want spoken and picking a voice. Before you start: - The free tier gives a monthly credit allowance and lets you try text-to-speech and cloning — but its output is non-commercial and must credit ElevenLabs. Commercial use starts on the cheapest paid plan, Starter (official FAQ, 2026-10-14); Creator, Pro and Scale add more credits and features. - If you're going to clone a voice, sort out consent first — this is the single most important thing about the platform (below). - Credits are metered by characters of text and reset monthly; unused credits generally don't carry over. Security: Voice cloning: consent is the rule, not a formality: - You may only clone a voice you own (your own) or one you have explicit, documented permission to use. Instant cloning asks you to confirm you have that right; professional cloning adds a verification step before it will train. - Cloning a public figure or any third party without consent is both against ElevenLabs' terms (accounts get suspended) and, increasingly, against the law: a dozen-plus US states have voice-likeness laws (Tennessee's ELVIS Act among them), and the EU requires AI-generated media to be labeled. - For anything client-facing, commercial, or persuasive, keep written consent on file. Treat a cloned voice like licensed talent, not a free asset. Security: - When you upload a voice sample or generate audio, ElevenLabs keeps a broad, long-running license to use your voice data and content to improve its models. It commits not to commercialize your voice on its own without permission, but the training use is part of the deal — factor that in before uploading a sensitive or valuable voice. Security: - On the free tier, output is non-commercial and must attribute ElevenLabs; commercial rights begin on the paid plans (and persist for audio you already generated, even if you later cancel). Don't build a paid project on free-tier audio. Best at: - Natural-sounding narration for videos, audiobooks, and e-learning - Cloning your own voice to scale narration you'd otherwise record by hand - Multilingual speech and dubbing across dozens of languages Okay at: - Highly emotional or performance-heavy delivery (very good, occasionally uncanny) - Real-time/low-latency voice agents (a specific model/plan territory) - Fine control over pacing and emphasis (improving) Avoid: - Cloning anyone's voice without documented consent - Impersonation, fraud, or anything deceptive (prohibited and often illegal) - Commercial use of free-tier output, which requires attribution and isn't licensed for it Starter: Hear a stock voice read your words The fastest way to judge the quality — pick a built-in voice, paste text, generate, and listen. No cloning, no consent questions. Try the same line in two different voices to hear how much the voice choice matters. Prompt: Welcome to the show. Today we are talking about something small that changed everything: the humble paperclip. Starter: Clone your OWN voice Cloning your own voice is the standout feature and the one clone you can make with zero consent worries. Use a quiet room and a minute or two of clear audio for a noticeably better clone. Prompt: [Record or upload a clean sample of your own voice reading a paragraph, then generate this line in your cloned voice:] Thanks for listening - I recorded this without touching a microphone twice. Starter: Try a short dub Shows the multilingual side that makes ElevenLabs useful beyond English narration. Swap in a language your audience actually uses. Prompt: Translate and speak this in Spanish, keeping a warm, friendly tone: "Hi everyone, thanks so much for being here today." Pitfalls: - Consent isn't a checkbox to click past. Cloning someone else's voice without documented permission is a terms violation and, in many places, illegal. - Free output is watermarked-by-attribution and non-commercial. You can't quietly ship it in a paid project. - Credits are character-metered. Long scripts eat the monthly allowance faster than the "minutes" framing suggests. Where to next: - Voice & Speech: https://www.launchpadhq.io/discover?category=voice-speech - Video Creation & Editing: https://www.launchpadhq.io/discover?category=video-creation-editing ### Getting Started with Elicit Canonical: https://www.launchpadhq.io/platform/elicit/guide Data: https://www.launchpadhq.io/data/v1/guides/elicit.json Last reviewed: 2026-07-27 Tool: Elicit (https://www.launchpadhq.io/platform/elicit): starting price Free, from $10/mo; free tier yes; data as of not dated Searches and extracts from academic literature at speed — with a published accuracy gap you need to know about. You type a research question in plain language rather than a keyword string, and Elicit searches the academic literature and returns papers. You then name the columns you want — sample size, method, outcome — and it fills a table with those values from each paper, each cell linked to the passage it came from. What it is: - A research assistant for academic literature, searching well over a hundred million papers. - Built for finding relevant work and extracting comparable data across many papers at once. - Deliberately narrow — it works on published academic literature and cannot help with current events, industry topics, or anything whose evidence base is not academic. - Transparent by design: every extracted claim links back to its source passage. Before you start: - There is a genuinely usable free tier that needs no card. Its exact limits are described inconsistently across third-party sources — check the current allowances in the product rather than trusting a comparison article, including this one. - Paid tiers for individuals are a modest monthly subscription, with a substantially more expensive tier adding large-scale screening and systematic-review workflows. - It searches published literature. If your question is about the last six months of anything, this is the wrong tool. - It is an accelerator, not a replacement for a proper search. See the security section — this is the single most important thing on this page. Security: - The risk on this page is not privacy — it is trusting the output too much. The vendor's own benchmark reports very high screening sensitivity, tested against a large set of established systematic reviews. An independent study published in 2025 found that sensitivity dropped dramatically, to under 40 percent, when queries followed the search strategies real systematic reviews actually use. Security: - Both numbers can be true. Benchmarks measure controlled conditions; the independent test measured field conditions. The gap between them is the thing to internalise: this tool performs far better on tidy questions than on the messy ones you will actually bring it. A search that feels comprehensive may have missed most of the relevant literature without telling you. Security: How to use it responsibly: - Treat it as one search among several, never the only one. Run your database searches too. - Verify extracted data against the source passage before you cite it. The link is provided precisely so you will. - Be especially careful with anything approaching clinical, safety, or policy conclusions, where a missed paper has consequences. - Record what you searched and how, so the gap in your method is documented rather than invisible. Best at: - Finding relevant papers on a niche topic where keyword search fails - Pulling comparable data out of many papers into one table - Getting oriented in an unfamiliar literature quickly - Monitoring for new publications on a defined topic Okay at: - Comprehensive recall, which is exactly where the independent testing found it weakest - Very recent publications, where indexing lags - Interdisciplinary questions spanning academic and non-academic evidence Avoid: - Being your only literature search for anything that must be complete - Clinical or policy conclusions without full expert review - Questions whose evidence base is news, industry reporting, or anything unpublished Starter: Ask a real research question in plain language The semantic search is the genuine advantage over keyword databases, and it only shows up when you stop writing keyword strings. Compare the results against the same question in a keyword database. The overlap, and the gap, are both instructive. Prompt: What does the evidence say about [your specific question]? I want the strongest findings on both sides, not a consensus summary. Starter: Build one extraction table Pulls specified fields out of every paper in your result set into columns. This is the time saving that justifies the tool, and it is the feature people discover last. Start with three columns you would have written down by hand anyway. Ambitious tables are harder to verify. Starter: Verify five cells against their sources Clicks the link behind five cells in your extraction table and reads the passage each was taken from, checking the value matches. The single most important habit to build with this tool, and the reason the citation links exist. Pick the five that matter most to your argument, not five at random. Starter: Run the same question in a database search and compare Puts the same question, as a keyword search, into a conventional academic database and compares the two result lists side by side. Directly exposes the recall gap the independent research identified, on your own topic. Count what the other search found that this one missed. That number is your calibration. Pitfalls: - Vendor benchmarks and field performance diverge sharply. Independent 2025 testing found screening sensitivity far below the vendor's reported figure when using realistic systematic review search strategies. Do not treat a search here as comprehensive. - It is a supplement, never a substitute. Run your proper database searches alongside it. Using this alone for anything that must be complete is a methodological error, not a shortcut. - Extraction is fast and still needs checking. Every cell links to its source passage because that link is meant to be followed. Cite what you verified, not what appeared in the table. - It only knows published academic literature. Current events, industry trends, and unpublished work are outside its scope entirely. It will not tell you that your question was a bad fit. - Free-tier limits are described inconsistently online. Sources disagree about what the free plan includes. Check inside the product before planning work around a number you read elsewhere. Where to next: - Other research tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Working with documents and PDFs: https://www.launchpadhq.io/discover?category=document-pdf-processing ### Getting Started with Fathom Canonical: https://www.launchpadhq.io/platform/fathom/guide Data: https://www.launchpadhq.io/data/v1/guides/fathom.json Last reviewed: 2026-10-24 Tool: Fathom (https://www.launchpadhq.io/platform/fathom): starting price Talk to sales; free tier yes; data as of 2026-10-14 A free AI note-taker for Zoom, Google Meet and Teams that records, transcribes and summarises your calls — generous on the free plan, with sharing defaults worth changing. You connect your Google or Microsoft calendar and install Fathom's desktop app. When a Zoom, Meet or Teams call starts, Fathom records it — usually with a visible "Fathom Notetaker" bot — then gives you a transcript, an AI summary and highlights you can clip and share. On a Mac there's also a newer bot-free mode. What it is: - Fathom is an AI meeting assistant. Its homepage describes it as an AI notetaker that summarizes your meetings so you can focus on the conversation. It works with Zoom, Google Meet and Microsoft Teams, and its free plan includes unlimited recordings, storage and transcription, per its help centre. - Most people meet Fathom as a bot: a participant called "Fathom Notetaker" joins the call and records. Fathom's help centre says a newer bot-free mode is rolling out on Macs, which can capture just a transcript or audio without a bot. For Microsoft Teams, bot-free recording is so far only a limited Windows alpha, so expect the bot on Teams calls for now. Fathom records only while you are on the call and leaves when the meeting ends. - After the call you get a transcript in one of 38 languages, an AI summary using a meeting template, and highlights you can turn into short clips. Free includes limited use of Ask Fathom, a chat that answers questions about your calls; Fathom's pricing page says paid plans make it unlimited for your own calls, with unlimited lookback, and add AI action items, follow-up emails and CRM syncing. An iPhone app records in-person conversations. Before you start: - You need a Google or Microsoft account. Fathom's help centre says it only supports Google- or Microsoft-based email addresses. Signing up with a personal address like Gmail has an extra hurdle: Fathom's quick-start guide says you need a Zoom, Meet or Teams meeting on your calendar in the next seven days to finish sign-up. - The free plan is genuinely useful. Fathom's help centre lists unlimited recordings, storage and transcription on Free, with its advanced AI summaries for a limited number of calls each month and a simpler summary after that. New accounts also get a 30-day free preview of the paid Premium features. - Paid plans are per person. Premium, Team and Business are each billed per user, with a discount for paying annually; current prices are on its Launchpad page. Team and Business need at least two users. Fathom's pricing page shows both monthly and annual prices as "/month", so check the Monthly/Annually toggle before you buy. Getting set up safely (Sign up at fathom.ai and download the desktop app from fathom.video/download. Fathom's help centre warns about impostors posing as Fathom, so don't install anything from a link in an unexpected email.): - Turn on consent requests. Fathom has an opt-in setting, Auto Request Recording Consent, that emails guests 24 hours before an external, scheduled meeting — but only in some cases. Its help centre says that with the bot, you must be the host and there must be exactly one guest; in bot-free mode it covers scheduled external meetings, and on Zoom only if you have host rights. Same-day meetings get no email. If a guest declines, auto-record is switched off for that call — but if nobody replies, Fathom still records. - Check who gets the notes. Fathom's sharing guide says that if auto-share is turned on, notes go to everyone on the calendar invite, including people who didn't attend. Leave it off, or limit it, until you know what a recap contains. - Lock down links. On Free and individual plans, recording links default to "Anyone with the link can view". Restrict sensitive recordings one by one, and remember a share can't be unsent, only revoked. - Run a test call. Start a meeting on your own and confirm a recording and summary appear in Fathom afterwards. Security: - Recording a meeting records other people. Fathom's help centre says there's no way to silently record — the bot is visible, Zoom shows a recording banner, and bot-free mode posts a notice in the chat — and it warns that recording without consent can be illegal in some places. Telling people at the start of the call is still on you. Security: What Fathom does with your recordings:: - Fathom's privacy policy (updated August 16, 2026) says its outside AI providers, such as OpenAI, Anthropic and Google, may not train on your data — but Fathom may use de-identified meeting data to train its own models unless you opt out in your settings. - Fathom says it is SOC 2 Type II, GDPR and HIPAA compliant, and stores data in the US. A signed HIPAA agreement is an Enterprise feature. - Deleting a call can't be undone. Automatic retention policies are Enterprise-only. Best at: - Recording and summarising lots of video calls for free, with no cap on recordings. - Turning a long call into short clips you can share with people who weren't there. - Mac users who'd rather not have a bot in the call, via the bot-free mode as it rolls out. Okay at: - Asking questions of your calls on Free. Ask Fathom is limited there; paid plans make it unlimited for your own calls. - Platforms beyond the big three. Fathom's help centre lists Chromebooks, Linux, iPads, webinars and breakout rooms as unsupported. Avoid: - Keeping the default sharing settings for sensitive meetings. - Assuming silence means consent — no reply to a consent email still means Fathom records. Starter: Set sharing before your first call In your settings, check whether auto-share is on and, if it is, limit who receives notes; then switch recording links away from the "Anyone with the link" default. If auto-share is on, Fathom sends notes to everyone on the invite, and on Free and individual plans recording links default to anyone-with-the-link. Checking both first means nothing leaves your account by surprise. Starter: Record a test call Start a Zoom, Meet or Teams call with yourself (or a friend who's agreed), talk for a few minutes, then end it and open the recording in Fathom. It confirms your calendar and platform are connected properly before a real meeting depends on it. Starter: Clip the moment that matters In a finished recording, highlight a key exchange and turn it into a clip to share. Clips are where Fathom saves other people time: a 90-second clip beats sending someone an hour-long recording. Starter: Ask your call a question Try this in Ask Fathom during your 30-day Premium preview. It shows whether the paid chat features are worth it before the preview ends. Prompt: What did we agree on next steps, and who owns each one? Pitfalls: - The Meet link has to be in the right place. Fathom's help centre says Google Meet links must be in the event's Location field, on your primary calendar. - The bot gets stuck in the lobby. On Zoom and Teams you may need to admit Fathom from the waiting room, and Teams meetings with end-to-end encryption can't be recorded. - Free summaries step down after a monthly allowance. The advanced summaries cover only your first few calls each month. - A blocked Gmail sign-up is final. If sign-up with a personal email is blocked, Fathom says you can't make another free account with that email. Where to next: - More meeting & note tools: https://www.launchpadhq.io/discover?category=meetings-notes - Voice & speech tools: https://www.launchpadhq.io/discover?category=voice-speech - Sales tools that use call notes: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai ### Getting Started with Figma AI Canonical: https://www.launchpadhq.io/platform/figma-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/figma-ai.json Last reviewed: 2026-09-09 Tool: Figma AI (https://www.launchpadhq.io/platform/figma-ai): starting price See pricing; free tier yes; data as of not dated AI that acts on the layers in your Figma file — if your plan and your credit balance let it. You work in a Figma file in your browser as normal. Select something on the canvas — a layer, a frame, a block of text — then right-click or open the actions menu and describe what you want in plain words. The change lands on the canvas, where you can undo it, adjust it, and ask again. What it is: - Figma AI isn't one button. It's a set of AI tools spread through Figma's products — inside design files, in FigJam boards, and in Figma Make, which builds a clickable web page from a written description — all drawing on one shared monthly balance of credits (the platform's unit of spend — each thing you make costs some). - Which of them you can reach depends on your plan rather than on how much you use. Figma's help pages put the in-file tools — generating and editing images, rewriting text, renaming layers — on paid plans, while a free account's practical way in is Figma Make inside its own drafts. - There is also a Figma agent (AI that takes actions on its own rather than only answering), which takes instructions in a side panel and changes the canvas itself. Figma's documentation still describes it as an open beta. Before you start: - You need a Figma account and a file open in your browser; there is nothing to install. If you have never used Figma at all, learn to draw a frame and move a layer first. - The honest money answer: a free Starter account does not unlock most of what people mean by "Figma AI". Full use of Figma Make needs a paid Full seat. A free account can still build unlimited Figma Make files in its own drafts, which is where the tour below starts. - Figma gives each Full seat a monthly credit allowance — smaller on Starter, which also has a daily ceiling, and larger on Professional, though by Figma's own estimate even that covers only a few dozen Figma Make prompts. Credits expire at the monthly reset rather than rolling over, and Figma's documentation puts buying more in the admin dashboard, so if you are not the admin you wait. - Switching AI on for a Starter account is a one-way door. Figma's settings documentation says that once AI features are turned on, they are "permanently enabled on your account". Decide before you flip it. Security: - Figma's default here is not the cautious one. Its own settings documentation says content training is turned on for Starter teams and for Professional teams, and off for Organization and Enterprise — so the cheaper the plan, the likelier your work is feeding model training. The switch is real and findable, but only a team or organisation admin can move it. On a solo Starter account that admin is you, which is the good news and the whole responsibility at once. Security: Three things to settle before you point any of this at a client's file:: - Turn content training off first, then design. Figma's AI Terms say that once an admin turns the toggle off, new content and edits added after that date will not be used to train AI models. It is forward-looking wording — it does not reach back. - What you select leaves Figma. Figma publishes a sub-processor list that names outside model companies as recipients of AI feature data — OpenAI for Figma Design, FigJam and Figma Slides; Anthropic, Google and Amazon Bedrock among the providers behind Figma Make. If a contract names which companies may touch the material, that list is what to check it against. - A draft is not a private corner. Figma's settings documentation says drafts organised into a team are treated the same as any other file in that team for training purposes, so "I only did it in a draft" protects less than it sounds like. Best at: - Bulk tidying nobody would do by hand — renaming a few hundred layers called Frame 47 in one pass. - Filling a mockup with plausible content instead of dummy text, so a stakeholder reacts to the design rather than to the placeholder. - Getting a rough idea clickable in Figma Make before anyone spends a day drawing it. Okay at: - Generating imagery. Figma's image documentation says you pick a model from a dropdown rather than using one house model, and it does not tell you which to pick for what — so quality is partly a choice you are making blind. - Wholesale layout generation. The agent works directly on the canvas, but treat what comes back as a sketch to react to. Avoid: - Letting a Figma Make project become the only place a design lives. Figma's documentation says you can copy a Make preview into Figma Design as layers, but that changes to those layers are not applied back to the Make file — so the two quietly drift apart. - Publishing what Make builds without reading it first. Figma puts a published Make project on the public web at its own address, and says plainly that you are responsible for the content that goes up there. - Planning around any of this on a school or government account. Figma's own notes carve those out unevenly — Government plans are excluded from some of the newer tools, and K-12 education accounts cannot use Figma Make at all. Starter: Make an inherited file searchable again When you open a file where every layer is named Frame 47 → then select the lot, right-click, and choose Rename layers so Figma names them from what they contain. (This one is in the paid in-file set.) Figma lists layer renaming among the handful of AI features that cost no credits, and it works on Figma's own layer tree rather than on pixels — Photoshop's generative tools cannot see a Figma layer at all. Rename a small selection first and read the result. Figma names from the content it sees, so a frame full of placeholder text gets named after the placeholder. Starter: Get dummy text out of a mockup When a layout is drawn but every text box says the same filler sentence → then select the frame and use Replace content to drop in plausible names, headlines and prices. Figma's docs describe it filling the text layers in the frame in place, rather than handing you a block of copy to paste. ChatGPT will write better copy and cannot put a word of it in the right layer. Run one duplicate through Adjust tone and keep both, so you can show the same layout reading formal and reading casual in the same review. Starter: Turn a flat picture into something you can recolour When you have a logo or icon that only exists as an image file → then use Vectorize image to convert it into vector layers — shapes, so it stays sharp at any size, unlike a photo, which blurs when you enlarge it. Figma's documentation describes the output landing as editable layers in the file you are already in. Illustrator's Image Trace does the same tracing job and leaves you with a separate document to bring across. Zoom in before you trust it. Tracing invents an edge wherever the picture was soft, and a logo is exactly where a wrong edge shows. Starter: Build a clickable version before you draw one When an argument about how a flow should work has gone three rounds → then open Figma Make, describe the flow in a sentence, and get something people can click through. This is the part of Figma AI a free account can genuinely reach, metered by credits rather than blocked outright. Lovable will build the same kind of clickable page from a sentence, but it does not run in the account where your design system already lives. Ask for one screen, not the whole product. Figma puts a Make prompt somewhere between 30 and 100-plus credits depending on complexity, so a sprawling first request is the quickest way to spend a free month's balance. Pitfalls: - Free today is not free next month. Figma said on 25 August 2026 that the Figma Design agent and the Weave tools would leave free beta "in a few weeks" and start consuming credits like everything else. That window has probably closed by the time you read this, so check what the agent costs before you build a habit around it. - Figma's own documentation disagrees with itself about who gets the agent. One help article says the agent beta is "available on all plans"; the getting-started page and the Config 2026 round-up both say Professional, Organization and Enterprise. Assume the narrower answer and be pleasantly surprised. - The model choice is yours to get wrong. If a result disappoints, change the model before you rewrite the prompt — nothing in the interface will suggest that as the cause. - On a free account, undo runs out. Starter keeps only 30 days of version history, so a file you reshaped with AI four months ago has no earlier state left to compare against. Duplicate a file before a big AI pass. - "Figma AI" is a moving label. Weave is Figma's new name for Weavy, a product it acquired that currently runs at its own address and keeps a credit balance separate from your Figma AI credits. Two balances, two products, one brand — read which one a help page is actually about. Where to next: - AI Plugins & Business Software: https://www.launchpadhq.io/discover?category=ai-plugins-business-software - Image Generation & Editing: https://www.launchpadhq.io/discover?category=image-generation-editing - AI Coding & Development: https://www.launchpadhq.io/discover?category=ai-coding-development ### Getting Started with Fireflies.ai Canonical: https://www.launchpadhq.io/platform/fireflies-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/fireflies-ai.json Last reviewed: 2026-10-23 Tool: Fireflies.ai (https://www.launchpadhq.io/platform/fireflies-ai): starting price Free, from $18; free tier yes; data as of 2026-10-14 An AI note-taker that joins your video calls, transcribes them and sends summaries — once you've told it which meetings to join and who should get the notes. You connect your Google or Outlook calendar. Fireflies' note-taker, Fred, then joins your video meetings as a participant, records and transcribes them, and afterwards sends a summary with action items. You can search transcripts or ask AskFred, its chat assistant, questions about one meeting or many. What it is: - Fireflies.ai is a meeting assistant built around a bot. Its help centre describes Fred as Fireflies' bot, which joins calls as [email address on the web page] on Zoom, Google Meet, Microsoft Teams and other platforms, then produces a transcript, summary and action items. The pricing page lists transcription in more than 100 languages. - If you don't want a bot on the call, there are other ways in. The desktop app can capture your computer's audio with no bot joining — Fireflies says that mode keeps only the transcript and summary, not audio or video. The Chrome extension can record Google Meet calls without a bot joining — audio only, and only for Google Meet in Chrome. The iPhone and Android apps record in-person meetings, and you can upload audio or video files. - After the meeting, AskFred answers questions about what was said — inside one meeting, or across past meetings and connected apps. You can also clip short soundbites of key moments, and send notes automatically to tools like Slack, Notion or a CRM; the pricing page says every plan includes all integrations. Before you start: - Sign-up connects your calendar. You sign in with Google or Microsoft (Outlook) — the only options — and grant calendar access. Fireflies supports one calendar per account. During setup it asks two important questions: which meetings Fred should join, and who should get the recap emails. - Change the defaults before your first meeting. Fireflies' account-setup guide says the defaults are to join every calendar meeting with a video link and to send recaps to everyone on the invite — including outside guests. Most beginners want "Only meetings that I own" or "Only when I invite [email address on the web page]", and recaps sent only to themselves at first. - Free has catches. Fireflies' free plan has limited storage, a length limit per recording and a one-time batch of AI credits. Its help centre says unlimited transcription on Free only applies with auto-join turned on; with auto-join off or set to invite-only, accounts created since September 1, 2026 get only a few meeting transcriptions. Downloading transcripts needs a paid plan. - Paid plans are per seat. Pro is $18/seat/month, with Business and Enterprise above it, and annual billing costs less. Separately, AI credits pay for extra AI features; Fireflies' credits guide says paid plans are automatically subscribed to a small credit pack when they run out. Getting set up safely (Sign up at fireflies.ai and install the desktop app, mobile apps or Chrome extension only from links on fireflies.ai or the official app stores. Fireflies asks for calendar access, so review the permission screen to make sure it's the real thing.): - Set who Fred records. In Settings → Recording & Privacy, choose your auto-join option. "Only when I invite [email address on the web page]" gives you the most control; on the Free plan, Fireflies' help centre says it also caps how many meetings get transcribed. - Set who gets the notes. In the same area, change recap emails from "Everyone on the invite" to just you, or only teammates, until you're sure what a recap contains. - Check sharing. Fireflies' help centre says meeting privacy defaults to "Teammates & Anyone with Link", meaning anyone with a forwarded link can view it. Tighten it if your meetings are sensitive; Fireflies offers to apply the change to past meetings too. Security: - A bot that joins meetings records other people, so consent is the first thing to manage. Fireflies' consent guide says it follows your meeting platform's recording rules, and most meetings are opt-out — the bot joins, and people have to remove it. Recording-consent laws vary by place, so tell people up front. Fireflies has two settings you switch on yourself: an email to attendees an hour before with a link to decline recording, and a message pinned in the meeting chat. Security: Handy controls:: - Type /ff pause or /ff leave in the meeting chat to pause or remove Fred. Fireflies says no transcript is made if Fred is removed within 3 minutes. - Recording Rules can stop Fred joining meetings that match a keyword, email or domain — useful for HR or medical calls. - Only a meeting's owner can delete it, and deletion is permanent. Security: - Fireflies' privacy policy says personal information isn't used to train AI models and that its AI vendors keep no meeting content. It lists SOC 2 Type II and GDPR compliance; HIPAA agreements require Enterprise. Best at: - Recording and summarising lots of video calls without anyone taking notes. - Finding what was said across weeks of meetings by search or by asking AskFred. - Pushing meeting notes into Slack, Notion or a CRM automatically. Okay at: - Accuracy in hard conditions. Fireflies' own guide says background noise, accents, people talking over each other and poor connections all reduce it. - Meetings where a visible bot feels awkward — use the desktop app's no-bot mode instead. Avoid: - Leaving the defaults on: every meeting recorded, recaps to every guest, links viewable by anyone. - Recording people who haven't been told. Starter: Lock down the defaults In Settings → Recording & Privacy, set auto-join to "Only meetings that I own", send recaps only to yourself, and set meeting privacy to teammates only. Fireflies' defaults record every meeting with a video link and email notes to every guest. Starting narrow means nothing goes out that you haven't seen. Starter: Ask about one meeting Open a processed meeting and ask AskFred. Fireflies' help centre recommends specific questions like this, and checking the answer against the transcript before you act on it. Prompt: What were the key decisions in this meeting, and who agreed to do what by when? Starter: Ask across meetings This is one of Fireflies' own example questions for AskFred across meetings (Ctrl/Cmd+J). It's the best way to see why a searchable meeting history is useful. Prompt: List my action items from this week's meetings. Starter: Try it without a bot Install the desktop app and record your next call from your computer's audio instead of inviting Fred. Fireflies says this mode keeps only the transcript and summary, with no audio or video saved — a gentler option for small or sensitive meetings. Starter: Turn on the courtesy notices Enable the pre-meeting email that lets attendees decline recording, and the pinned chat message announcing that the meeting is being recorded. Fireflies' consent guide treats both as settings you switch on yourself. Turning them on makes telling people the default rather than an afterthought. Pitfalls: - Fred turning up uninvited. If it joins a meeting it shouldn't, check your auto-join setting — and remember it follows your calendar, not your intentions. - Outsiders getting your recaps. Recap emails go to everyone on the invite by default, including people without a Fireflies account. - Running out of Free storage. The free plan's storage cap doesn't reset monthly; you'll need to delete old meetings to open new ones. - Surprise AI-credit charges. Fireflies says its Daily Digest and Meeting Prep features are on by default and use AI credits; turn them off if you don't use them. Where to next: - More meeting & note tools: https://www.launchpadhq.io/discover?category=meetings-notes - Voice & speech tools: https://www.launchpadhq.io/discover?category=voice-speech - Automate what happens after the call: https://www.launchpadhq.io/discover?category=workflow-automation ### Getting Started with Fireworks AI Canonical: https://www.launchpadhq.io/platform/fireworks-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/fireworks-ai.json Last reviewed: 2026-09-08 Tool: Fireworks AI (https://www.launchpadhq.io/platform/fireworks-ai): starting price See pricing; free tier yes; data as of not dated A hundred-plus open AI models behind one login — try them in a browser chat box, then wire the one you like into software you're building. You open Fireworks in a browser, pick an AI model (the AI "brain" that does the actual thinking) from a long list, and type into a chat box to see what comes back. Swap to a different model and send the same message again. When you want it inside your own software, the same models answer from code. What it is: - Fireworks is a hosting company, not a model maker. Labs like DeepSeek, Alibaba, Meta and Mistral publish their AI models as files anyone may download and run; Fireworks runs them on its own machines and rents you access. That is why one login reaches a hundred-plus models rather than one company's. - It is built for people putting AI inside something they are making. The browser chat box is the right place to start, but there is no phone app, no saved conversations, no sidebar of documents; the paid layers rent dedicated hardware or tune a model on your own examples. If you want an assistant to talk to daily, a consumer chatbot is the better door. Before you start: - Understand the money before anything else. Fireworks runs on prepaid credits (the platform's unit of spend — each thing you make costs some), not a subscription. A new account currently gets a small amount of free credit — a one-off welcome grant, not an allowance that refills. The documentation is explicit: when it runs out with no card on file, the account is suspended until you add one. - It goes further than it sounds. The cheapest text models currently list at cents per million chunks of text handled (a chunk is roughly a short word); only the flagships cost dollars. An afternoon in the browser spends a fraction of the grant. - Until a card is on file the account is capped at roughly ten requests a minute — ample for typing questions by hand, far too little to run anything real. - Nothing to install. Sign in with Google, GitHub, LinkedIn or an email address; Fireworks' own onboarding puts experimenting in the model playground before anything else, so the model list and chat box are where to head first. Security: - Fireworks' stated default is unusually strong for a hosted service. Its documentation says that for open models it does not log or store what you type or what comes back at all without an explicit opt-in — the data lives only in working memory for the length of the request. Fireworks also lists SOC 2 Type II certification and says it supports HIPAA workloads. Security: Two things that qualify it: - One documented exception, and it is opt-out rather than opt-in: a newer interface called the Response API saves conversations by default so a chat can carry across calls, deleting them after 30 days. - "Not stored" describes storage, not travel. What you type still crosses to Fireworks' machines and runs through a model another company built. Best at: - Breadth — DeepSeek, Qwen, GLM, Kimi, MiniMax, Llama, Gemma and Mistral behind one login and one balance, so comparing them is a dropdown rather than four signups. - Growing past the trial without moving house: the same account rents dedicated hardware and tunes models on examples you supply. Okay at: - Being explored by someone who does not write code. The chat box is the way in, though the quickstart makes a key before anything else, so you may find you need one; everything around it is written for developers. Avoid: - Assuming Fireworks' terms are the only ones that apply. Its agreement requires you to obey each model's own licence, written by the lab that published it — so whether you may use a model commercially is a question about that model, not about Fireworks. - Passing the results off as your own writing. The terms specifically forbid representing output as human-generated, and let Fireworks tell you at any time to stop using an output it believes may infringe someone's rights. Starter: Ask the same question of a cheap model and an expensive one, and read both price rows Every Fireworks model page prints three prices, not one — text going in, text coming back, and a much lower third rate for input the system has already seen. Read those beside a model priced at a small fraction of them and you learn both things that matter here: whether the cheap model was good enough, and why repeating a long standing instruction costs far less than the headline rate. Use a question with a right answer you can check. Open-ended writing makes every model look fine. Prompt: How many of each? Show your working. Starter: Try a model you cannot get in a consumer app Picks DeepSeek, Kimi, GLM or MiniMax from the model list in the browser chat box and asks it the question you used in the first card. DeepSeek, Kimi, GLM and MiniMax are among the strongest openly published models anywhere, and almost none reach Western consumer chatbots. Fireworks serves them from its own machines under the no-logging default above, rather than forwarding your request to whichever host is cheapest that minute — so trying them does not also mean choosing where your text lands. Starter: Only if you write code: make a key and change two lines Creates an API key in your account, then points an existing script at Fireworks' address with that key in place of the one it had. The quickstart lists three ways to call the service — Fireworks' own toolkit, OpenAI's and Anthropic's — so a script already written against either of those usually needs a changed address and key, not a rewrite. If you do not write code, stopping after step two is a legitimate ending. An API key is a password that identifies your app, and that spends your money. Make a separate one per project. Pitfalls: - Adding credits does not always restart a stopped account. A separate monthly spend limit can pause one that still holds a balance; the fix is raising that cap in billing settings, not buying more. - Auto Reload is a standing instruction to charge your card whenever the balance dips. Switch it on deliberately. - The pricing page currently flags a rate change on dedicated hardware. Check any figure here against it before relying on one. Where to next: - More AI APIs and developer services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - Running open models on your own computer: https://www.launchpadhq.io/discover?category=local-open-source-ai - If you just want an assistant to talk to: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with FLUX Canonical: https://www.launchpadhq.io/platform/flux/guide Data: https://www.launchpadhq.io/data/v1/guides/flux.json Last reviewed: 2026-10-14 Tool: FLUX (https://www.launchpadhq.io/platform/flux): starting price ~$0.048/image (768×768, 1 image). Open-weights licensing (Builder/Platform) is separate and custom.; free tier yes; data as of 2026-10-14 The image model that finally gets text right — but you'll reach it through someone else's front door. FLUX is a model — the AI "brain" that does the actual work — not a website you log into. Whichever route you reach it through, the motion is the same: describe the picture you want, look at what comes back, change the wording, go again. What it is: - An image and editing model family from Black Forest Labs, the German lab founded by the researchers behind Stable Diffusion. The current line is FLUX.2: [max], [pro], [flex], [dev] and [klein]. Two are open-weight — the model file itself is published, so you can run it yourself, which no closed rival allows. The rest sit behind a paid API, a way for programs to talk to each other without a person clicking. - Its pitch against Midjourney: legible text inside images, and tight instruction-following. BFL says typography, infographics and UI mockups "now work reliably", and that it holds a character consistent across up to ten reference images. Before you start: - Going direct to Black Forest Labs' dashboard at bfl.ai is pay-as-you-go, no subscription. You pay per image, at a few cents each, and per-model rates vary by resolution; the calculator on bfl.ai/pricing shows the current rate for the model and size you pick. - Most beginners should come in through someone else's app. Krea and Freepik are ordinary image-making websites — sign up, type, get pictures — that license FLUX and run it for you. Both have a free tier, you pay them rather than BFL, and their terms set your rights. - Running the files yourself is free, and the barrier is hardware, not skill: the freely licensed FLUX.2 [klein] 4B wants a graphics card with roughly 13GB of memory on it — most laptops don't have that — plus software like ComfyUI to drive it. Below that bar, a third-party app is the same model for cents an image. Security: - Privacy differs sharply by route. [klein] on your own machine is genuinely private — no server is involved. Every other route sends what you type to someone's computer, BFL's or a third-party app's, which may pass it on. Open source is not the same as private. Security: Free to download is not free to sell from: - Images made through BFL's paid API are yours: their Developer Terms say you own the Output and may use it commercially. - FLUX.2 [klein] 4B is Apache 2.0 — free, and free to sell from. The 9B is under the FLUX Non-Commercial License: free to download, not for commercial use. - FLUX.2 [dev], the large open-weight model, is non-commercial by default; selling needs a paid self-hosted licence. - For FLUX these permissions genuinely come apart — check your variant before anything reaches a client invoice. Best at: - Readable text inside an image — signage, packaging, poster headlines - Long, multi-part instructions, without quietly dropping half - Editing by instruction: change the jacket, leave the rest alone Okay at: - Painterly art — Midjourney is prettier unasked - Photorealistic faces: good, occasionally uncanny - Fast iteration on the small variants, trading fidelity for speed Avoid: - Expecting a polished consumer app with a gallery — BFL doesn't make one - Body text: headlines hold, paragraphs degrade into plausible nonsense - Anything commercial on a non-commercially-licensed variant — a legal limit, not a quality one Starter: Test the text-rendering claim first Legible text is the capability BFL leads with for FLUX.2, so a failure here tells you about your route immediately. Prompt: A weathered enamel shop sign on a brick wall, cream lettering on deep green, reading exactly: HARBOUR ROAD BAKERY - est. 1974. Soft overcast daylight. Starter: Stack five requirements and count what survives FLUX.2 is tuned for adherence to complex structured instructions, so counting how many hold is a fair test, not a gotcha. Prompt: A ceramicist's workbench seen from directly above. Exactly three unfired bowls in the upper left. A blue cloth over the right edge. Warm light from the left. No hands. Starter: Edit a picture with words instead of tools FLUX.2 generates and edits in one model at up to 4MP, and change one thing, touch nothing else is the job it was built for. Prompt: Keep this photograph as it is, but change the parked car from silver to matte dark green. Change nothing else — not the background, the lighting, the reflections. Pitfalls: - It is not a website you log into. There is no consumer FLUX app the way there is a Midjourney — BFL runs a developer dashboard with a playground, billed against prepaid credit. A free browser tool with a gallery means a third-party product with FLUX inside. - Confusing the generations. FLUX.1, FLUX.2 and FLUX 3 are different things; FLUX 3 is a combined image, video and audio model, not a successor to the image line. Check the release notes when what you read doesn't match what you see. Where to next: - Image tools you can just sign up for: https://www.launchpadhq.io/discover?category=image-generation-editing - Running open models yourself: https://www.launchpadhq.io/discover?category=local-open-source-ai ### Getting Started with Gamma Canonical: https://www.launchpadhq.io/platform/gamma/guide Data: https://www.launchpadhq.io/data/v1/guides/gamma.json Last reviewed: 2026-10-15 Tool: Gamma (https://www.launchpadhq.io/platform/gamma): starting price Free, from $12; free tier yes; data as of 2026-10-17 Describe a presentation, document or simple website in a sentence, and Gamma designs the whole thing — then you edit it like any slide deck. You type a request into a box on your dashboard — a sentence is enough. Gamma's agent (AI that takes actions on its own rather than only answering) asks a few questions, shows you a plan, then builds finished, designed slides. You change them by chatting with it, or by clicking in and editing yourself. What it is: - Gamma builds finished-looking presentations, documents, one-page websites and social posts from a description. You type what you need — "a ten-slide talk for new volunteers" — and it writes the text, picks a layout and adds images, so your first draft already looks designed rather than being a blank template you fill in. - Since its October 2026 update, called Gamma 5, the starting point is a conversation. In its default Plan mode the agent asks about your goal, audience and priorities, then shows a written plan for the story and the look, and builds nothing until you approve it. An Express mode skips the questions and goes straight to a draft. - It does not need you to start from nothing. You can paste in your own notes, or import a PowerPoint, Word file, PDF, Google Doc or Google Slides deck, or a web page's address, and Gamma restyles that material rather than inventing new content. - The writing and images come from AI made by other companies — Gamma names Claude, OpenAI, Gemini, Flux, Ideogram and Luma — rather than its own. When you are done you can present straight from Gamma, share a link, or export a file to PDF, PowerPoint or an image, or bring it into Google Slides. Before you start: - You can start free in a browser at gamma.app. There are also iPhone and Android apps. The free plan gives you a one-time batch of credits (the platform's unit of spend — each thing you make costs some) at signup. Gamma's help centre is explicit that free credits "do not refresh", so treat them as a trial budget rather than a monthly allowance. - Credits are spent only on AI work: generating a deck, adding slides or images with AI, asking the agent to change something, trying new layouts. Typing, moving things around and fixing a slide by hand cost nothing, and Gamma says it does not charge for AI errors or for the agent's clarifying questions. Costs vary with the length of what you ask for and, for images, with which image generator you pick — Gamma shows the cost before an image is made. On the free plan a single request can produce only a limited number of slides. - Anything you export on the free plan carries a badge. Gamma's help centre says a "Made by Gamma" badge is stamped on PDF and PowerPoint exports whenever your workspace is on the free plan at the moment you export, and a "Made with Gamma" badge shows on decks you share as links. A paid plan removes it from future exports only — files you already exported keep it. - Whether you will need to pay depends on whether that badge or the one-time credit budget gets in your way. Paid plans are billed per person: Plus, Pro and Ultra each add a monthly credit allowance that grows with the tier, and paying yearly is cheaper. Security: - The thing to manage on Gamma is who can see what you make, because its finished work is built to be shared as a link. A public link lets anyone who has it open your gamma without a Gamma account, and you set what they can do with it — view, comment or edit. Gamma's own terms add that anything you put in a 'public' area of the service may be visible to other Gamma users. Security: Before you share or upload:: - Check the link setting every time you share. Choose View unless you actually want strangers editing, and remember that forwarding a link forwards access. - Password protection and keeping a gamma out of search engines are not on every plan; without them, a shared link is the only lock on it. - Gamma 5 can pull material in from connected tools such as Slack threads and call transcripts. Connect only what you would be comfortable seeing on a slide, because that is where it can end up. Security: - Gamma's help centre says that on every individual plan, including Free, its default setting lets Gamma use what you make to train its AI. You can switch that off under your avatar → Settings → Data Controls — do it before you upload anything you would not want reused. Best at: - Getting from a topic, or a pile of notes, to a designed first draft in a few minutes — with a plan you can correct before it spends credits building. - Restyling the same content: ask for a new look and Gamma keeps your words while changing the design. - Reusing one piece of work as a deck, a document or a one-page website, since all three live in the same editor. Okay at: - Producing a deck you will finish in PowerPoint or Google Slides. Exports work and keep editable tables, but Google Slides swaps Gamma's fonts for its own, slide masters are left empty, and gradients and some effects come out flatter. - Being accurate. Gamma's terms describe what it generates as initial drafts, 'not intended as any final, definitive or comprehensive presentation' — read every slide before it goes in front of anyone. Avoid: - Handing over a free-plan export as finished work, unless you are happy for it to carry Gamma's badge. - Spending your free credits regenerating a whole deck to fix one slide. Edit by hand, or point the agent at that slide; the one-time budget does not come back. Starter: Let the plan do the work Leave Gamma in its default Plan mode for this. It will ask a few questions and show you a plan before it builds, and Gamma's own guidance calls adjusting that plan the highest-leverage step — fixing the story there is free, while fixing it after the build spends credits. Ten slides is also the most one request can produce on the free plan. Answer any question you don't care about with 'Decide for me'. Switch to Express only once you know what you want. Prompt: Make a 10-slide presentation introducing our community garden to new volunteers: what we grow, when we meet, what a first shift looks like, and who to contact. The audience is mostly retirees and families, so keep it friendly and light on text. Starter: Bring your own words Gamma's help centre recommends exactly this instruction when your content is final: it builds the design around your text instead of rewriting it. That keeps the result accurate to what you meant, which matters on a tool whose own terms call its writing a first draft. Have the notes in a Word file, PDF or old PowerPoint? Import the file instead of pasting. Prompt: Use this text as-is — don't rewrite it. Turn it into a short presentation, one idea per slide: [paste your notes] Starter: Change the look, not the words Gamma treats restyling and regenerating as two different things — restyling keeps your content, regenerating starts the content over. Asking for the first is how you try a new look without losing edits you have already made, or paying for a fresh draft you didn't want. Prompt: Restyle this deck in a dark, moody look with bold headings. Keep all of my content exactly as it is. Starter: Fix one slide, not the deck The agent remembers your original goal and shows what it changed, so a narrow request like this gets a targeted fix. Gamma also lets you click a single element in AI-editing mode to limit the change to just that — the narrower the request, the fewer credits you spend undoing side effects. Prompt: Slide 4 is too text-heavy. Cut it to three short points and add one image that fits. Starter: Reuse it as a page Decks, documents and websites are all made in the same Gamma editor, so the content you already approved can become a shareable web page without starting over in a separate website builder. Prompt: Turn this presentation into a one-page website with a short intro at the top and a sign-up section at the end. Pitfalls: - Exporting on the free plan stamps a badge. Gamma's help centre notes one exception: printing from your browser while signed in to your own gamma does not add it. - Taking a PowerPoint export at face value. Open the exported file before you present from it; swapped fonts can change text sizes and push words off a slide. - Regenerating when you meant to restyle. Asking for the wrong one loses your edits and spends credits. - Forgetting a paid plan renews itself. Gamma's terms say subscriptions renew automatically, there are no refunds, and they ask you to cancel at least 30 days before the renewal date — so cancel well ahead, not on the last day. Where to next: - Image & design tools: https://www.launchpadhq.io/discover?category=image-generation-editing - Chat assistants for drafting the content first: https://www.launchpadhq.io/discover?category=text-conversational-ai - Document & PDF tools: https://www.launchpadhq.io/discover?category=document-pdf-processing ### Getting Started with Gemini Canonical: https://www.launchpadhq.io/platform/gemini/guide Data: https://www.launchpadhq.io/data/v1/guides/gemini.json Last reviewed: 2026-10-14 Tool: Gemini (https://www.launchpadhq.io/platform/gemini): starting price Free, from $4.99/month; free tier yes; data as of 2026-10-26 Google's assistant, free and genuinely capable — with a privacy model built on human review rather than a training switch. You open Gemini in a browser or its app, signed in with your Google account, and ask in plain language. It answers, and you keep going — attach a file, ask a follow-up, ask it to try again — in one running conversation. The same assistant also appears inside Google's other products, working on whatever you have open there. What it is: - A general-purpose assistant with an unusually capable free tier — chat, image generation and editing, deep research, and live voice conversation at no cost. - Tied into the Google account you already have, with paid tiers bundling cloud storage and other Google subscriptions alongside higher usage. - Strong on very long inputs, with paid tiers accepting document sets far larger than most assistants handle. - Governed by Google's activity and human-review model rather than by a single train-or-don't switch. Before you start: - The free tier needs nothing but a Google account — no card, no trial clock. It includes chat, image generation and editing, deep research, live voice, and the notebook research tool. - Paid tiers are now named Google AI Plus (2× Free usage), Google AI Pro (4×), and Google AI Ultra, the top tier; current prices are on its Launchpad page. Do not use the retired “Gemini Advanced” name. - Limits are compute-based rather than message-counted, refresh on a five-hour cycle against a weekly ceiling, and can be extended by buying credits. - Decide what you are willing to have a human read before you start typing. See the security section — this is the part that differs most from its competitors. Security: - Gemini's privacy model is not a training toggle. Google's stated position is that a portion of conversations are read and annotated by human reviewers, including reviewers at service providers, to improve its models and services. Google's own guidance is unusually direct about the implication: do not enter anything confidential that you would not want a reviewer to see. Security: What deleting does and does not do: - Conversations selected for human review are kept separately from your account and retained for up to three years. - Deleting your activity does not remove them. Deletion clears your history; it does not reach into the review pipeline. - Turning the activity setting off still leaves a floor — conversations are held for roughly 72 hours to operate the service. - Temporary conversations are excluded from being used to improve Google's AI. - Custom assistants you configure carry the same exposure as ordinary conversations, including any files you attach to them. Security: - Set against a tool where one switch turns training off, this is a different shape of risk rather than a larger or smaller one. There is no configuration that gets you to zero retention, and the meaningful control is not a setting — it is what you choose to type. Use temporary conversations for anything sensitive, and treat the sensitive material itself as the thing to withhold. Best at: - Very large document sets, where its context capacity outclasses most assistants - Multi-source research reports assembled in one pass - Work that touches Gmail, Docs, or Drive, where it is already inside the tools - Image generation and editing without leaving the conversation Okay at: - Hard multi-step reasoning, where it is competent but not the category leader - Consistency across long sessions, which varies more than its rivals - Code, which it handles but is not the reason to pick it Avoid: - Confidential, regulated, or client material on a personal account - Any workflow where you need a guarantee that deleting means deleted - Treating its answers about its own settings as current, since the product's interface changes frequently Starter: Hand it more material than feels reasonable Its capacity for large inputs is the clearest advantage it has, and almost nobody uses it. Ask which document is the outlier and why — that surfaces the assumption the set is built on. Prompt: I'm attaching several documents. Read all of them, then tell me where they disagree with each other. I care more about the contradictions than the summary. Starter: Run a deep research report before you commit to a direction Searches across many sources and returns a structured report rather than a chat answer. Available on the free tier, and it is a different tool from ordinary chat — worth learning as its own thing. Give it the decision you are trying to make, not the topic. "Should I do X or Y, and what would change my mind" beats "tell me about X". Starter: Use a temporary conversation for anything sensitive The one concrete control that meaningfully reduces exposure, and it is off the beaten path. Make it a habit tied to content, not to mood — anything with a name, a number, or an employer in it. Prompt: [Start a temporary conversation, then ask your question as normal.] Starter: Ask it to show its sources and then check one Builds the verification reflex early, while the stakes are low. Pick the claim that most surprised you and check that one. Surprise is a good relevance filter. Prompt: Answer the question, then list the specific sources you used and which part of the answer each one supports. Pitfalls: - Deleting your activity does not delete everything. Conversations that went to human review are kept separately for up to three years and survive your deletion. Deletion manages your history, not Google's review pipeline. - There is no zero-retention setting. Even with activity off, conversations persist for around 72 hours. The real control is what you type, not what you toggle. - The free tier's model is not the paid tier's model. Free access to the top model is described as varying rather than guaranteed, so a result you got yesterday may not reproduce today. - Third-party pricing articles are frequently wrong. Plan prices were restructured in 2026 and stale figures are widespread online. Check the official subscriptions page rather than a comparison post. - Interface details move faster than any guide. Setting names and menu locations change often. If a setting is not where you expect, it was probably renamed rather than removed. Where to next: - Other chat assistants: https://www.launchpadhq.io/discover?category=text-conversational-ai - Research and academic tools: https://www.launchpadhq.io/discover?category=research-academic-tools ### Getting Started with Gemini CLI Canonical: https://www.launchpadhq.io/platform/gemini-cli/guide Data: https://www.launchpadhq.io/data/v1/guides/gemini-cli.json Last reviewed: 2026-09-08 Tool: Gemini CLI (https://www.launchpadhq.io/platform/gemini-cli): starting price Free; free tier yes; data as of not dated Google's terminal coding agent — still shipping, but since June 2026 it no longer opens to a free Google account. You open a text window on your computer, start it inside a project folder, and type what you want in plain English. It reads the files itself, proposes edits, runs commands to check its work, and asks before changing anything. You read what it did and say what to fix. What it is: - Gemini CLI is Google's coding agent — AI that takes actions on its own rather than only answering — driven from a terminal, the text window where you type instructions instead of clicking. It works on a whole project at once, not one file. - Since 18 June 2026 it no longer serves free personal Google accounts, or Google AI Pro and Ultra subscribers. Google moved individuals to a separate tool, Antigravity CLI, which is where a beginner should now start. - The software is alive: the code is public and releases ship most weeks. It just answers only to an organisation's Code Assist licence or a paid Gemini API key — a password that identifies your app and spends your money. Before you start: - Settle access before you install. A personal Google account no longer works; the two live routes are a Gemini Code Assist Standard or Enterprise licence, bought per seat by an organisation, or a paid Gemini API key. Without one it installs fine and then refuses to answer. - There is no monthly price for an individual — a paid key bills you for what you use, so a heavy week has no ceiling until you set a budget yourself. - Expect Google's documentation to lag Google's product: the quota page still lists a free 1,000-requests-a-day tier, and the sign-in page still recommends a personal account. The banner above them is current; the pages beneath are not. - You need code on your own disk and enough comfort with a terminal to install a package. If not yet, Google's Gemini assistant in a browser writes and explains code with nothing to install. Getting set up safely (Google's official Gemini CLI documentation): - Install from Google's own listing only — npx @google/gemini-cli, a global npm install -g @google/gemini-cli, or brew install gemini-cli. Lookalikes exist and third-party instructions run a version behind. - Keep it to one project folder whose changes are tracked. It edits real files and runs real programs. Security: - Nothing runs on your own hardware. Every request leaves your machine for Google's models, so a codebase you are barred from sharing is barred from this too. Security: Which door you came in decides your data terms: - On a paid Gemini API key, Google's terms say it does not use your prompts or responses to improve its products. On the unpaid tier the same terms say the opposite, and that human reviewers may read them. The tool never tells you which you are on. - It also sends usage statistics by default — which tools ran, which model, how long, though Google states not the content of your prompts. One setting turns that off: privacy.usageStatisticsEnabled: false in ~/.gemini/settings.json. Best at: - Reading a project you did not write and explaining how it fits together - Changes that touch several files and must stay consistent - Working from a picture — app-from-an-image-or-PDF is Google's own headline claim Okay at: - Costing what you expect, since the bill tracks how long a session stays open, not the size of the task - Being learnable from what is written about it, since most guides predate the June change Avoid: - Its own setup pages, which still recommend the sign-in Google switched off in June - Depending on a free Google developer tier; this one ended a month after it was announced Starter: Work out which door is open to you, before you install Checks whether you hold a Gemini Code Assist Standard or Enterprise licence, or a Gemini API key with billing switched on; if neither, installs Antigravity CLI instead. Claude Code's entry check is one question — is your plan paid. Here the sign-in Google's own docs recommend is still there and still fails, so the only check that holds is what you own: a Code Assist licence, or a key with billing on. If it is neither, Antigravity CLI is where Google now sends you. Starter: Hand it a picture instead of a description The @ prefix pastes a file straight into the prompt with no upload step, and app-from-an-image is the line Google leads with — not where Claude Code or Cursor put their emphasis. Prompt: @mockup.png Build a single plain HTML and CSS page matching this layout, then list the three things you had to guess. Starter: Write a GEMINI.md in the project root Context files load in a hierarchy — home folder, then workspace, then more as it touches directories — so a project file applies every session, and the filename setting takes a list, so a conventions file you already have can be reused. Keep it short. Every session pays for it whether it needs it or not. Prompt: Read this project and draft a GEMINI.md: what it is, how to run its tests, the conventions you can infer, and three things never to change without asking me. Pitfalls: - Nearly every Gemini CLI guide online opens with the free personal-account login it no longer has. - A Google AI Pro or Ultra subscription does not cover it — it did until 18 June 2026, and the CLI simply stopped answering it. Where to next: - Other AI coding tools: https://www.launchpadhq.io/discover?category=ai-coding-development - Google's assistant in a browser: https://www.launchpadhq.io/discover?category=text-conversational-ai - Build with an API directly: https://www.launchpadhq.io/discover?category=ai-apis-developer-services ### Getting Started with Genspark Canonical: https://www.launchpadhq.io/platform/genspark/guide Data: https://www.launchpadhq.io/data/v1/guides/genspark.json Last reviewed: 2026-10-23 Tool: Genspark (https://www.launchpadhq.io/platform/genspark): starting price See pricing; free tier yes; data as of 2026-10-14 An all-in-one AI workspace: ask its Super Agent for a finished deck, spreadsheet, document or podcast, and it plans the steps and builds it for you. You describe what you want in one message. Genspark's Super Agent breaks the job into steps, picks the right tools — web research, a slide maker, a spreadsheet builder — and works in its own cloud computer until it hands back a finished file. Each action spends credits, Genspark's unit of usage. What it is: - Genspark is an AI workspace built around agents — AI that carries out tasks rather than only chatting. Its help centre describes the Super Agent as an autonomous assistant that thinks, plans and acts: it splits a request into steps, chooses tools, and delivers a finished result. It can browse the web, read files you upload and work with accounts you connect. - Alongside it are specialist agents for particular outputs. AI Slides builds presentations you can export to PDF, PowerPoint or Google Slides. AI Sheets gathers and analyses data in spreadsheets. AI Docs writes documents you can export to Word, PDF or Markdown. AI Pods turns a topic, document or link into a podcast episode. There is also a plain AI Chat, which Genspark says offers more than 40 models and web search with sources. - Genspark doesn't rely on one AI company. Its privacy policy names OpenAI, Anthropic, Google, xAI and ElevenLabs as its main AI providers, and its help centre says the standard slide mode runs on Genspark's own model. It runs in a web browser, on iPhone and Android, and as a free desktop app for Mac (Apple Silicon), Windows and Linux. Before you start: - Choose your sign-in carefully. You can sign in with Google, Apple or email. Genspark's help centre says the sign-in method can't be changed later, the account email can't be changed, and signing in a different way creates a separate, empty account that support can't merge. - Free is a small daily allowance. Genspark's membership page lists the Free plan as a daily credit refill that stops once you reach a lifetime limit. It doesn't say how big that lifetime limit is. Free users can present and share slides but need a paid plan to export them. - Plus is the usual paid step. Plus is sold monthly or, for less per month, yearly, and comes with a far larger monthly credit allowance and commercial use of what you make. Pro is a much more expensive tier above it. Credits reset each billing period and don't carry over. - Refund rules are strict. Genspark's help centre says refunds are possible within 72 hours of a charge for most customers (longer in the EU, UK, Turkey and Korea), only if you've used less than half the subscription, and only once per customer. Security: - Genspark sends your prompts to several outside AI companies, so treat it like any cloud service: no passwords, ID numbers or confidential client data in prompts or uploads. Its privacy policy says data is stored on Microsoft Azure and deleted within 30 days of closing your account. Neither its privacy policy nor its terms says plainly whether your content is used to train Genspark's own models, and its terms grant it a licence to use your content to operate and improve the service. Security: Settings and features to handle with care:: - Share links. In Slides, Sheets, Docs and Pods you can share by email invite or with "Anyone with the link", which is public. A slides "Project share" also shows your whole creation process, not just the finished deck. - Connectors. Connecting an app gives Genspark real access — its help centre says connecting Slack grants read and search access to channels, direct messages and files. Connect only what a task needs. - The Chrome extension. It can click, fill in forms and extract data on pages you're using. Genspark advises testing automation on unimportant pages first and confirming before anything is ordered or sent. Security: - Genspark's terms say you can't rely on generated content and ask you to credit the source when republishing it. Check facts and figures before a deck or report goes to anyone else. Best at: - Going from a one-line request to a finished, designed slide deck, document or spreadsheet. - Combining web research with a deliverable in one go, without stitching tools together yourself. - Trying many AI models in one place through AI Chat. Okay at: - Cost control. The Super Agent does many steps on its own, and Genspark's credits guide says regenerating, retrying and long chats all add up. - Accuracy. Genspark's own docs recommend human review, and AI Slides includes a fact-check option for that reason. Avoid: - Making a project link public when it contains anything private. - Burning a small free allowance on repeated regenerations. Starter: Start simple in AI Chat This is one of Genspark's own examples. Genspark's help pages suggest working out what you want in AI Chat first, then handing a clear request to Super Agent so it doesn't spend credits on back-and-forth. Prompt: Explain the differences between Web3 and traditional web applications in simple terms. Starter: Research into a deck Another of Genspark's own examples, and a good test of what the Super Agent adds: it researches, writes and designs in one task. Open the sources it cites before you present. Name your audience — "for a high-school economics class" — to change the depth and tone. Prompt: Research electric vehicle market trends and create a 10-slide presentation. Starter: Turn reading into listening AI Pods is one of Genspark's more distinctive tools, and this is its own example. Genspark says an episode usually takes 5 to 15 minutes to make. Prompt: Give me a daily news briefing podcast on today's tech headlines. Starter: Summarise the page you're on Use this in the Chrome extension on a long article. It's a low-risk way to try the extension, which only reads a page when you actively use it. Prompt: Summarize the key points of this page. Pitfalls: - Regenerating costs the full price again. Genspark's credits guide says each regeneration costs the same as the original; edit instead where you can. - Ultra mode is much more expensive. Genspark's help centre puts an Ultra slide at roughly ten times the credits of a standard one. - Two subscriptions by accident. Subscriptions bought in the iPhone or Android app are managed and cancelled in the app store, not on the website — and deleting your account doesn't cancel them. - Credits that don't last. Plan credits reset every billing period; Genspark says separately purchased top-ups last three months. - Signing in the wrong way. Using a different Google account or method later creates a new, empty account rather than opening your existing one — and support can't merge them. Where to next: - More workflow & automation tools: https://www.launchpadhq.io/discover?category=workflow-automation - Document & PDF tools: https://www.launchpadhq.io/discover?category=document-pdf-processing - Research & academic tools: https://www.launchpadhq.io/discover?category=research-academic-tools ### Getting Started with GitHub Copilot Canonical: https://www.launchpadhq.io/platform/github-copilot/guide Data: https://www.launchpadhq.io/data/v1/guides/github-copilot.json Last reviewed: 2026-10-14 Tool: GitHub Copilot (https://www.launchpadhq.io/platform/github-copilot): starting price Free, from $10/mo; free tier yes; data as of 2026-10-25 An AI pair-programmer that lives inside your code editor, suggesting lines and answering questions as you work. Copilot watches the code around your cursor and suggests what probably comes next — a line, a function, a block — which you accept with a keystroke or ignore by carrying on typing. Alongside that, you can ask it things in plain English in a chat panel and get answers grounded in the code you have open. What it is: - GitHub Copilot is an AI coding assistant that runs inside your editor — it autocompletes code as you type, and answers questions in a chat panel about the file or project you're in. It's built by GitHub with models from several AI labs, and it works in the popular editors (VS Code, Visual Studio, the JetBrains IDEs, and Vim/Neovim) rather than being its own separate app. It's aimed at anyone who writes code, from first-project beginners to professionals; you don't need to be an expert to get value from the autocomplete. Before you start: - There's a Free plan (limited chat and agent usage; GitHub no longer publishes a single completion-count we could confirm on 2026-10-14). Paid individual plans step up through Pro, Pro+ and Max, plus a GitHub AI Credits layer; current prices are on its Launchpad page. Verified students and popular open-source maintainers can still get paid features free. - Install it from the official source only — the Copilot extension inside your editor's built-in marketplace, or via github.com. Don't install a "Copilot" plugin from an unofficial third-party listing. - You need a GitHub account and one of the supported editors. If you don't already have an editor, VS Code is free and the most common starting point. Security: - Copilot sends the code around your cursor, plus your chat messages, to GitHub's servers to generate suggestions. On a work codebase, check your employer's policy before turning it on. Security: The data-training default worth knowing: - On the individual plans (Free, Pro, Pro+), GitHub may use your interactions — including your inputs, the suggestions you get, code snippets, and surrounding context — to train and improve its models unless you opt out in settings. - The Business and Enterprise plans are excluded from that training by default. - If you're on an individual plan and don't want your code used for training, the opt-out is a settings toggle — flip it before you start on anything sensitive. Security: - Suggestions can occasionally mirror public code, and generated code can carry bugs or security flaws. Treat every suggestion as a draft to review, not a finished answer — you're still the author on the hook for what ships. (The paid Business/Enterprise tiers add IP indemnity; the individual tiers don't.) Best at: - Autocompleting boilerplate and repetitive code you'd otherwise type by hand - Explaining unfamiliar code in plain language - Drafting tests, comments, and small functions from a description - Keeping you in flow without tab-switching to a separate chatbot Okay at: - Larger multi-file changes via agent mode (workable, but review closely) - Niche or less-common languages (quality tracks how much public code exists) - Debugging — it helps, but can confidently suggest wrong fixes Avoid: - Trusting generated code without reading it - Security-critical or licensing-sensitive code without human review - Treating it as a substitute for understanding what your code does Starter: Pick your editor and install the extension Install Copilot from the official marketplace inside VS Code (or your JetBrains IDE / Visual Studio), then sign in with your GitHub account. Everything else depends on this; the official in-editor marketplace is the safe source. No editor yet? Install VS Code first — it's free. Starter: Start on the free tier Choose Copilot Free to try autocomplete and limited chat with no card on file before deciding on a paid plan. You can judge whether it fits your workflow at zero cost and zero commitment. Student or OSS maintainer? Check eligibility for free paid-tier access. Starter: Set your training preference In Copilot settings, decide whether GitHub may use your interactions to train its models, and toggle the opt-out if you'd rather it didn't. On individual plans this is on by default; setting it deliberately before you code is the one thing not to skip. Starter: Try it on real code Uses both halves of Copilot at once — the autocomplete drafts the function, chat drafts the tests — so you see how it actually feels. Swap in a small task from something you're actually working on. Prompt: Write a function that takes a list of numbers and returns the average. Then write three tests for it. Pitfalls: - Credits ≠ autocomplete. Inline completions are unlimited on paid plans; only chat/agent/review draw down credits. Heavy agent use is what burns the monthly allowance. - The free tier is a taste, not a workflow. Its monthly caps are for evaluating, not for daily reliance. - The training opt-out is per-account and easy to miss. If your code is sensitive, set it before your first session, not after. Where to next: - AI Coding & Development: https://www.launchpadhq.io/discover?category=ai-coding-development - Local & Open-Source AI: https://www.launchpadhq.io/discover?category=local-open-source-ai ### Getting Started with Glasp Canonical: https://www.launchpadhq.io/platform/glasp/guide Data: https://www.launchpadhq.io/data/v1/guides/glasp.json Last reviewed: 2026-09-15 Tool: Glasp (https://www.launchpadhq.io/platform/glasp): starting price Free; free tier yes; data as of not dated A highlighter for everything you read — and a decision about who gets to see it. You install a small add-on in your browser, then select text on any page the way you would to copy it, pick a color, and it is saved. A sidebar shows what you have marked on that page, and everything lands in one library you can read back later. What it is: - Glasp is a highlighter for the web. It saves the passages you mark on articles, PDFs and YouTube transcripts into one searchable library, so the good bits of what you read stop vanishing into closed tabs. - What makes it unusual is the social layer. Glasp is built on the belief that highlights are worth sharing — its stated mission is "to democratize access to other people's learning" — so you get a profile page other people can visit, and you can see where other readers stopped on the same article. - There is AI on top of the highlighting: summaries of pages and YouTube videos, a chat box for questioning a PDF, and an "AI Clone" that answers from your own saved highlights instead of the open web. Before you start: - The first thing to do is not to highlight anything. Open Settings and find the visibility control for highlights — and the separate one under AI Clone & Hatch — before you mark a single passage. Which way those switches start is genuinely unclear from outside Glasp (see Privacy below), and they decide whether your reading is public. - Free to use with a sign-up and no card. The thing you would pay for here is privacy, not capability: Pro — a monthly plan, cheaper if you pay for a year, with a student discount — is what adds unlimited private highlights. Whether a free account gets any private highlights at all is not something Glasp's own pages settle either way. - Highlighting itself is unlimited and unmetered, but the AI features are rationed on the free plan: a small monthly allowance of chat credits (the platform's unit of spend — each thing you ask costs some), a daily cap on basic YouTube summaries, and a cap on PDF uploads. Chatting with a PDF is metered too, on a separate allowance Glasp's pages do not describe consistently, so check the current limits in the app rather than planning around fixed numbers. - It is a browser add-on, so it lives in a browser: Chrome, Safari, Edge and Firefox have published versions, and Brave and Opera install the Chrome one. Phone browsers cannot run add-ons at all, which is why there are separate iPhone, iPad and Android apps with their own built-in reader. - You are not locked in. Glasp states plainly that export is not a paid feature — "your highlights are yours, and we would rather you could leave than feel stuck." Security: - Glasp is a publishing tool wearing a highlighter's clothes, and that is the single most important thing to understand before you use it. Its mission statement is about sharing what you read, and its free plan is sold on public highlights. Some of that is not optional: Glasp's privacy policy (last updated 31 August 2026) says it will publicly share your profile picture, name, user handle and any social media handles you have added to your profile. Anything marked public "can be indexed by search engines." Security: Check these three things before you highlight anything sensitive: - The visibility default, because Glasp's own pages contradict each other. Its support page says "Highlights are private to you unless you choose to make them public." Its plan comparison sells the free tier as "unlimited public highlights", and its own how-to guide says "By default, Glasp highlights are public." Those cannot all be true. Look at the setting in your account and trust that rather than any of the three pages. - Your AI Clone, which is a separate switch. It is set to Public, and Glasp says other people can ask someone's clone a question without an account at all. Settings → AI Clone & Hatch changes it to Private. - What Glasp does with the content regardless. The policy says it "may use information we collect through the Services, including content you create and save, to develop, train, evaluate, and improve machine learning and artificial intelligence models." It names no opt-out, and it names no exception for private highlights either. Security: - The one genuinely reassuring thing here is where Glasp refuses to run. The extension is switched off entirely on online banking, sign-in pages, password managers, tax and immigration portals, health records and adult sites, and on your own network's router and storage pages — so those never produce anything to save in the first place. Security: - Deletion is partial rather than instant. Deleting an item removes it from view while Glasp keeps a limited record of it to enforce usage quotas; deleting your account removes your content and profile but keeps billing records and a snapshot of account and usage facts for up to two years. Best at: - Highlighting a YouTube transcript — Glasp's own support pages discuss timestamps on highlights, so it appears to keep the timestamp alongside the line - Seeing where other readers stopped on the same page — a genuinely different way of finding what is worth reading - Pulling your Kindle highlights out of Amazon and into something you control Okay at: - Being your whole reading app; there is a built-in reader and a feed, but the highlighter is what the product is actually good at - PDFs you did not open in a browser — they upload and work, but count against the monthly file cap Avoid: - Anything you would not publish under your own name. Your profile picture, name and handle are public by Glasp's own policy whatever you do with individual highlights. - Assuming the visibility setting is the whole privacy story — it governs who else can read a highlight, not what Glasp itself does with it (see Privacy above). Starter: Decide who your highlighting is for, before you do any Set both switches — highlight visibility, and the separate one under AI Clone & Hatch — then open your profile page, where an already-saved highlight can be flipped between public and private one at a time. Readwise gives you no public profile at all and nothing a stranger can query; Glasp ships you both by default, so your first five minutes here are spent switching things off rather than on. Starter: Highlight one article end to end Pick a long article you were going to read anyway. Select a passage, choose a color, and add a tag while you are there. Open the sidebar to see everything you marked on that page in order. Glasp captures the color and the tag at the moment you select the passage, not at an import step afterwards, so the reason it mattered is recorded while you still know it — which is the part you have forgotten by evening. Turn on the crowd-highlights view on the same article to see where other readers stopped. Starter: Ask your own reading a question The AI Clone answers only from passages you saved, so this question hands your own reading back to you — something ChatGPT cannot do unless you paste the material in first. Name a topic you highlight a lot to narrow it down. Prompt: Across everything I've highlighted, what are the three ideas that keep coming back? Quote the passage behind each one and tell me which article it came from. Starter: Put a PDF through the chat box Glasp keeps the chat box and the highlighter on the same file, so an answer worth keeping becomes a highlight in the same library — rather than a screenshot in a second app, which is what a standalone tool like ChatPDF leaves you with. Prompt: Summarize this PDF in ten bullet points, then list every claim it makes that it doesn't back up with a citation. Starter: Set up the way out before you need it In Settings → Integrations, connect Notion, Obsidian, Readwise, Roam, Zotero, Tana, Workflowy or Anki — or just run Export All Highlights and take a Markdown or CSV file. Re-running an export only sends the pages that changed since the last one, so this behaves as a backup you repeat weekly rather than a one-time escape hatch — which is the reason to wire it up on day one instead of the day you leave. Pitfalls: - Kindle highlights are imported, not synced — you re-run the import after each reading session. They only exist for books bought from Amazon and read on a Kindle device or app, and Amazon caps how much of any one book can be exported, so a long book can arrive incomplete. That cap is Amazon's, not Glasp's. - Highlighting will sometimes do nothing at all, and the reason is rarely a bug. Text drawn as part of an image or inside an embedded frame from another site cannot be highlighted by any browser add-on, and Glasp also switches itself off on the excluded sites listed under Privacy above. Reloading the page first fixes a third case: the add-on has to be running before the page finishes loading. - If a highlight shows in the add-on but is missing from the website, it is almost always an account mismatch — the add-on popup can be signed into a different account from the one in your browser tab. - There is no way to share with just a few people. Glasp said in August 2026 that it has no group feature, so your only two settings are effectively "just me" and "anyone." - Do not read the public changelog as a sign of life. Glasp's "What's new" page has not been updated since February 2025, even though the mobile apps were rebuilt in 2026 and prices changed in May 2026. The blog and the pricing page are the pages that move. Where to next: - PDF and document tools: https://www.launchpadhq.io/discover?category=document-pdf-processing - Notes and meeting tools: https://www.launchpadhq.io/discover?category=meetings-notes - Learning tools: https://www.launchpadhq.io/discover?category=education-learning-ai ### Getting Started with Gong Canonical: https://www.launchpadhq.io/platform/gong/guide Data: https://www.launchpadhq.io/data/v1/guides/gong.json Last reviewed: 2026-09-15 Tool: Gong (https://www.launchpadhq.io/platform/gong): starting price Custom; free tier no; data as of not dated What to do in your first week on a Gong account your employer pays for. Gong sits on your work calendar and records your sales calls. Afterwards you open a call in your browser and see the recording, a written transcript and a summary. A chat panel sits beside it: you type a question about what was said, read the answer, and keep asking follow-ups. What it is: - Gong is a call-recording and analysis platform for sales teams. It captures the meetings and phone calls a company's reps run with customers, then analyses what was said across all of them. - Gong's core licence is called Gong Foundation, and your employer assigns you a seat on it. Specialised applications — Enable Essentials, Forecast Essentials, Gong Forecast, Gong Engage and Data Cloud — are bought separately on top of it. - The part a new user actually touches is Gong Assistant, the chat panel that answers questions about your calls, accounts and deals. Gong announced in March 2026 that it was renaming Ask Anything to Gong Assistant, but the change has not reached everywhere: the mobile help page still calls it "Ask anything" and still says it is available on any Gong plan, while the web pages say Gong Assistant requires a Gong Foundation seat. Gong has not reconciled the two, so what you see on your phone may not match the browser. - Most of what Gong produces — call summaries, activity, deal signals — flows back into your CRM (the system a sales team keeps its customer records in), and that round-trip is usually why a company buys it. Before you start: - There is no way to sign yourself up, so this is not a tool you can go and try tonight. Gong's pricing page publishes no prices, offers no trial and no self-serve plan — only a form to "prepare a customized proposal for you." You get access because your employer buys Gong and an admin assigns you a seat. If what you want is hands-on practice with AI on recorded meetings this week, Otter has a free tier you sign up for yourself in a browser, and it will teach you most of the same habits — asking a transcript questions, checking what it got wrong — on calls you own. - What your company pays is per person, plus a platform fee: Gong states on its own pricing page that "Licenses are priced per user" and that "There is a platform fee based on the number of users supported," but publishes no figures for either, and no vendor page states a seat minimum or a floor price. Per-seat numbers circulating on comparison sites are someone's estimate rather than Gong's price. As a user you never see a bill or a usage meter; the cost question is settled above you. - Your admin, not you, decides which of your calls are recorded and who can see what. If your calls are not appearing in Gong, that is almost always a company-level setting rather than anything you did. Gong also sells a cheaper Call Intelligence seat alongside Gong Foundation; Gong marks that help page as excluded from its own search, so it may not reflect what is currently sold, but if the chat panel is missing for you it may be that your company bought that seat rather than a Foundation one. - Gong records you as well as your customer. Your manager — and, depending on how permissions are set, other colleagues — can play back your calls, read the transcript and score them. Treat every external call you run as reviewable, including the small talk before the customer joins. Security: - Gong is a recording system, and that makes its legal position different from a chatbot's. Recording law varies by country and, in the United States, by state: some places need only one person on the call to be aware, others require everyone to agree, and when the two sides sit in different places the stricter rule usually applies. Gong's own guidance is blunt about where that lands — decisions about whether and how to record in a particular jurisdiction "belong exclusively to the customer and their legal counsel," and its recording-law article carries a note that it "does not constitute legal advice." Gong also flags its own recording-consent help page as stale in its documentation, so check the current version with your admin rather than treating the detail below as settled. Security: How participants are told they are being recorded: - A consent page Gong hosts, opened before the participant joins the meeting - An audio prompt played as they join - A pre-call email sent about 10 minutes before the meeting - Zoom's own recording disclaimer, where a company uses native Zoom recording - Which of these run is set by your admin in a consent profile and applies to every call — you cannot switch them on or off for one meeting Security: Controls worth knowing on day one: - Marking the calendar event private stops Gong recording the call at all — but it does not make an already-recorded call private - Setting a recorded call to private hides its content, not its existence: colleagues still see that the call happened, who owned it and when, it still counts in Gong's statistics, and it still exports to your CRM flagged as private - Your chat history with Gong Assistant is personal to you and not visible to teammates - Your company chooses whether its Gong data lives in a US or an EU data centre; US is the default, and the choice is made when the company onboards Security: - Under European and UK privacy law your employer is the party responsible for the recordings; Gong acts on your company's instructions rather than on its own account. In practice that means a customer asking to be deleted from your records is handled by your admin, who has tools for exactly that — it is not something you can action from the chat panel. Gong deletes a company's data within 30 days of its contract ending. Security: - One question the public documents do not settle: whether your conversations are used to train Gong's own AI. Gong's terms permit it to use customer data to "provide insights, recommendations and value analysis" without addressing model training either way, and its privacy policy and help centre are silent on it. Gong's trust centre lists an Artificial Intelligence section, but its contents did not open without an account, so this page cannot say what those documents commit to. If it matters to you, ask your admin. Best at: - Answering "what did they actually say" about a call you were on — responses on a call page can cite the moment in the transcript they came from - Finding a pattern across a pile of calls at once — objections, competitor mentions, recurring themes — without listening to any of them - Getting up to speed on an account before a meeting or a handover, using calls plus the emails Gong has captured for that account - Turning conversations into reusable material: crib sheets, FAQs, talk tracks, onboarding docs and scorecards Okay at: - Calls in languages other than English. Gong transcribes more than 70 languages, but several analysis features — email labels, red-flag alerts, recommended to-dos and example-based smart trackers — only produce results for English calls. Gong marks that language-support page as stale in its own documentation, so the list may have moved on since it was last reviewed. - "Either/or" and "everything except" questions. Gong's own documentation says to set filters first for those, because a typed question combines conditions with "and" - Reading Gong in your own language: the interface currently displays in English, French and Canadian French only, and Gong says localisation is still rolling out, so parts of it stay in English Avoid: - Treating an answer as the whole conversation when someone refused consent partway through. If a participant joins late and declines, Gong keeps the portion recorded before they joined as long as roughly 70% of the planned call was captured — and the answer is built on that partial transcript without making the gap obvious - Settling who said what in an internal meeting. Many companies exclude internal calls from recording entirely, and Gong only classes a meeting as internal when every invitee works at your company — one outside address changes the rule that applies - Trying to force a recording through by inviting the recorder by hand. That does not override an enforced consent requirement that was not met, or a colleague on the call whose settings say never record — the call simply will not be recorded Starter: Ask the call what actually happened On a call page the panel reads the transcript together with the account, deal and participant records attached to that call, so it can answer about the relationship and not just the hour — Otter.ai, working from the same recording, has only the transcript to go on. Or skip typing: the panel offers Summarize this call, Suggest next steps, List objections and Draft follow-up as one-click actions. Prompt: What did the customer commit to on this call, what did I commit to, and what was left unresolved? Starter: Ask one question across a month of calls Gong resolves timeframes, deal status, deal value, named accounts and named participants straight out of the sentence by matching them against your CRM records — a generic meeting recorder like Fireflies has no deal or account fields to match against, so the same sentence is just a keyword search there. Prompt: What concerns did customers raise in closed-lost deals during the last 30 days? Starter: Set the scope by hand when the sentence won't carry it Open Search, filter down to the calls you mean, then click Assistant — or open the Assistant page, click Add context, and pick calls, folders or streams. Gong publishes the list of question shapes its natural language does not handle reliably — CRM fields like industry or territory, sentiment, specific transcript phrases — and tells you to filter first for those rather than leaving you to discover it from a wrong answer. Starter: Turn what you found into something the team can use Gong Assistant produces documents and Gong assets — scorecards among them — that open in the Assistant page where you can read them properly and ask for revisions, rather than returning a wall of chat text you have to copy out. Prompt: Using these calls, write a one-page crib sheet on the objection that came up most: the objection in the customer's own words, the three best responses reps actually used, and what to avoid saying. Starter: Give it something the calls don't contain Click Assistant, choose Add context, then Upload files — up to 10 files per message, PDF, MD, TXT or CSV, each up to 20 MB — then ask your question. Uploading the template, price list or competitor deck you want the answer shaped around lets Gong answer from your calls and the document together, which is the difference between "summarise this call" and "score this call against our own checklist." Pitfalls: - Not every call reaches Gong. Private calendar events, meetings booked less than 30 minutes ahead, calls hosted by somebody outside your company, and internal calls your company has excluded are all skipped unless you invite the recorder yourself — so "it isn't in Gong" is not evidence a conversation never happened. - The chat only knows the context it was opened with. From a call it sees that call; from Search it sees your current search results. Change the filters and your next question uses the new set, while the answers already on screen still refer to the old one. - Answers inherit transcript errors. Gong's own FAQ says an incomplete transcript, or one transcribed in the wrong language, is the usual cause of a wrong or thin answer — the fix is to re-transcribe the call in the right language and ask again, not to rephrase the question. - Capital letters matter more than you would expect. Gong matches account names and participant names in your question against the names in your CRM, and its documentation notes that the matching is case sensitive. - A new chat is a blank slate. Gong holds the thread of a conversation only while that chat stays open, and moving between Your library and Company library starts a fresh chat whether you meant it to or not — so keep one line of enquiry in one chat. Where to next: - Meeting recorders you can actually sign up for: https://www.launchpadhq.io/discover?category=meetings-notes - Sales, marketing and SEO tools: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai - AI inside business software: https://www.launchpadhq.io/discover?category=ai-plugins-business-software ### Getting Started with Google Antigravity Canonical: https://www.launchpadhq.io/platform/google-antigravity/guide Data: https://www.launchpadhq.io/data/v1/guides/google-antigravity.json Last reviewed: 2026-09-08 Tool: Google Antigravity (https://www.launchpadhq.io/platform/google-antigravity): starting price Free; free tier yes; data as of 2026-10-20 Google's free AI coding app: hand it a whole task and it plans, edits your files, and clicks through a browser to check its own work. You install it like any desktop app, open a folder of code, and type what you want in plain language. It works through the task on its own — reading files, editing them, running things, even clicking through a browser — then hands back a plan and screenshots of what it did for you to approve or send back. What it is: - Google Antigravity is a free desktop app for building software with agents (AI that takes actions on its own rather than only answering): you hand out tasks, several run in parallel, and you review what comes back. - The unusual part is the reach. As well as your files, an agent can drive a real browser, opening what it just built and clicking through it. - Cursor and GitHub Copilot sit beside you as you type. Antigravity is built to be handed a task and checked on later. Before you start: - You must be 18 or over. Google says Antigravity is currently unavailable to under-18 users, and checks the age on your Google account. It is a hard stop, and it rules the tool out as a school or family starting point. - Sign in with a personal @gmail.com address — Google's own FAQ points people there when a work or school (Workspace) account misbehaves. Eligibility also follows the country on your Google account: most of the world, not everywhere. - It costs nothing and asks for no card; the paid tiers are Google's general AI subscriptions, AI Pro and AI Ultra, and they buy headroom rather than features. How the allowance refills matters more than the price. The free allowance refreshes weekly, while paid plans also get a top-up roughly every five hours, so running dry on the free plan can cost you days rather than an evening. - You need a folder of code to point it at. To build something without that background, Lovable in this same category works from a written description in a browser. Installs on macOS 12+, Windows 10 (64-bit)+ and Linux. Getting set up safely (Google's official Antigravity download and documentation): - Download only from antigravity.google. The name has attracted lookalike sites with their own downloads and their own privacy policies. - First launch signs you in with Google and puts a security and data use policy on screen. That screen is worth reading rather than clicking past. - Point it at a throwaway project for the first hour. Its agents change real files and run real commands on your computer, and the undo that matters is the copy you kept outside the app. Security: - The work happens on your machine; the thinking happens at Google, and the files an agent touches are sent there. Free and paid use fall under the same terms, which are blunt about it: interactions are used to develop and improve Google's products and machine learning, and Google employees and contractors may access, view and review them. The opt-out is a toggle called Enable Telemetry, in Settings under Account. Security: Three things that change the answer: - Which model you pick changes whose rules apply — Anthropic's and OpenAI's models sit alongside Google's, and using them means agreeing to those companies' terms too. - Personal and work use are different contracts. Organisations get Antigravity through Google Cloud under Google Cloud's terms, so your own Gmail sign-in is not covered by what your employer negotiated. - The browser agent drives a real browser window, and whatever is loaded there during a task — a signed-in dashboard, say — is within its reach. Best at: - Being handed a whole task and left to it - Several agents working at once across different projects - Checking its own work by clicking through the app and recording it Okay at: - Being your everyday code editor - Small edits, where explaining the task costs more than doing it - Predicting what a task will spend Avoid: - Anyone under 18 — Google refuses the account outright - Work or school Google accounts; Google's own FAQ sends people to a personal Gmail - Code a person at Google should not read, unless you have turned telemetry off Starter: Ask it to explain a project before it changes anything Antigravity answers as an artifact — a written walkthrough saved next to the project, not a chat reply that scrolls away — so your first task also shows you how it reports back on every task after it. Ask which parts look unmaintained. That answer beats the tour. Prompt: Read this project and explain what it does, where a request enters, and the three files I should read first. Do not change anything yet. Starter: Give it one small change and make it prove the change worked The browser agent is the piece GitHub Copilot has no equivalent for: it clicks through what it built and hands back a recording, so "it works" becomes something you watch rather than take on trust. Ask for a phone-sized window too. Prompt: Change the homepage heading to "[new text]", then open the site in the browser and show me a screenshot proving it worked. Starter: Make it write the plan before it writes the code Where Claude Code shows you each step in a session you sit and watch, Antigravity is built to be left alone with several agents running, so an approved plan is often your only checkpoint before they act. Change one line of the plan before approving, and watch how it takes the correction. Prompt: Before you touch any files, write a plan for [the change] — what you will change, in which files, and how you will know it worked. Wait for me to approve it. Starter: Find out where your free week actually ends Opens your usage after a normal afternoon of tasks and notes how much of the weekly allowance is gone. Work an ordinary afternoon, then check your usage. Running dry on Windsurf costs you until tomorrow; here the free allowance is weekly, and the five-hourly top-up that rescues a bad afternoon is a paid-plan feature. Pitfalls: - Signing in with a work or school account and hitting an eligibility error — a personal Gmail is the path of least resistance. - Treating the age check as a formality. Under-18 accounts are refused, and the fix is age verification on your Google account, not a setting in the app. - Reading the write-up as proof. The plan and the prose are what the agent believes it did; the screenshot and the recording are the evidence. Where to next: - More AI coding tools: https://www.launchpadhq.io/discover?category=ai-coding-development - Start with a chat assistant: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Google Imagen (Now Nano Banana) Canonical: https://www.launchpadhq.io/platform/google-imagen-3/guide Data: https://www.launchpadhq.io/data/v1/guides/google-imagen-3.json Last reviewed: 2026-09-12 Tool: Google Imagen 3 (https://www.launchpadhq.io/platform/google-imagen-3): starting price See pricing; free tier yes; data as of not dated Google's image generator has been renamed and moved — here's where it actually lives now, and what you get without paying. You open Google's Gemini app in a browser or on your phone, pick the image tool, and type a description of the picture you want. Pictures come back a few seconds later. Then you talk to it — "make the sky darker," "lose the hat" — and it redraws, keeping the parts you liked. What it is: - Imagen is a model — the AI "brain" that does the actual thinking — not an app you sign into. You reach Google's picture-making through one of Google's own products, and the model underneath can be swapped for a newer one without your being asked. - It has been swapped twice. Google's list of retired models records Imagen 3 as shut down on 10 November 2025, and the Imagen 4 models that replaced it as scheduled to shut down on 17 August 2026, with a Gemini image model as the replacement. What you get today is a model Google calls Nano Banana 2 — its real name, not a nickname; the Gemini app puts a banana on the button. - For a beginner the swap is good news: what replaced Imagen lives inside the free Gemini app and needs only a Google Account. This page covers that surface — gemini.google.com and the Gemini phone app — not Google Cloud's Vertex AI, where the Imagen name still appears in the developer documentation and which assumes cloud billing. Before you start: - There is no Imagen website, app or download. Go to gemini.google.com, or install the Gemini app, and sign in with a free Google Account — that is the entire setup. Searching for "Imagen" lands people in Google Cloud's developer console, which opens by asking about projects and billing, or on a lookalike site with nothing to do with Google. - You can make pictures without paying. Google's plan comparison lists "image generation and editing" on the free plan, alongside several paid plans. Its two plan pages disagree — the Google One version lists image access only on the paid rows — so assume the free allowance is real but modest. - The clearest free-versus-paid line is the download: Google's help page says images download at 1K resolution without a Google AI plan and 2K with one. 1K is fine for a screen or a slide; small once you want to print or crop in hard. - The limits are not counted in pictures. Google describes usage as compute-based across everything you do in the app, refreshing every five hours until a weekly ceiling — so a long chat earlier in the day eats into the images you were saving. Subscribers who hit it drop to a lighter, faster model; Google does not say what a free account gets there, so treat running out as "come back in a few hours". - Google's help page puts generating images at 13 and over, or the applicable age in your country, and editing at 18 and over. Availability follows the Gemini app's supported countries and languages, and school accounts may be restricted further. Security: - The unusual thing here is not what you type in, it's what comes out. Every picture carries a watermark — a mark on generated output identifying it as AI-made — and here there are two: an invisible one Google calls SynthID, designed to survive cropping, filters and compression, plus a visible Gemini mark on the free plan. What to weigh before pressing send is where the picture ends up, and whether anyone downstream needs to know a machine drew it. Security: Which rules actually govern your pictures: - The standalone Generative AI Additional Terms stopped applying on 22 May 2024, when Google folded AI into its main Terms of Service. The governing document today is the Google Terms of Service, version effective 30 July 2026 — read that, not the friendlier help pages. - They say your content stays yours: you keep whatever intellectual-property rights you have, while granting Google a broad licence to host, reproduce and use it to operate and improve its services. - Google's Gemini Apps privacy hub, last updated 10 August 2026, asks you to get permission before using someone's face or voice in Gemini — guidance rather than a licence term, but the closest thing Google publishes to a plain rule about other people. - Google may remove images when its systems detect a possible violation of the Terms of Service, including the Prohibited Use Policy — a picture appearing is not the same as one allowed to stay. Best at: - Changing one thing in an existing picture by saying what to change. - Words inside the picture: signs, posters, labels. Google's product page leads on spelling and typography — the traditional weak spot of image generators, and the easiest claim to test. - Keeping the same character recognisable across pictures — Google's documentation claims this for the current model. - Starting cold: a free Google Account is the whole setup — no card, no waitlist, nothing to install. Okay at: - Precise control. You steer by describing the change in words, not by masking or brushing a region — "a bit further left" is a negotiation, not a setting. - Matching something outside the chat — an exact brand colour, a real product's shape. It gets close and confidently misses. Avoid: - Anything that has to pass as human-made. Between the marking above and Google's Terms of Service, which forbid misleading others into thinking generative-AI content was created by a human, this is the use the platform is built to defeat. - Putting a recognisable person into a situation they were never in. Google's Generative AI Prohibited Use Policy — the version in force when this page was checked was dated 17 December 2024 — bans impersonating an individual "living or dead" without explicit disclosure to deceive, and separately bans non-consensual intimate imagery. - Feeding the output into training your own image model. Google's Terms of Service prohibit using AI-generated content from its services to develop machine learning models or related AI technology. Starter: The spelling test Send it once, then switch the app's model setting from Fast to Thinking and send the same words again — the Gemini app offers that choice next to the description, so one prompt (the message you type) buys two results. Miscounted letters are the one error you can judge without any taste. Change only the words in capitals; that isolates the lettering from everything else. Prompt: A vertical poster for a neighbourhood plant swap. The words on the poster read exactly: PLANT SWAP — SATURDAY 10AM — BRING ONE, TAKE ONE. Hand-lettered chalk style on dark green, one large monstera leaf in the corner. No other text anywhere. Starter: The one-thing-at-a-time edit Editing is another sentence in the same conversation, so the picture has to still be in that thread — and it is the feature Google gates at 18 and over while generating opens at 13. Stay in the same chat, one change per message: "Keep everything identical but make it raining outside the window." Two changes at once is where the picture drifts. Prompt: A photo-realistic cat sitting on a windowsill at sunset, shot on a 50mm lens, warm light, shallow depth of field. Starter: The same character twice Its reference is the picture sitting in your chat history, not a set of images you upload and train a style on — so testing Google's consistency claim costs one more message. In the same chat: "Same fox, same scarf and glasses, now at a desk writing a letter." Judge the claim on a third scene rather than the second. Prompt: A friendly cartoon fox wearing a red scarf and round glasses, flat vector style, plain cream background, front view, full body. Starter: Same words, two Google boxes Send a description you already like twice: once in the Gemini app, once in Google Search's AI Mode. Note which gave the better picture, and whether either interrupted you to suggest an upgrade. Google's subscription page sells "more access to Nano Banana in Search" as something a paid plan buys, while the free plan's own line item is image generation outright — so the same sentence can be free in one Google box and gated in another. Starter: Ask it whether it made the picture Save one of your generated images, open a new chat, upload it and ask whether Google AI created it. Do the same with a photo you took yourself, and compare. Google states you can upload an image and ask the app whether Google AI made or altered it, reading back the invisible marking above — the detector is something you already have, not a promise to take on trust. Pitfalls: - Following a guide written for Imagen 3. The buttons, the model picker and the results all moved on when that model was retired, so a step-by-step accurate last year will not match your screen. - Opening a new chat to make a change. Editing works on the picture in the conversation you are in; start fresh and you are describing it from scratch, with no way to say "the same one, but." - Expecting a named real person to come out right. Between the age gate on editing and the policy limits on impersonation, this is where refusals cluster and results go generic. Where to next: - Image generators: https://www.launchpadhq.io/discover?category=image-generation-editing - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai - Video creation and editing: https://www.launchpadhq.io/discover?category=video-creation-editing ### Getting Started with Google Lyria Canonical: https://www.launchpadhq.io/platform/google-lyria-3/guide Data: https://www.launchpadhq.io/data/v1/guides/google-lyria-3.json Last reviewed: 2026-09-15 Tool: Google Lyria 3 (https://www.launchpadhq.io/platform/google-lyria-3): starting price Free, from See pricing; free tier yes; data as of not dated Describe a song in a sentence and Google's Gemini app writes, sings and mixes it — free. You open Gemini's music page, pick a template or type a sentence describing the song you want — mood, genre, instruments, what it's about. A minute or so later a finished track plays back with its own cover art. You listen, change the description, and generate again until one sounds right. What it is: - Lyria is Google's music-generating model (the AI "brain" that does the actual thinking) — it writes, performs and mixes a complete track, vocals and all, from a plain-English description. - You never meet it by name. It runs inside the Gemini app, on the Music page, behind the same login you would use to chat with Gemini — the button says "Music generation," not "Lyria." - Google keeps shipping new versions under the same name: the one currently serving the Gemini app is Lyria 3.5, from September 2026. Businesses and developers can also buy Lyria through Google's paid Vertex AI and developer products, but you need none of that to make a song. Before you start: - You need a Google account, and Google requires you to be 18 or older to use music generation at all. - Nothing to pay to start. Google's plan comparison lists music generation on every Gemini plan including the free one, so a free account can make tracks. What the paid Google AI Plus, Pro and Ultra plans buy you is a higher ceiling on how much you can generate — not a better music model. - The one setting that will stop you: Gemini refuses to generate music unless Keep Activity is switched on. That is the setting that saves your chats to your Google account, so the price of making a song here is that the session is kept. A temporary chat or an activity-off account will not work. - Full-length tracks — around two to three minutes — need the Gemini model switched to Pro in the model picker. Free accounts get some Pro access rather than unlimited, so you will run out of long tracks sooner than short ones, which run about a minute. - Available in every country where the Gemini app itself is available. Google's launch announcement named eight interface languages — English, German, Spanish, French, Hindi, Japanese, Korean and Portuguese — and that list may well have grown since; what language your *lyrics* are in is a separate question from what language the app speaks. Security: - Because the feature cannot run with your activity switched off (see Before you start), music generation is the rare Gemini feature with a privacy floor you cannot lower. Security: - Under the Gemini Apps Privacy Notice, last updated 29 June 2026, human reviewers read some of what you type; chats are kept 18 months by default, adjustable to 3 or 36 months; and a conversation picked for review can be held for up to three years. Google's own wording is not to enter anything you would not want a reviewer to see. Security: - None of this makes normal use risky. Write about your cat, your commute, your friend's birthday. Just do not treat the lyrics box as private. Best at: - Background music for a video, a slideshow or a podcast intro, where you need something that fits the mood and is not a famous song - Novelty and occasion songs — a birthday track, an in-joke, a song telling your roommate to do the dishes, which is an example Google itself publishes - Trying one idea in four genres in a single sitting Okay at: - Song structure — asking for a shape works better than it once did, but you are still describing a shape, not arranging one. - Long-form music — anything past the full-length ceiling has to be several tracks joined together outside Gemini. - Revising one moment in a track you otherwise like. The documented flow is describe-and-regenerate; if there is a way to edit an existing track in place, Google's help page does not describe one. Avoid: - Sounding like a named artist. Google states the model treats a named artist as broad creative inspiration rather than something to imitate, so "in the style of [famous singer]" gets you the genre and not the voice — and the gap between those two is where beginners waste an afternoon. - Anything that needs to be undetectable as AI. Every track the Gemini app produces carries SynthID, an inaudible watermark (a mark on generated output identifying it as AI-made), by design and with no opt-out. - Assuming you may sell it. Google's Terms of Service, in force since 30 July 2026, say Google does not claim ownership of your content, and separately forbid using AI output to train other AI models — but no Google page we could find grants commercial music rights in so many words. Check the terms before you put a track behind a paywall or monetise a video with it. Starter: The one-sentence song Google's own Lyria prompting guidance asks you to name genre, era, instruments, vocal character and subject, and this line supplies all five into a single box — the one field is doing work that Suno splits across a separate style field and a separate lyrics field. Change one word — swap "hopeful" for "resigned" — and regenerate, to hear how much of the result that single adjective was holding up. Prompt: A warm, mid-tempo indie folk song about moving out of the first apartment you ever loved. Acoustic guitar, brushed drums, a female vocal that sounds a little tired. Hopeful at the end. Starter: Hand it your own words The word Lyrics: is a real instruction to Lyria rather than decoration — Google's prompting guide uses that exact prefix to switch the model from writing *about* your subject to singing the text you supply. Keep supplied lyrics short. Google's guidance is that a track has limited room for words, so a full page of them gets compressed or dropped rather than sung. Prompt: A 90s-style pop punk track, fast drums, a shouty male vocal. Lyrics: Verse: the sink is full again / the mugs have grown a skin / Chorus: Ryan, do the dishes / Ryan, do the dishes Starter: Switch to Pro and ask for a shape Open the model picker in Gemini, choose Pro, set the length control to the full option, then ask for a track with named sections — an intro, two verses, a chorus and a bridge. Prompting for named intros, verses, choruses and bridges is a capability Google added specifically with Lyria 3 Pro; paired with the full-length setting it is the difference between three minutes of loop and a track that goes somewhere. Ask for the bridge to drop the drums. Naming what an individual section should *stop* doing tests the structure control harder than naming what it should do. Starter: Start from a template, not a blank box On the music page, open the template gallery and pick a genre — Pop, Hip-hop and rap, Rock, K-pop and others are offered — then edit the description it hands you rather than writing from nothing. The templates are Google's own worked examples of the prompt shape Lyria responds to, so editing one puts you inside the format Google's prompting guidance describes rather than guessing at it. Notice how many words the template spends on instruments versus mood, and copy that ratio into descriptions of your own. Starter: Prove the watermark is real Download a track you made, then upload the file back into an ordinary Gemini chat and ask whether Google AI generated it. Gemini reads SynthID straight out of a file you hand it, which means the same trick works on audio someone *else* sends you — the check is not limited to your own downloads. Try it on a piece of music you know a human made, to see what a negative answer looks like. Pitfalls: - Writing one word and judging the tool on it. "A love song" is a valid description and Google lists it as one, but it hands every decision to the model. - Forgetting the cover art is generated too. Downloading the MP4 gets you a video with an AI-made image attached to the music; if you only want audio, take the MP3. - Expecting a published daily allowance. Google says only that limits exist and that you will get a notification when you pass one, with the choice of waiting for a refresh or upgrading — so plan on discovering your ceiling rather than reading it. - Assuming your tracks are saved somewhere. The output arrives as a file to download or a link to share, and Google's help page does not promise a permanent library of what you have generated, so save the ones you want to keep. Where to next: - Other music generators: https://www.launchpadhq.io/discover?category=music-generation - Video creation and editing: https://www.launchpadhq.io/discover?category=video-creation-editing - AI chat assistants: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Google Veo Canonical: https://www.launchpadhq.io/platform/google-veo-3-1/guide Data: https://www.launchpadhq.io/data/v1/guides/google-veo-3-1.json Last reviewed: 2026-09-07 Tool: Google Veo 3.1 (https://www.launchpadhq.io/platform/google-veo-3-1): starting price See pricing; free tier yes; data as of not dated Describe a shot, get a few seconds of video with the sound already in it. You describe a shot — what's in frame, how the camera moves, what you hear — and the model (the AI "brain" doing the work) returns a few seconds of video with its own sound. Each attempt is a fresh roll, not an edit, so you steer by rewriting and going again. What it is: - A text-to-video tool from Google DeepMind, reached through Google's own apps — a Google account is the only login you need. - It makes picture and sound in one pass — Google's docs list native audio as always on, where most rivals in this class hand you silent video you score yourself. - That, a daily allowance Google's own pages describe as free, and no new account or card is why you'd pick it over Runway, Pika or the Sora-class tools. It lives in Google Flow, a studio with projects and a timeline rather than a bare text box. Before you start: - Start in Google Flow (labs.google/fx/tools/flow), not the Gemini chat app — video there currently needs a paid Google AI plan. - The free allowance is small and spent per attempt — and Google is inconsistent about whether you get one. Flow's own page and its credits help article say a signed-in Google account currently gets a small daily allowance of Flow credits, the unit of spend; Flow's "Get started" article lists a paid Google AI plan as a requirement. Check which one your account lands on before planning around it. On that reading, the top-quality Veo setting costs about ten times the cheapest: a handful of short clips a day, or a single top-quality one. Credits don't roll over, and mush costs the same as your hero shot. - You'll pay past the experimenting stage, because ten attempts at one shot is a normal session. Paid Google AI plans run from a low-cost entry tier through a mainstream tier to expensive top tiers; check the plan page, these move. - First step: sign in, pick the cheapest model, and run one description several times. How much a single wording varies is day one's real lesson. You must be 18 or over and in a supported country. Security: - Flow runs on your Google account and what you type is treated like other Gemini activity: Google says a subset of conversations is read by people (disconnected from your account first) and kept up to three years, and that your activity trains its generative AI models. Turning off Keep Activity stops that and cuts retention to around 72 hours. Assume a person could read what you type. Security: Watermarks, and what you may not generate: - Every video from Google's video models carries SynthID, an invisible watermark — a mark identifying output as AI-made. You can't turn it off, and Google's own detector reads it: disclose rather than be caught not disclosing. - A visible watermark is a separate toggle in Flow's profile menu, applied automatically for users in India, South Korea and Vietnam. - Google's Generative AI Prohibited Use Policy bans impersonating a person, living or dead, without explicit disclosure in order to deceive; non-consensual intimate imagery; and using personal data or biometrics without legally required consent. So don't animate a photo of a real colleague or public figure — expect refusals, and occasional false ones on invented characters that merely look real. Best at: - Short establishing shots and B-roll where nobody speaks - Atmosphere — weather, light, texture, room tone, ambient sound - Animating a still image you already like Okay at: - Dialogue and lip-sync — it'll try; results vary shot to shot - Continuing a clip you liked — on some Flow model options, not others - On-screen text, signage, product and logo accuracy Avoid: - Surgical edits to a clip you already have — a generator, not an editor - Client work signed off without a human watching every frame Starter: Write like a camera operator, not a customer Veo's controls are built around camera movement and framing, so film-set words work as instruction: handheld means the camera is carried rather than fixed, shallow depth of field means only the woman is sharp. The rain and the hiss arrive as sound in the same pass. Swap handheld for a locked-off tripod. Same story, colder temperature. Prompt: Handheld medium shot, shallow depth of field: a woman in a yellow raincoat steps off a bus into heavy rain, rain drumming on her hood, bus doors hissing shut behind her. Starter: Animate a photo you own Image-to-video is core to the Veo family, and because audio is native, a silent photograph comes back with a soundscape. Ask for a push in instead of a drift right. Prompt: Animate this still: the camera drifts slowly to the right, steam rises from the mug, soft morning room tone with faint birdsong outside the window. Starter: Bridge a first frame and a last frame First-and-last-frame control is a documented Veo capability — you fix where the shot starts and ends, and it solves only the middle. Use two near-identical frames with one moving object. Prompt: Move from the first frame to the last frame in one continuous handheld take with no cuts, keeping the ambient street sound running throughout. Pitfalls: - A "video" here is a handful of seconds. Flow's Veo options currently produce roughly four, six or eight seconds, the newer companion model about ten. Anything longer is you assembling clips, and continuity between them is the real work. - The model names change faster than any tutorial. Google ships new video versions constantly and is rolling out a successor, Gemini Omni, which replaces Veo in the Gemini app while both stay selectable in Flow. Trust the in-product menu over any guide, including this one. - Consistency across clips stays hard. The same character generated twice gives two similar-ish people; reference images improve the odds, not guarantee a match. Where to next: - More video tools to compare it against: https://www.launchpadhq.io/discover?category=video-creation-editing - Make the still images you feed it: https://www.launchpadhq.io/discover?category=image-generation-editing ### Getting Started with GPT4All Canonical: https://www.launchpadhq.io/platform/gpt4all/guide Data: https://www.launchpadhq.io/data/v1/guides/gpt4all.json Last reviewed: 2026-09-13 Tool: GPT4All (https://www.launchpadhq.io/platform/gpt4all): starting price Free; free tier yes; data as of not dated A double-click desktop app that runs AI on your own computer — free, offline, and able to answer questions about a folder of your own files. You install the app, pick an AI model — the "brain" that does the thinking — from a list, and wait while it downloads onto your hard drive. Then you type into a chat box and read the reply, which your own computer worked out. Ask a follow-up and it keeps going. What it is: - GPT4All is a free desktop app from a company called Nomic that downloads AI models onto your own machine and lets you chat with them with the internet switched off. Where Ollama started life as a tool you typed commands into and grew a window around it later, GPT4All has been a plain install-and-chat window from the beginning — and it ships with something a browser chatbot doesn't have: LocalDocs, which points the model at a folder of your own files and then shows you which file each answer came from. Before you start: - Hardware decides this, not how technical you are. Nomic's own minimum-specification sheet asks for 16 GB of RAM — your computer's short-term memory, what runs out when too much is open — and drops that to 8 GB only for the smallest models; on Apple hardware the minimum it lists is an M1 with 16 GB. Check what you have first (Task Manager on Windows, "About This Mac" on a Mac). Below the line the app still opens and the model still loads; it just crawls — if your machine is under the line, Ollama runs the same small models with a lighter footprint and is the gentler place to start. - There are installers for macOS (Monterey 12.6 or newer, and it is tuned for Apple Silicon), Windows, Windows on ARM, and Ubuntu-style Linux on Intel/AMD chips. One wrinkle worth knowing before you count on it: Nomic's own two pages disagree about ARM — the download page offers a Windows ARM installer while the requirements sheet says ARM PCs are not currently supported, so on a Snapdragon laptop treat this as worth a try rather than as promised. - Check how current it is before you commit to it. As of this review the newest tagged release is v3.10.0 from February 2025, the installers on the official download page were last changed in early February 2025, and the code repository's last push appears to be around May 2025. It has not been archived, renamed or handed to anyone else, and Nomic still hosts and links it — but Nomic's business today is an AI platform for architecture and construction firms, and GPT4All currently reads like a product being kept online rather than one being built. It installs and it works; expect no new features. - You will not need to pay for any of this. The app is free, there is no account, no card and no sign-in, and models cost nothing to download or run. The paid monthly plan advertised across Nomic's website is for its separate business platform and has nothing to do with GPT4All. Getting set up safely (Download from one place only — Nomic's own GPT4All page at nomic.ai/gpt4all, which is also where the old gpt4all.io address now sends you. This matters more here than anywhere else on the page: you are installing a program with the run of your whole machine, and "GPT4All" is a well-known enough name that copycat download sites and bundle installers exist. Not a search result, not a download aggregator, not a link from a forum post.): - Know where the big files land. The app itself is small; the models are not. Each one is roughly 2 to 7 GB, they are saved outside the app's own folder, and they stay there when you uninstall. A model's size is usually quoted in parameters — a rough measure of a model's size, where more is smarter but heavier — and the documentation lists a memory figure beside each example model, so you can decide before you click Download rather than after. - Confirm it worked, then confirm it is really local. Load the model, send one message, get a reply back — that is your setup checkpoint. Then turn your wi-fi off and send another. If it still answers, you have proved the thing the whole app is for. Security: - There is no sign-up, no server and no session here. The model file sits on your hard drive and your own processor does the thinking, which makes GPT4All about as private as an AI tool gets — and it changes what you have to watch. Nothing you paste is the risk. The risk is a switch inside the app, or a file you chose to install. Security: Four things to know:: - The first time you open it, GPT4All asks you two yes/no questions: whether to send anonymous usage statistics, and whether to share your chats with the GPT4All open data collection. The chat-sharing one is documented as off by default, and its own dialog spells out the consequence — that with it on you 'should have no expectation of chat privacy.' Answer no to both if privacy is why you are here. (These wordings come from the app's published source and settings documentation; the exact screen you meet may have moved.) - Adding an API key turns it into a different product. The app can be pointed at outside providers, and Nomic's documentation is refreshingly clear about what that means: prompts 'leave your computer to the API provider' and come back afterwards. Nothing wrong with the feature, but it is the one thing that undoes the reason you installed this. - There is a built-in local server, off by default, that lets any program on your device talk to your model on port 4891. Leave it off unless you have a specific reason and know what you are connecting to it. - Models are pulled from the public HuggingFace library, which is enormous and not curated. Stick to what appears in the app's own suggested list until you have a reason not to; treat anything else the way you would treat a program from a publisher you have never heard of. Best at: - Asking questions of a folder of your own files with the internet off - Privacy in the strongest form on offer — nothing you type crosses your network - Unlimited use for nothing once the model is on your disk — on a plane, in a basement, with the wi-fi switched off Okay at: - Answer quality — the models this app is built around sit in the 3-to-13-billion size band, and that band has a ceiling - How fast an answer arrives, which is decided by the machine and not by anything you can change in the app - Helping you choose: the Explore screen searches the whole public HuggingFace library, which is thousands of files with very little to tell a newcomer which one is right Avoid: - Windows or Linux machines with ARM chips, and older Intel Macs — the requirements sheet's minimum Apple entry is an M1, and the vendor's own two pages contradict each other on whether Windows ARM is supported at all - Anything you intend to sell or publish without checking the individual model's licence first — the app's MIT licence is not the model's - Making it your only AI when the quality of the answer matters more than where it runs — a cloud chatbot will beat it on hard questions Starter: On 8 GB: start with the smallest model, on purpose Downloads a 3-to-4-billion-size model — roughly a 2 GB file — which loads quickly, leaves room for everything else you have open, and proves your setup works end to end before you gamble on anything bigger. Nomic's published floor carves out 8 GB only for the smallest models, so on a modest laptop GPT4All tells you in advance which side of the line you are on and the small model here is a deliberate choice rather than a consolation prize. Judge it on the second answer. The first one carries the wait while the model is read off your disk, and that wait is not the model's speed. Starter: With 16 GB: step up to a 7-to-8-billion-size model Moves you to a mid-size general model — around a 4 to 5 GB download — which is where local AI stops feeling like a demo and starts being something you would actually reach for. The documentation's example table pairs every model with the memory it needs — a 4.66 GB Llama 3 file listed against 8 GB of memory — so you can decide whether this whole approach fits your machine while still reading a web page. Starter: Read the licence beside the model, not the one on the app Before you download anything you plan to use for work, look at the licence attached to that specific model. The app is MIT-licensed; the models are separately licensed by whoever made them, and they do not agree with each other. GPT4All's documentation states outright that "license agreements differ regarding personal and commercial applications," and its own example list runs MIT, Apache 2.0, GPL, Meta's Llama licence and a non-commercial CC-BY-NC-SA entry side by side — while the project's repository headline advertises "available for commercial use." That gap between the headline and the list is specific to how this project presents itself, and it is the reason to check. Starter: Point it at one folder of your own files (LocalDocs) Creates a named collection tied to a folder on your computer, reads through it on-device, and then lets you switch that collection on during a chat so the model answers from your documents instead of from general knowledge. The indexing runs on your machine using Nomic's own on-device models, and every answer gets a Sources control listing which file it drew from — private document search plus per-answer attribution in one window, built in rather than assembled. Start with one small folder, not your whole Documents directory. Wait for the collection to show Ready before you judge the answers. Starter: Ask something only your own files can answer The last clause is doing real work here, because GPT4All gives you a way to check it: the Sources control names the files behind the answer, so you can open one and see whether the quoted line is in it. A local model will invent a confident answer as readily as any other, and this is the one setup on the page where you can catch it in about ten seconds. Works well on things you would never upload anywhere — a lease, medical paperwork, old bank letters, a folder of personal notes. Prompt: Using only my documents, answer this: [your question]. Quote the exact sentences you based it on, and tell me if the documents don't actually say. Pitfalls: - The Explore screen will cheerfully hand you a model your memory cannot hold, and that is where most people wrongly conclude that local AI is useless. Match the model to what you actually have. - Looking for an update inside the app and finding none. Given how long it has been since the last release, "no update available" is the expected answer rather than a sign something is broken — the official download page is the place to check. Where to next: - Local & Open-Source AI: https://www.launchpadhq.io/discover?category=local-open-source-ai - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Grammarly Canonical: https://www.launchpadhq.io/platform/grammarly/guide Data: https://www.launchpadhq.io/data/v1/guides/grammarly.json Last reviewed: 2026-07-27 Tool: Grammarly (https://www.launchpadhq.io/platform/grammarly): starting price Free, from $12; free tier yes; data as of 2026-10-26 A writing assistant that follows you across every app — which is both the point and the thing to think about. Grammarly installs as a browser extension or desktop app and watches what you type wherever you type it — email, documents, forms, chat. It underlines what it would change as you go; you click a suggestion to accept it, or dismiss it and keep writing. What it is: - A writing assistant that works inside the applications you already use rather than in its own window. - Free for spelling, grammar, and basic tone; paid for rewrites, tone adjustment, and detection features. - Now one product within a company that renamed itself Superhuman — the writing tool is still called Grammarly. - Best understood as an always-on layer, not a tool you open. Before you start: - The company is now called Superhuman. It renamed itself on 2025-10-29, uniting the writing tool with an email client and a workspace product. You will see Superhuman branding and prompts for other products. The writing assistant you came for is still Grammarly. - The free plan covers spelling, grammar, and tone signals plus a limited monthly allowance of AI prompts — roughly a hundred. - The main paid tier is priced per person, cheaper if you pay annually, and raises the AI prompt allowance many times over. There is no published top-up if you exhaust it. - Decide where it runs before you install it. The extension can see what you type on the sites it is enabled for, and that choice is more consequential than any setting inside the product. Security: - The company states it does not sell user content or use it to train third-party models, holds recognised security certifications, and performs basic checks on your device rather than in the cloud. Taken on its own that is a reassuring posture, and there is no reason to doubt it. Security: - The exposure is a different shape: surface area. This is software with read access to text as you type it, across whatever sites and apps you enable. That is not a claim about misuse — it is a description of what the product is. The relevant question is not whether the vendor is trustworthy, but whether you want any third party positioned to read everything typed into your banking portal, your medical records, or your employer's internal systems. Security: Worth doing at install time: - Disable it on financial, medical, and government sites explicitly rather than assuming it is off there. - Check with your employer before installing it on a work machine. Many organisations have a position on this, and some have a policy. - Review what the newer assistant features connect to — email, calendar, files — and enable only what you need. - Revisit the enabled-sites list occasionally. It grows quietly as you accept prompts. Best at: - Catching mistakes in the moment, before you send - Consistency across everything you write, without switching tools - Tone checks on messages where the stakes are social rather than grammatical - Non-native writers wanting continuous, low-friction correction Okay at: - Long-form structural editing, where a dedicated assistant does more - Style and voice, where it tends to flatten toward neutral professional prose - Technical or specialist writing, where it flags correct jargon as errors Avoid: - Sites handling money, health, or credentials - Work systems without your employer's agreement - Accepting every suggestion, which produces bland and slightly generic writing Starter: Turn it off somewhere before you turn it on everywhere Restricts the extension to the sites where you actually want help. The default is broad. Narrowing it deliberately is the highest-value five minutes you will spend with this tool. Start with banking, health, and work systems disabled, then add exceptions as you find you want them. Starter: Run one important email through the tone check before sending Drafts the email somewhere Grammarly is enabled and reads the tone reading it shows before you press send. Tone detection is more useful than grammar checking for most adults, and almost nobody uses it deliberately. If the tone reading surprises you, that is the point. Rewrite before you send, not after they reply. Starter: Reject a suggestion on purpose Dismisses a suggestion you disagree with instead of accepting it, and keeps your original wording. Builds the habit of treating suggestions as opinions rather than corrections, which is what keeps your writing sounding like you. Pay attention to how often it wants to remove a deliberate stylistic choice. Those are the ones to keep. Starter: Watch your monthly AI allowance for one cycle Keeps track of how many AI prompts you use in a month against the free plan's small allowance, without changing how you write. Free and paid tiers differ mostly in this number, so knowing your real usage tells you whether upgrading would change anything. If you never approach the free limit, the paid tier is buying you features rather than capacity — decide on that basis. Pitfalls: - The company renamed itself; the product did not. Superhuman is the company, Grammarly is still the writing assistant. Expect unfamiliar branding and prompts for products you did not ask for. - It sees what you type where you enable it. That includes forms, portals, and internal systems. Disable it on anything financial, medical, or confidential, and check your employer's policy before installing at work. - Accepting everything flattens your writing. Suggestions optimise for correct and neutral. Applied uncritically, they strip the voice out of anything with personality. - It flags correct specialist language as wrong. Technical terms, names, and domain jargon get marked routinely. In specialist writing you will override it constantly. - The AI allowance is capped with no top-up. Exhaust the monthly prompt limit and you wait for the cycle to reset or move up a tier. There is no published option to buy more. Where to next: - Research and academic tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Assistants for longer writing: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Granola Canonical: https://www.launchpadhq.io/platform/granola/guide Data: https://www.launchpadhq.io/data/v1/guides/granola.json Last reviewed: 2026-10-23 Tool: Granola (https://www.launchpadhq.io/platform/granola): starting price Free, from $14; free tier yes; data as of 2026-10-14 An AI notepad for meetings: it listens through your computer instead of sending a bot to the call, then turns your rough notes into a clean write-up. Granola runs on your computer or phone and transcribes the meeting from your microphone and speakers — no bot joins the call. You jot a few words as you go. When the meeting ends, it combines the transcript, your jottings and the calendar invite into tidy notes, and you can ask it questions about any meeting afterwards. What it is: - Granola is a notes app for people who spend their day in meetings. You open a note for a meeting, type whatever short reminders you like while people talk, and Granola listens in the background. Afterwards it writes enhanced notes — a structured summary built from the transcript, your own notes and the calendar event. Your own words stay in black, and Granola's help centre says a magnifying glass next to each point shows whether it came from the transcript or from you. - The difference from most AI note-takers is that nothing visible joins the call. Granola captures your computer's audio directly, so it works with Zoom, Google Meet, Microsoft Teams, Slack huddles and in-person conversations. Its help centre says it does not record or save audio or video; audio is held only long enough to transcribe and then deleted. - Once you have a few meetings in it, the chat becomes the main event. Ask from the home screen and it searches across all your meetings; ask inside one meeting and it answers about just that one. Granola's own example question is "What did I promise to do in my meetings this week?" Prompts you use often can be saved as recipes and reused with a /. - It runs on Mac (macOS 14 or newer), Windows, iPhone and Android. The phone apps are aimed at in-person meetings and conversations away from your desk. Before you start: - Free gets you started; history is the limit. Granola's free Basic plan lets you see only your last 30 days of notes, and chat and AI connections only look at those 30 days too. Older notes are kept, not deleted — they reappear if you upgrade. - The paid step is a team plan. Granola's pricing page lists Business at $14 per user per month, which removes the history limit and adds integrations with tools like Notion and HubSpot, and Enterprise at $35 per user per month. Its help centre says Business is billed monthly; annual billing is offered only on Enterprise contracts. - You sign in with Google or Microsoft. Those are the only two sign-in options, and Granola works best with the account that holds your work calendar, because it reads your calendar to know which meeting you are in. Microsoft works with cloud Microsoft 365 calendars, not on-premises Exchange. - On a Mac, allow both audio permissions. Setup asks for the microphone and for system audio. If you skip system audio, Granola's troubleshooting guide says it will hear you but not the other people on the call. Security: - The first responsibility is to the people you are meeting with. Because no bot appears, they may not realise anything is transcribing. Granola's own guidance calls it best practice to make sure participants know, including in person, and says recording-consent rules vary by place — when in doubt, check with your legal or IT team. It offers two built-in ways to tell people: an automatic chat message in the meeting, or a watermark on your camera. Security: Settings worth checking on day one:: - Model training. On free and Business accounts, Granola's help centre says it may use anonymised data to improve its services by default. Turn off "Use my data to improve models for everyone" in Settings → Preferences (on desktop or iPhone). - Transcript auto-deletion. Transcripts are kept indefinitely unless you set auto-deletion (from one day to 12 months) under Settings → Preferences → Data & sharing. It deletes transcripts, not your notes, and can't be undone. - Sharing. Notes are private by default. Anything you move into a shared team space or folder becomes visible to colleagues in your workspace. Security: - Granola states it holds a SOC 2 Type II report and stores data on AWS in the US. If you discuss something you would not want on paper, you can pause transcription during that part of the meeting. Best at: - Back-to-back calls where you want to listen rather than type, then get a clean summary a minute after hanging up. - Finding what was said or promised across many meetings by asking in plain language. - Meetings where a visible recording bot would feel awkward — though you should still tell people. Okay at: - Knowing who said what. In Zoom, Meet and Teams it can use participants' names; elsewhere the transcript just says "Me" and "Them". - Noisy setups. It hears everything your computer plays, so music or videos in another window can end up in the transcript. Avoid: - Using it on a free plan as your permanent record — you only see recent meetings. - Leaving people unaware they are being transcribed. Starter: Your week, in one question This is Granola's own headline example. Asked from the home screen, chat searches across all your recent meetings, which is the thing a stack of separate notes can't do. On the free plan, chat only looks back over recent meetings — ask about recent weeks. Prompt: What did I promise to do in my meetings this week? List each item with who it's for and any deadline mentioned. Starter: Check a summary against the source Ask this inside a single meeting. AI summaries can smooth over disagreement or get a number wrong; asking for the transcript wording lets you confirm before you act on it. Prompt: What exactly was agreed about the budget in this meeting? Quote the relevant part of the transcript. Starter: Draft the follow-up Granola's chat can draft follow-up emails from the meeting it is open in. You review the draft before anything is sent, so you stay in control of what goes out. Ask for bullet points instead if your team prefers them. Prompt: Write a short follow-up email to the attendees summarising the decisions and next steps from this meeting, in a friendly but brief tone. Starter: Spot patterns across calls Granola's recipes guide suggests uses like spotting common sales objections and summarising feedback across calls. It works for any repeated meeting: interviews, customer calls, parent meetings. Prompt: Across my meetings from the last two weeks, what questions or objections came up more than once? Starter: Save it as a recipe Once a prompt earns its keep, save it as a recipe so you can run it after every meeting by typing /. Granola also has ready-made recipes under Discover in the Recipes tab. Prompt: List the action items from this meeting as a checklist, grouped by person, with anything unassigned at the end. Pitfalls: - Skipping the system-audio permission on a Mac. You get a transcript of only your own voice. - Bluetooth headsets can lower the quality. Granola's troubleshooting guide suggests setting your laptop's built-in microphone as Granola's mic instead. - VPNs and strict firewalls can block transcription. If nothing appears, try without them. - Templates are desktop-only. Notes taken on your phone can have a template applied once they sync to your computer. - Trusting the summary blindly. Use the magnifying glass or ask chat for the transcript wording before quoting a decision. Where to next: - More meeting and note tools: https://www.launchpadhq.io/discover?category=meetings-notes - Voice & speech tools: https://www.launchpadhq.io/discover?category=voice-speech - Automate what happens after the meeting: https://www.launchpadhq.io/discover?category=workflow-automation ### Getting Started with Grok Canonical: https://www.launchpadhq.io/platform/grok/guide Data: https://www.launchpadhq.io/data/v1/guides/grok.json Last reviewed: 2026-10-14 Tool: Grok (https://www.launchpadhq.io/platform/grok): starting price Free, from $8/month; free tier yes; data as of 2026-10-14 A chatbot with a live feed — it reads the web and X as you ask, so it can talk about what happened an hour ago. You sign in at grok.com or the phone app, type a question into a box, and read the answer. If it is about something current, Grok searches the web and X before replying, then shows you the links it used. What it is: - Grok is a general-purpose AI assistant from the company that also owns X (formerly Twitter). You can chat, have it write and edit text, explain code, make images in the same box, or talk to it out loud in voice mode, which is included free. - The reason to pick it over other chatbots is its wiring into X: most assistants search the open web, while Grok also reads live X posts — which is the mechanism behind "what are people actually saying about this" and the one source a web-search assistant cannot reach. - The reason to be careful is the same reason: when Grok says "people are saying," it is summarising an argumentative social network, not verified reporting. Its edge is speed and pulse, not accuracy — loud does not mean true. Before you start: - It costs nothing to start — a real free tier, not a trial that dies in a week. Sign-in is by X account or email, so an X login is the fastest way in. - The free tier works on a usage pool: you get an allowance, spend it, and it refills on its own schedule. The exact numbers change often, so treat the mechanism as the durable fact — you will hit a wall, and the wall lifts by itself. - A few questions and some writing a day costs nothing. You'll feel the limits if you run long research sessions, lean on it daily for work, or make a lot of images — image generation burns an allowance far faster than chatting. Grok is a standalone app (web/iOS/Android). Since 12 October 2026, new customers sign up through The X Subscription — not only as an X Premium+ perk — and existing SuperGrok plans continue for current subscribers unless they switch. Beyond the free plan, The X Subscription runs from Plus at $8/month up through Premium, Super and Ultra, each with more usage. Grok Bot is a separate product and a separate catalog page. - First step: go to grok.com, sign in, and ask it something that happened today. Security: - Grok runs in the cloud: everything you type goes to the company's servers, and by default your conversations help train future versions of the model (the AI "brain" that does the actual thinking). Change that before you paste anything you wouldn't want a stranger to read. Security: What to do in your first five minutes: - Turn training off. Website: Settings, Data, deselect "Improve the Model." App: Settings, Data Controls, same toggle. It only affects new conversations, not ones you have already had. - For a one-off sensitive question, use Private Chat — the ghost icon at top right. Those chats stay out of your history, aren't used for training, and are deleted within 30 days. - If you signed in with X, Grok's answers already draw on public X posts, including yours; opting out of that is a separate setting inside X, under Privacy and safety. Best at: - Questions about right now — breaking news, live scores, a stock moving today - What people on X are saying, and which way the argument is going - Explaining a confusing viral post, meme or bit of jargon - Fast everyday writing: emails, summaries, rewrites, first drafts Okay at: - Coding help — capable, but not the first tool most developers reach for - Making images, and summarising documents you upload - Long multi-step reasoning, which works better on the paid tiers Avoid: - Publishing a Grok image as your own — everything it makes carries a visible watermark (a mark identifying output as AI-made) that you can't remove - Treating it as a fact-checker on live events; it has repeatedly produced confident errors during fast-moving news - Questions where you want a careful, hedged answer — Grok is built to have an attitude Starter: Ask it what happened today Grok searches X and the open web as you ask, and shows its links. Put the same question to ChatGPT and you get an article write-up, not what the feed is saying. Prompt: What are the three biggest news stories from the last 12 hours? For each one, give me two sentences of context and link the sources you used. Starter: Decode something you saw on X Grok can read the replies and quote-posts, not just an article about them. Ask another assistant and you get a definition; ask Grok and you get the room. Prompt: Explain what people mean when they talk about this and why it is trending. Assume I have no background at all. Tell me who is arguing what, and be clear about which parts are opinion. Starter: Make an image from a sentence Image generation sits in the same chat box, no second tool and no separate signup — and every picture comes out watermarked, worth discovering on something throwaway rather than on real work. Prompt: Make an image of a small bookshop on a rainy street at night, warm light in the window, soft and painterly. Pitfalls: - Confident is not correct. Grok answers breaking news in a fluent, certain tone even when wrong, and has a documented record of getting fast-moving events wrong. Click through to the sources it cites. - The personality is a feature, and it can bite. Grok is deliberately less filtered than other assistants, which makes it refreshingly blunt and occasionally crude or badly judged. Don't send anything customer-facing without reading it first. Where to next: - Compare it with other chat assistants: https://www.launchpadhq.io/discover?category=text-conversational-ai - Tools built just for making images: https://www.launchpadhq.io/discover?category=image-generation-editing ### Getting Started with Groq Canonical: https://www.launchpadhq.io/platform/groq/guide Data: https://www.launchpadhq.io/data/v1/guides/groq.json Last reviewed: 2026-09-07 Tool: Groq (https://www.launchpadhq.io/platform/groq): starting price Free, from See pricing; free tier yes; data as of not dated Absurdly fast answers from open AI models — try it in a browser before writing any code. You open Groq's console in a browser, pick a model (the AI "brain" that does the actual thinking) from a dropdown, and type a question into a chat box. The answer arrives about as fast as the page loaded. Swap the model, ask again. What it is: - First, the name. Groq is not Grok. Grok is xAI's chatbot, the one on X. Groq makes computer chips and sells fast AI hosting to developers. - It didn't build the models it offers; it runs open ones from other labs on its own machines. Most AI services use GPUs, chips designed for video game graphics — Groq built its own chip for language models, which is why words arrive at page-load speed. You're renting speed, not intelligence. - Infrastructure, not a finished product. No app, no memory of past chats — the console is a testing bench for people building things, still worth twenty minutes of yours to feel what speed does to a tool. Before you start: - Free and genuinely free — an email or Google account, no card, nothing to install. Start in the playground rather than with a key — though Groq's docs never quite say the playground runs without one, so you may need to make a key once you are signed in. - The free plan gives you nothing to spend. It caps how often you can ask. It counts requests and tokens (chunks of text, roughly three-quarters of a word) per minute and per day; hit a cap and requests fail until it resets. Caps cover your whole organisation, not each key, and only the limits page in your console settings is current — published numbers change. - You probably won't pay. Usage beyond that is currently priced per million tokens, cents rather than dollars for the smaller models — check the pricing page. You'd upgrade for real users hitting the caps, for batch or flex processing, or for the spend cap under Settings, Billing, Limits, which is paid-plan only. Security: - What you type runs on Groq's servers. Groq's documentation says it doesn't retain customer data from the requests you send, but that inputs and outputs may be logged temporarily — currently up to 30 days — for troubleshooting and abuse investigation. Data Controls has a zero-data-retention setting that switches that off, at the cost of features needing stored data. Security: Practical rules: - It describes retention but says nothing explicit about training on what you send. Absence of a claim is not a promise either way — read the Services Agreement and Data Processing Addendum if it matters. - Make an API key — a password that identifies your app, and that spends your money — only when you have code to put it in. Never paste one into a webpage, chat, screenshot or public code; bots scan publicly posted code for keys within minutes. - Keep customer records, credentials and medical or legal detail out of the playground. - The realistic risk isn't Groq reading what you type. It's you leaking your key and someone else spending your money. Best at: - Speed — enough that live transcription and instant chat feel ordinary - Running well-known open models without owning a graphics card - Dropping into existing code: its interface matches OpenAI's, so most OpenAI tutorials work after changing a web address and key - Speech, where its Whisper-family models transcribe at a fraction of real time Okay at: - Model choice — a curated open catalogue that changes often - Being explored without code: it works, but it behaves like a developer tool - Long documents; respectable capacity, not the largest anywhere Avoid: - Expecting frontier intelligence — the top proprietary models are still ahead on hard reasoning and coding - Everyday chatbot use: no conversation history worth relying on, no integrations, no mobile - Building on a model you expect to exist next quarter — preview ones can vanish at short notice, and models move to enterprise-only Starter: Ask something you already know the answer to You're not judging the answer, you're calibrating what fast means — the one thing Groq's own chip buys you, and the one thing a screenshot can't show. Ask for 500 words and you watch a stream rather than a burst. Prompt: Explain what an API key is to someone who has never written code, in about four sentences. Starter: Send one question to a small model, then a large one The catalogue runs from very small and fast to larger and slower, and switching is a dropdown, not a new account. The small ones sometimes get this wrong; the larger ones take longer and usually don't. Groq sells speed, so look for a tokens-per-second readout beside each reply — that number, not the answer, is what you came to compare. Prompt: List every number between 40 and 80 that divides by 3 and whose digits add up to more than 10, and show your reasoning. Pitfalls: - It's Groq, not Grok. Groq is the hosting company at groq.com; Grok is xAI's chatbot at grok.com. A tutorial that mixes them up didn't check anything else either. - Speed is not intelligence. Answers arrive fast enough to feel authoritative. These are good open models, not frontier ones, and they make things up at ordinary rates. - The model list isn't stable. Models get deprecated and moved to enterprise-only, so a hardcoded name can stop working. Check the models page, and prefer production models over preview ones. Where to next: - More developer services like this one: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - Running open models on your own machine: https://www.launchpadhq.io/discover?category=local-open-source-ai ### Getting Started with Gumloop Canonical: https://www.launchpadhq.io/platform/gumloop/guide Data: https://www.launchpadhq.io/data/v1/guides/gumloop.json Last reviewed: 2026-09-07 Tool: Gumloop (https://www.launchpadhq.io/platform/gumloop): starting price Free, from See pricing; free tier yes; data as of not dated Build automations by dragging boxes onto a canvas — with AI steps built in, not bolted on. Gumloop gives you a blank canvas. You drag boxes onto it, and each box does one small job: fetch the new email, read the attached PDF, ask an AI for the total, add a row to your spreadsheet. Lines between the boxes set the order. Each box is a node — Gumloop's word for one step. What it is: - A visual automation builder. You connect your accounts once — Gmail, Slack, Drive, Notion, Airtable and a few hundred others — then build workflows that move information between them while you do something else. - AI steps are ordinary boxes on the canvas: "Ask AI", "Extract Data", "Categorizer". Summarising or pulling fields out of messy text sits mid-workflow instead of in a separate service. Gumloop's own published examples lean that way — contract clause extractors, invoice processors, PDF readers. - Why this over Zapier, Make or n8n, all already on this site? For simple "when this, then that" plumbing, Zapier is easier and free. Gumloop earns its price when the middle of the job needs judgement — reading, classifying, extracting — across a batch. Before you start: - There is no free plan. Gumloop dropped its permanent free tier in mid-2026; the pricing page now lists Pro and Enterprise plus a 14-day trial. The entry plan is a monthly subscription with an allowance of credits — the platform's unit of spend. Check the pricing page for today's figure, because this company repackaged recently. - So the real question is what fourteen days can show you. Bring a card, and one job worth automating — a folder of invoices, a weekly report you assemble by hand. Open the template gallery rather than a blank canvas and run someone else's finished workflow first. - Simple steps are cheap or free; the AI steps are what you actually pay for, billed on how much text the AI reads and writes, with the bigger, smarter ones costing more per word. Credits currently don't roll over. Security: - Building the workflow is the half you can see. The half you can't is the standing permission that runs it: link your Gmail or Drive and Gumloop can open, read and write in that account on a schedule, at three in the morning, without asking you again. Security: Sensible habits when connecting accounts: - Prefer workflows that read over ones that send, delete or overwrite. - Point early workflows at one test folder or label, not your whole inbox. - Every few months, disconnect apps no live workflow uses. Security: - Documents you push through AI steps go out to third-party AI providers to be processed. Gumloop's privacy policy says that for paying customers it does not use your uploaded data, flows or chat sessions to train any AI, and that those providers are contractually committed not to train on data sent through Gumloop. Better than silence — but treat confidential material as a deliberate decision, not a default. Best at: - Reading a batch of documents — PDFs, contracts, invoices, resumes — and pulling the same fields from each - Summarising and classifying messy text so it lands somewhere already sorted - Repetitive research and reporting: gather, condense, drop it somewhere Okay at: - Plain two-step plumbing with no thinking in the middle — more tool than you need - Real-time reactions; scheduled and triggered runs are the natural rhythm - Being learned from a blank canvas rather than a template Avoid: - Free-forever tinkering — Gumloop dropped its free plan in mid-2026, leaving Pro, Enterprise and a 14-day trial - Pointing an AI step at a batch before you've priced a single run — two hundred documents is not twenty times ten - Confidential material on the strength of the no-training promise — a policy, not a technical guarantee - Saving up an allowance for a big month — credits currently don't roll over Starter: When a PDF invoice arrives in Gmail, then extract the totals into a spreadsheet A Gmail trigger — the event that starts a workflow — watches one label, a document step reads the attachment, an AI step pulls supplier, date and amount, and a row lands in Sheets. Elsewhere, reading a PDF means bolting on a separate parsing service; here it is another box on the same canvas. Flag anything over a threshold into Slack instead of filing it silently. Starter: When a file lands in a Drive folder, then summarise it and file the summary A Drive trigger watches one folder, a document step reads each new file, an AI step summarises it, and the last step appends the summary to a running doc. It writes a plain-language summary of each new document into a running doc. Looping over the batch costs nothing — only the summarising does — so forty documents is a reasonable trial-week test. Ask for three bullets and one open question instead of a paragraph. Starter: When a form is submitted, then classify it and route it to the right place A form or Airtable trigger fires, a categoriser step decides which bucket the submission belongs in, and a router sends it onward. Classification is a first-class node, so "which of these five things is this?" is one box rather than a chain of brittle keyword rules. Add a catch-all branch so anything uncertain reaches a human. Pitfalls: - AI steps are where the money goes. Run one, check what it consumed, then multiply before you point it at the pile. - Test on a single item before switching it on. Batch automations don't make one mistake, they make two hundred identical ones. - Don't start on a blank canvas. Open a template, run it, break it, rebuild it — most people who bounce off this kind of tool bounced off an empty screen. Where to next: - More ways to connect your apps together: https://www.launchpadhq.io/discover?category=workflow-automation - Tools built for documents and PDFs: https://www.launchpadhq.io/discover?category=document-pdf-processing ### Getting Started with Harvey Canonical: https://www.launchpadhq.io/platform/harvey-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/harvey-ai.json Last reviewed: 2026-09-13 Tool: Harvey AI (https://www.launchpadhq.io/platform/harvey-ai): starting price Custom; free tier no; data as of not dated The AI most large law firms are buying — sold by contract, not by credit card, and here is the one part you can get for free today. You sign in through your firm, type what you need in plain English into a text box, and point it at your own case files and documents. It works through the task and hands back a draft or a summary with links to the pages it drew from. You read it, correct it, and ask again. What it is: - Harvey is legal AI built for the inside of a law firm or a company's legal department: it reads the matter files, drafts and reviews documents, and researches law in the jurisdictions whose sources it has licensed. - It is sold to organisations rather than to individuals, so if you are reading this on your own, Harvey is something you will meet through an employer or a school, not something you can start tonight. - It is worth understanding anyway, because a large part of the profession is standardising on it: Harvey says more than 200,000 lawyers at 2,400+ organisations across 70 countries use the platform. Before you start: - The barrier is not skill or hardware, it is procurement. As of September 2026 harvey.ai has no pricing page and no sign-up link — the only two buttons on the site are "Request a Demo" and "Login" — and Harvey's own Evaluation Terms of Service, last updated 9 January 2026, describe "an authorized free trial" you use "in accordance with Your instructions from Harvey," which means evaluations are granted by Harvey rather than requested by you. You will not realistically pay for this, because you cannot: nothing is sold in a unit one person buys, and the per-seat figures that circulate in the legal trade press are secondhand and should not be read as the price. If you want something in the same category you can start yourself this week, Spellbook currently offers a seven-day free trial you begin from a form rather than a sales call — it does not publish prices either, but you can be inside it the same day. - One part of Harvey is free to anyone right now. Harvey Academy runs online courses on the platform and on using AI in legal work, and Harvey's help centre currently states that "Both Level 1 Foundations and Legal Engineering certifications are available and free for anyone to take. A Harvey Academy account is all you need." Separately, if you are a law student or faculty member, ask your library: Harvey's Law School Program gives partner schools "free access to the Harvey platform," and enrolment runs through the institution rather than the individual. - Where you are matters more here than with most software. Harvey's research quality depends on which publishers it has licensed, not on how good the AI is — around August 2026 it added 100+ regional sources across Germany, South Africa, the UK, France, Poland, Brazil and Vietnam, plus Spanish localisation for Spain and Mexico. That expansion is also the tell: coverage is still uneven by country, and the same question will be better answered in some jurisdictions than others. - Everything on this page about what the screens look like comes from Harvey's own product pages, help centre and developer documentation rather than from hands-on use, because there is no way to see the product without a firm account. Treat the interface detail as Harvey's account of its own software; you may need to check it once you are signed in. Security: - With Harvey the document worth reading is not the security page but the Platform Agreement — a firm signs it, Harvey publishes it in full, and it states in writing what happens to privileged material once it is inside. Security: What Harvey commits to in writing: - Its Platform Agreement, last updated 9 January 2026 and the version in force today, states: "Harvey will not train any AI models using Your Content or Customer Data. Subprocessors will not train any AI models using Your Content or Customer Data." The second sentence is the one worth reading twice — it binds the outside AI providers Harvey runs on, not just Harvey. - The same agreement sets a deletion window: "Within 30 days of termination, Harvey will securely delete any remaining Customer Data or Content unless otherwise instructed by You." - Log-ins are per-person and "may not be shared, including within the same organization" — so stretching one licence across a team is a breach of the agreement, not a workaround. - A customer can choose where the data physically sits. Harvey's security page says you can "keep everything in-region—EU or Switzerland, US, or Australia," which is how a firm answers a client whose own contract forbids its files leaving a territory. - It lists SOC 2 II (an audit of how a company handles customer data), ISO 27001, ISO 27701 and ISO 42001, plus GDPR and CCPA compliance. Security: - The control you would not find in a consumer tool is the ethical wall. Harvey says it "syncs and enforces your firm's existing ethical wall policies, blocking restricted users from accessing or sharing walled content" — those are the internal barriers that stop lawyers on opposite sides of a conflict seeing each other's files — and around August 2026 it added automatic syncing of those restrictions from Intapp. This is the sharp edge of putting AI in a law firm: if the conflicts rules do not reach into the AI tool, the AI tool is the hole in them. Best at: - Working across a whole matter's documents at once — you query the pile and get review tables and cited reports back, rather than opening files one at a time - Producing output with links back to the source pages, which is the difference between a draft a partner will read and one they will not - Living inside Word and Outlook where the drafting actually happens — around August 2026 Harvey added playbook reviews in the web app and matter-grounded reply drafting in Outlook - Jurisdictions where it has licensed the local publishers — Tirant lo Blanch for Spain and Mexico, Otto Schmidt for Germany, FromCounsel for UK restructuring, Wolters Kluwer in the US - Showing the people who signed the contract whether anyone is using it — its Command Center reports adoption by practice group Okay at: - Being one person's tool. Nearly everything good about it depends on your organisation's documents, templates and matter history already being loaded into it - Judgement. It drafts, extracts and compares well; deciding what position to take is still the lawyer's job Avoid: - Reading "fully cited" as "already checked." Harvey markets review-ready cited output, while the Platform Agreement the same company asks you to sign calls the Service "a research tool" whose output "is not legal advice" and "may contain errors and misstatements or may be incomplete." Both of those are Harvey's own words - Using it to work out what AI can do in general. It is a set of legal work surfaces wired into a firm's document systems, not a sandbox Starter: Take Level 1 Foundations at Harvey Academy Harvey's own online course walks through how it frames legal AI — what each piece of the platform is for, what "legal engineering" means in its vocabulary, and how it expects work to be reviewed before it leaves the building. It ends in a certificate and a LinkedIn badge. Open enrolment is the mechanism: it puts the vendor's actual method — not its marketing — in front of a non-customer. Neither CoCounsel nor Legora has published anything equivalent; their training sits behind a customer relationship, so this is the only one of the three you can study from the outside. Starter: Understand what Vault is before you ever meet it Vault is the document side of the platform. A matter's files go into one project and you ask questions across all of them at once, getting back extraction tables and cited reports rather than one answer per document. The mechanism is a published number: Harvey's Vault page currently states capacity of around 100,000 files stored per vault. Legora publishes no comparable figure, and that number is what makes "ask the whole matter one question" a different motion from uploading a contract and asking about it. It is also why Harvey turns up in diligence and investigations work specifically. Starter: Look for the approval step before anything runs Harvey's Agents are handed a task — a research memo, a set of extractions — and produce the finished work rather than a chat reply. Its platform pages say you "preview the plan, adjust the scope, and approve work before Harvey begins." That gate is a supervision decision, not a convenience feature: an agent (AI that takes actions on its own rather than only answering) running unattended across privileged files is a professional-responsibility problem before it is a technical one, and Harvey's published answer is a plan a human signs off first. CoCounsel has stated no equivalent preview-and-approve step, which makes this the single thing worth watching for in a demo. Starter: Check whether your jurisdiction is actually covered Harvey's research is only as good as the publishers licensed into it. Named publishers are checkable in a way that "global coverage" is not — you can ask whether the source you already rely on is in the list. Legora's list is a different set of names, and that difference is the product: for legal research the licensed source list, not the AI itself, is what you are buying. Starter: Notice which model is doing the work Harvey does not run on one model (the AI "brain" that does the actual thinking). Around August 2026 it said Mistral Medium 3.5, Opus 5 and Fable 5 were available as reasoning models, opt-in for the most complex work. The menu being published and opt-in is the mechanism: a firm can pin sensitive work to a named provider and tell a client which one touched their matter. That list is Harvey's own, published in its monthly product notes, and it turns "which model saw this file?" into a question a client can now actually ask. Pitfalls: - Expecting the mobile app to be a way in. Harvey's iOS and Android app is free to download, which looks like consumer availability — it is a companion to a firm workspace, and without one there is nothing behind the login. - Confusing an Academy certificate with product access. Finishing Level 1 Foundations gives you a credential and Harvey's vocabulary. It does not give you a workspace. - Bringing real client material to an evaluation before your firm's own agreement is signed. An evaluation runs under Harvey's Evaluation Terms of Service rather than the full Platform Agreement, so which document governs your test files is a question to settle before anything is uploaded. - Judging it from a demo run on Harvey's data. The things that make Harvey different — the ethical walls, the matter history, the firm's templates — only exist once your organisation's own material is inside it, so a demo shows you the least Harvey-specific version of the product there is. Where to next: - Legal AI tools: https://www.launchpadhq.io/discover?category=legal-ai - Document & PDF Processing: https://www.launchpadhq.io/discover?category=document-pdf-processing - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Healthee Canonical: https://www.launchpadhq.io/platform/healthee/guide Data: https://www.launchpadhq.io/data/v1/guides/healthee.json Last reviewed: 2026-09-12 Tool: Healthee (https://www.launchpadhq.io/platform/healthee): starting price Custom; free tier no; data as of not dated Ask plain questions about the health insurance you get through work — if your employer gives you the app. You open its app or website and type a question about your health insurance — that typed message is your prompt. Zoe, its assistant, answers in plain language from your own plan's details, and a human specialist steps in when Zoe can't answer. What it is: - Healthee is a health-benefits assistant: it holds the details of the health insurance your employer gives you and answers questions in ordinary language. The assistant is called Zoe, and you will see that name far more than "Healthee" once you are inside. - It is not a general-purpose AI; its only subject is your own coverage. A general chatbot has no idea what your plan covers, how much you have already spent on care this year, or which doctors near you your plan has a price deal with. Before you start: - You cannot buy this for yourself. Healthee is sold to employers, insurance brokers and benefits administrators — never to individuals — at a quote-only price set by "the size of the employer, the specific modules and capabilities deployed." Its App Store listing, published by developer Insurights Inc., puts it plainly: "If your company offers Healthee, you can use our AI-powered app." So the first step is your workplace benefits portal, or your last enrollment email: look for "Healthee" or "Zoe" by name. - If it does not, Healthee's employee page invites you to fill in a form and says "We'll team up with HR to integrate with your current benefits program." That is a request routed to your HR team, not an account today. Meanwhile a general AI tool can read the plan PDF itself — and because a plan summary carries your name and coverage, use Ollama, which keeps it on your machine, or ChatGPT with the identifying details stripped. - It costs you nothing personally: the app is free to download, and Healthee's terms of use say your employer "is generally responsible for paying us for your access and use of the Platform." There is no upgrade to buy and no free-tier limit to hit. - United States only, adults only: the terms say "You must be located in the United States to be a Registrant" and "at least eighteen (18) years old." Healthee calls itself "carrier-agnostic," so it is not tied to one insurer — but it still needs your employer's plan data loaded. Security: - The privacy question here is about what was loaded before you typed anything: your plan documents, what you are enrolled in, and how far into your deductible you are — the amount you pay yourself each year before insurance starts covering things. Security: What Healthee says, and when it said it: - Its privacy policy, last modified 27 March 2026, says it shares your information "to your employer (our customers) in order to manage the services we provide to you," and that after you ask for deletion it may keep backups for up to 90 days. - Its FAQ says: "Healthee does not use protected health information (PHI) or personally identifiable employee data to train AI models." - Its security page claims HIPAA compliance — the US law governing health information — plus SOC 2 Type II (an audit of how a company handles customer data), ISO 27001, ISO 27017 and ISO 27018, with no audit dates published for any of them. Security: - Two Healthee pages disagree about what your employer can see. Its security page, headed "We keep your info safe from outside parties–even your employer," says: "As a processor of your information, we receive health plans from your employer. We do not transfer back any information to them." Its FAQ says HR teams "can see how employees are using the platform, what questions they're asking, how they're making plan selections." Neither says whether that HR view is individual or pooled, and we could not establish which without an account. Until Healthee says plainly, ask Zoe what you would be comfortable asking if HR might later read the transcript. Best at: - Answering "what will this cost me" — the question a plan document technically answers and practically does not. - Comparing the plans your employer offers at enrollment against how much care you actually expect to use. - Healthee says the assistant runs around the clock and handles over 50 languages. Okay at: - Anything that is not your benefits. It is a specialist, not a general chatbot. - The hardest problems. Healthee's FAQ says appeals, plan-change requests and "nuanced coverage disputes" are escalated to a human benefits specialist — the AI part stops where the painful problems start. Avoid: - Treating its answer as the last word on your coverage. Healthee's terms of use — the binding document — say "WE DO NOT PROVIDE ANY GUARANTEE, WHETHER EXPRESSED OR IMPLIED, REGARDING THE ACCURACY OR COMPLETENESS OF THE RESULTS" and "You should always verify all insurance information directly with their insurance provider," while its marketing page says Zoe answers "instantly and accurately." - Using it to decide whether you need care. Healthee's terms say the company "is not a licensed medical professional," that no doctor–patient relationship is created, and that it is "not a substitute for medical care." - Counting on it to outlast the job. Healthee does not publish what happens to your account or history when you leave — assume it goes with the job. Starter: Find out what a scan or procedure will actually cost you Healthee's FAQ lists this as a question Zoe is built to take, worked out against your own plan, your deductible position and live cost data rather than a list price. Ask ChatGPT the same sentence and the best it can honestly give is a national price range. Prompt: How much will I pay for an MRI at [name the exact clinic or hospital]? Starter: Check a specific doctor before you book "In network" means your plan has a negotiated price with that doctor; out of network the same appointment can cost several times more. Healthee's FAQ names this as a Zoe question checked against your plan's provider list, and says the assistant can carry you into booking. Prompt: Is Dr. [full name] in my network, and what will a first visit cost me? Starter: Make the enrollment choice with your own numbers in it Healthee's FAQ lists this comparison as something Zoe answers, scored against the plans your employer actually loaded. The second half of the sentence does the work: the terms warn results can miss "factors not captured in our questionnaire." Use the plan names on your own enrollment paperwork, and do this while your enrollment window is still open. Prompt: Should I enroll in the PPO or the high-deductible plan this year? I take [medication], see a specialist about [how many] times a year, and expect [no procedures / one procedure]. Starter: Get the paperwork after a visit explained Healthee's FAQ says employees can ask Zoe about the statement an insurer sends after an appointment. That document carries your name and a diagnosis — the exact thing this site's ChatGPT page tells you not to paste into a general chatbot. Here it is a question about data the system already holds. Prompt: I got a statement from my insurer for a visit on [date]. Explain what each line means and what I actually owe. Starter: Tell it the things a plan document cannot know about you Early on, give Zoe the facts in no document: your medications, the specialists you already see, any procedure planned this year, who else is on your coverage. Healthee's terms refer to a questionnaire whose answers feed its recommendations, and warn your real needs "may vary from what is described in the Results due to factors not captured in our questionnaire" — the vendor naming its own blind spot. Go only as deep as you are comfortable: it lands in the same system covered by the employer-visibility question above. Pitfalls: - Downloading the wrong app. The free listing is "Healthee - Smarter Benefits", developer "Insurights Inc." — Healthee's legal name, not a copycat. Anything under a different name is not it. - Assuming everyone's Healthee is the same Healthee. Zoe is "connected to each employer's specific plan data in real time," and pricing depends on "the specific modules and capabilities deployed" — so features a colleague raves about may not be in what your employer bought, and a thin answer means ask your benefits team, not that the tool is bad. - Using it as your record. We could not find a published retention period for Zoe conversations; the privacy policy speaks to deletion requests and 90-day backups, not chat history. Keep the insurer's own paperwork for anything you might need to point at. Where to next: - Healthcare AI: https://www.launchpadhq.io/discover?category=healthcare-ai - Chat assistants you can sign up for today: https://www.launchpadhq.io/discover?category=text-conversational-ai - Tools that read a PDF for you: https://www.launchpadhq.io/discover?category=document-pdf-processing ### Getting Started with Hedra Canonical: https://www.launchpadhq.io/platform/hedra/guide Data: https://www.launchpadhq.io/data/v1/guides/hedra.json Last reviewed: 2026-10-15 Tool: Hedra (https://www.launchpadhq.io/platform/hedra): starting price $20/month; free tier yes; data as of 2026-10-17 Turn one photo of a face and a voice recording into a talking video — under a rule about whose face that can be that is stricter than most beginners expect. You open Hedra in your web browser, attach a photo of a face and a recording of speech, and type a sentence or two about how the person should come across. Hedra sends back a short video of that face saying your audio, mouth moving in time with it. You watch it, then ask for changes in plain words. What it is: - Hedra makes talking-character videos. You give it one still image of a face and an audio track, and it animates the face so the mouth, eyes and head move with the speech. Hedra's documentation calls these avatar or presenter videos, and says a sharp, front-facing photo with nothing covering the mouth or eyes, plus clean audio, gives the most reliable result. - That job now sits inside a broader studio. Hedra's current workspace is built around an agent (AI that takes actions on its own rather than only answering): you describe the result you want in a chat box — that message is your prompt, the message you type — and the agent plans and generates video, images and speech onto a canvas beside the conversation. Each project lives in its own "Space", which keeps the conversation, your uploads and the results together. If you would rather pick every setting yourself, a set of manual tools does one job at a time, including an Avatar mode for exactly this task. - Behind one account it runs a range of video, image and voice models, and a live model picker shows what each one needs as input and what it costs to run. It is aimed at creators and marketers: its ready-made starting points include presenter videos, creator-style ads, product spots, storyboards and animated photos. - There is also a developer API for building Hedra into other software; nothing on this page touches it. Before you start: - The only real person you may put into Hedra is you. Hedra's Acceptable Use Policy (the live page carries no date; its download link is the version filed as 31 July 2024, seen 15 October 2026) forbids using the service "to input, generate, or create output that contains personal information related to any individual other than yourself, including (without limitation) any photos, audio, videos or any other media containing individuals other than yourself." The single exception is images of people who "died at least one hundred years ago." The same policy separately bans deepfakes — impersonating real people or creating fake personas "without consent or legal right." Note that the photo-and-audio rule has no permission exception: a friend, a parent, a colleague or a client saying "yes" does not, on the policy's wording, make their face or voice allowed. Hedra's own face-swap help article says only to "use faces you have permission to use", which is looser; where the two disagree, the policy is the binding document, so follow it. Your own face and voice, characters you generate inside Hedra, and Hedra's library voices are the safe inputs. Breaking the policy lets Hedra suspend or close your account. - Signing up is free and needs no credit card. Everything you make is paid for in credits (the platform's unit of spend — each thing you make costs some), and Hedra does not publish how many credits the free plan includes. Costs differ from model to model, so check the picker's number rather than guessing. - Paid plans are monthly subscriptions — Basic, Pro, Ultra and Teams, each with its own monthly credit allowance — with Enterprise priced on request. Hedra's billing help says monthly credits reset each cycle and do not carry over, while one-off credit packs you buy separately do not expire. Its terms say a subscription renews automatically at the then-current price until you cancel; cancelling stops billing, and you keep your credits until the period ends. - Hedra's privacy policy says the service is not intended for anyone under 18. Security: - What makes Hedra different from a text chatbot is what it extracts from your uploads. Hedra's Biometric Data Privacy Policy (effective 23 July 2024) says it may generate facial templates from photos and derive information from voice recordings to build avatars, and that it will not keep that biometric data longer than three years after your last interaction with Hedra. Its privacy policy (effective 4 September 2026) says this derived data may also be used to improve the service and develop new products, and that data stripped of identifying details may be used to train its AI models. Neither page describes a switch to opt out of those uses. Security: Three things to know before you upload your own face:: - Sharing a Space with someone shares its canvas, its files and the prompts used to make them. Hedra's docs say your chat with the agent stays private, but the prompts do not. - Cancelling your subscription does not remove anything. To remove your account and the personal data tied to it, Hedra's docs point to Settings → Privacy → Delete account, which also ends any subscription immediately; the privacy policy says deleted information can sit in backups for up to 30 days. - Hedra is US-based, and its privacy policy says your information may be moved to the United States or other countries whose privacy laws may be less protective than yours. Best at: - Making one still face speak an audio track you supply, with the mouth moving in time — the job Hedra's Avatar mode and presenter starting point exist for. - Keeping a made-up character recognisable across many clips, by saving it once as a reusable reference and calling it up by name. - Trying the same idea on several different video models from one account, with each one's cost shown before you spend. Okay at: - Scenes with more than one speaker. Hedra's docs say to make the intended speaker unambiguous in your instruction and to use speaker controls only where the chosen model offers them. - Animating a photo with no speech at all. The docs say small, physically plausible movement holds up far better than several competing actions in one short clip. Avoid: - Videos fronted by any real person other than yourself — a relative, a co-worker, a client's spokesperson — even with their blessing; the policy in Before you start has no permission exception. - Planning a budget from documentation. Hedra's own docs say exact model names, credit prices, duration limits and resolutions are deliberately not listed there, and they can change. Starter: Make yourself talk Before you paste this, attach a sharp, front-facing selfie and a short voice memo you recorded yourself. Under the yourself-only rule in Before you start, your own face is the one real face this page can send you off with. Keep the first one short. If the lips drift out of time, ask for one change — 'keep everything, but make the mouth movement match the audio more closely' — instead of rewriting the whole brief. Prompt: Make a 15-second presenter video using my photo and my voice recording that I've attached. Frame me from the chest up, keep my face and background exactly as they are, and make the delivery warm and relaxed, with small natural head movements and no hand gestures. Starter: Give a historical portrait a voice This uses the policy's one exception, for people who died at least a century ago. A public-domain portrait of a nineteenth-century figure, read aloud by your own voice or one of Hedra's library voices, is the most dramatic thing you can make here that sits squarely inside the rules. Ask for a still camera and one small movement first. Prompt: Animate this portrait so the person in it speaks the attached audio. Keep the face, clothing, background, lighting and painted style unchanged. Subtle head movement and natural blinking only; the camera stays still. Starter: Invent a presenter, then reuse them A character you generate is not a real person, so the policy's yourself-only rule does not bite. Once you like the result, save it under Library → Elements — Hedra's docs allow up to 10 reference images per Element — and type @ plus its name in any later prompt to bring the same face back, which is what keeps a recurring presenter looking like the same person. Pair your saved presenter with a voice from the Speech tool (next item) rather than a recording of a real person. Prompt: Create a portrait of an original, fictional presenter: a friendly woman in her forties with short grey hair and a green cardigan, front-facing, evenly lit, plain light background, looking straight at the camera. Starter: Write the script first, then hand it to the face Open Manual tools → Speech, choose a library voice, type a few sentences of script and generate the audio. Then, in the same Space, add that audio as the reference for your avatar video. Hedra's docs recommend this order because the agent receives everything already on the canvas when you send your next message. It also keeps you inside the rules — a library voice is not anyone's real voice — and the Speech tool lets you insert emotion and pacing tags by typing a forward slash. Hedra's docs suggest short sentences and spelled-out abbreviations wherever pronunciation matters. Pitfalls: - Generating a long video first. Hedra's own advice is to start with a short test before a long performance. If a generation fails, your billing page's generation history should show the charge and, often, an automatic refund — check there before assuming the credits are gone. - Letting the agent choose everything when you care about cost. Describing what you want in the main chat box lets Hedra pick the tools and model for you. If you want to see and choose the model and its price yourself, use Manual tools → Video → Avatar instead. - Feeding it too many references. Hedra's docs warn that more images are not automatically better if they disagree about identity or style. Two clear, consistent photos beat six mismatched ones. - Mixing projects in one Space. The docs advise one creative goal per Space, because the agent leans on earlier decisions in the conversation. Start a new Space for a new idea rather than piling it onto an old one. Where to next: - Video Creation & Editing: https://www.launchpadhq.io/discover?category=video-creation-editing - Voice & Speech: https://www.launchpadhq.io/discover?category=voice-speech - Image Generation & Editing: https://www.launchpadhq.io/discover?category=image-generation-editing ### Getting Started with HeyGen Canonical: https://www.launchpadhq.io/platform/heygen/guide Data: https://www.launchpadhq.io/data/v1/guides/heygen.json Last reviewed: 2026-07-13 Tool: HeyGen (https://www.launchpadhq.io/platform/heygen): starting price Free, from $29/mo; free tier yes; data as of 2026-10-25 Turn a script into a talking-head video with a realistic AI avatar — no camera, studio, or on-camera talent required. You paste a script, pick a presenter and a voice, and it renders a video of that avatar speaking your words with matched lips and gestures. You watch it back, fix the script or swap the presenter, and render again. What it is: - HeyGen makes AI avatar videos. You paste a script, pick a presenter — a stock avatar, one built from a single photo, or a digital twin of yourself trained from a short recording — choose a voice, and it generates a video where the avatar speaks your words with synced lips, gestures, and expressions. It also translates videos into 175+ languages (re-syncing the speaker's lips) and can build a whole video from a single prompt. It's used for training content, product explainers, marketing, and localization. Before you start: - The free plan (no card) gives a handful of short, watermarked videos a month plus one custom avatar — enough to see the quality, not to produce real content. Creator, the first paid plan, turns the watermark off and exports in 1080p. - Before you make an avatar of anyone — including yourself — read the consent section below. It's the thing that separates responsible use from a real problem. - Scripts are where the cost lives: every re-render burns credits, so write and polish the script before you generate. Security: Likeness needs consent — this is the whole ballgame: - A HeyGen avatar is a synthetic version of a real person's face and voice. Only create a digital twin (or clone a voice) of yourself, or of someone who has given you explicit, documented permission. - HeyGen's rules prohibit deepfakes and impersonation — creating a likeness to deceive, or of a public figure without consent, can get your account suspended and may be illegal. - Worth knowing: independent reviewers have flagged that HeyGen's consent checks are lighter than some competitors', which means more of the responsibility to do the right thing falls on you, not the platform's guardrails. Security: - The video translator re-animates a real person's mouth to speak another language. That's still that person's likeness — get their OK before translating and publishing footage of someone else. Security: - On rights: paid users generally own their exports for commercial use (subject to the terms and to consent), and stock avatars/voices are licensed for commercial use — but free-tier output is watermarked and more restricted. Read the terms before running client work or ads. Best at: - Talking-head videos from a script, no camera needed - Localizing one video into many languages with lip-sync - A reusable digital twin so you don't re-record every update - Training, explainer, and marketing content at volume Okay at: - Cinematic realism (good, but not indistinguishable from real footage) - Predictable cost (premium avatars + 4K + translation burn credits fast) - Off-axis or busy shots (front-facing, clean audio gives the best lip-sync) Avoid: - Making anyone's avatar or voice without documented consent - Deepfakes, impersonation, or anything deceptive (prohibited, often illegal) - High-authenticity content where a real human on camera matters Starter: A stock-avatar video from a script The fastest way to judge quality — pick a stock avatar and voice, paste a script, render, watch. No consent questions with a stock avatar. Keep the first script short; long scripts cost more credits and take longer to render. Prompt: Hi everyone! In this quick video I will show you three tips for staying organized at work. Let us jump right in. Starter: Make a digital twin of YOURSELF Record the short sample HeyGen asks for and create a digital twin of your own likeness, then have it deliver a line. Your own avatar is the standout feature and the one you can make with zero consent worry. Record in a quiet, well-lit spot facing the camera for a much better twin. Starter: Translate a clip Upload a short video of yourself and use the translator to produce a lip-synced version in another language. Localization with matched lip movement is HeyGen's other headline trick. Only translate footage of yourself or someone who's agreed to it. Pitfalls: - Consent is the rule, and the guardrails are light. Only make avatars/voices of yourself or with documented permission — HeyGen leans on you to do the right thing. - Credits vanish fast. Premium avatars, translation, and 4K each multiply the cost, and credits mostly don't roll over. - Free output is watermarked. Fine for testing, not for publishing. Where to next: - Video Creation & Editing: https://www.launchpadhq.io/discover?category=video-creation-editing - Voice & Speech: https://www.launchpadhq.io/discover?category=voice-speech ### Getting Started with Higgsfield Canonical: https://www.launchpadhq.io/platform/higgsfield/guide Data: https://www.launchpadhq.io/data/v1/guides/higgsfield.json Last reviewed: 2026-10-15 Tool: Higgsfield (https://www.launchpadhq.io/platform/higgsfield): starting price $19/month; free tier yes; data as of 2026-10-17 One subscription, many video models, and a camera you direct from a menu — with billing terms worth reading before you pay. Pick Image or Video at the top, choose a model (the AI "brain" doing the work) or keep the default, and type a description of the shot, optionally uploading a photo to start from. Set length and shape, check the cost shown on the Generate button, click it, and wait for the clip. Then adjust and try again. What it is: - Higgsfield, made by Higgsfield, Inc., is a studio for making short AI videos and images. It became known for letting you choose how the "camera" moves — a slow push-in, a drone-style sweep — from a menu instead of describing it in words, and that is still the part that sets it apart. - Much of what runs underneath is not Higgsfield's own. Alongside in-house tools such as the Soul image family, Cinema Studio 4.0, Canvas, Genjutsu, Katana and Popcorn, one subscription gives you access to models from other companies, including Kling, Veo 3.1, Sora 2, Seedance, Wan and Nano Banana. Higgsfield's job is the controls, presets and single bill wrapped around them — so two clips from the same description can look very different depending on which model you picked. - Cinema Studio 4.0 is where the camera control lives. Higgsfield's help centre describes a Director's Panel where you set the film look, the camera move (it names options such as POV, Robot Arm and Helicopter Shot), the colour palette and the lighting, then generate. - It also includes an AI assistant, Supercomputer, which Higgsfield says plans and generates whole projects from a plain-language request; it runs on GPT-6 Astra, a third-party model. You do not need it to make a first clip. Before you start: - You can start without paying: Higgsfield has a free tier, described simply as limited use. Paid plans for individuals start with Starter at $19/month, then step up through Plus and Ultra, which cost less per month if you pay for a year; teams pay per seat. Each plan comes with a monthly allowance of credits (the platform's unit of spend — each thing you make costs some), and the cost of a single clip depends on the model, the length and the quality, which is why the exact figure is shown on the Generate button before you commit. - Read the refund rules before your first payment, because they are tighter than most. Higgsfield's terms (updated 26 July 2026) allow a refund only within seven days of your first purchase and only if you have not used a single credit; renewals are not refundable, and its help centre says up to 6% is deducted as a service fee. Subscriptions renew automatically for the same length as the first term, so on an annual plan the next charge is a year's fee, and cancelling stops the next renewal without refunding any part of the current one. If you bought at a promotional price, the help centre says the renewal comes in at the standard price. - Credits do not pile up. Plan credits reset at each renewal and unused ones expire; on annual plans they refresh every 30 days. Extra credit packs last 90 days from purchase. Higgsfield also offers Auto-Refill, which charges your saved card whenever your balance drops below a threshold — it is off unless you switch it on, under Manage Account → Subscription. - You must be at least 18 to hold an account, per Higgsfield's terms. Security: - Higgsfield's privacy policy, updated 26 July 2026, says it may use what you upload and what you generate to train its own and its affiliates' AI models. The only way it describes to stop that is deleting the content or your account, and it says anything already used cannot feasibly be removed from a trained model. Deleted content can sit on its servers for up to 30 days, and copies may remain in backups after that. Security: Before you upload anything:: - Your files may pass to other companies. The policy says data can be shared with 'third-party AI model and compute providers' — expected, since many of the models here belong to someone else — without naming them. - Your data is processed in the US. The policy says personal information 'will necessarily be accessed and processed in the U.S.', and that Higgsfield also has a location in Kazakhstan. - Don't upload someone else's face or voice without their permission. The terms require you to hold 'all necessary rights, releases, and consents' from anyone whose likeness or voice appears in what you upload. Best at: - Directing camera movement by choosing it from a menu — a named move, a lens, a lighting setup — rather than hoping a written description gets it right. - Trying one idea across several companies' video models without signing up to each of them separately. - Short, stylised clips for social posts and ads, which is what most of its presets and studios are built for. Okay at: - Longer pieces: Cinema Studio 4.0 can extend an uploaded clip forwards or backwards, but you are stitching short generations together, not editing on a timeline. - Predicting cost: because each clip is priced separately, a monthly allowance is hard to translate into a number of videos in advance. Avoid: - Paying for a year up front on your first day, given the refund and renewal terms in Before you start. - Counting on an 'unlimited' offer to be fast. Higgsfield's own terms say it may throttle or move to a slower queue any usage that materially exceeds typical individual use, and do not guarantee generation speed; in August 2026 WION reported paying users waiting hours, some days, for a clip on a shared unlimited queue. - Anything you would not want used as training material, since deletion is the only exit the privacy policy offers. Starter: One shot, one camera move The prompt (the message you type) describes only what happens in the scene, not how the camera moves. In Cinema Studio 4.0 you choose the move from the Director's Panel instead — a slow push-in, a POV, a Helicopter Shot — so you can keep the words fixed and change only the camera, which is the control Higgsfield was built around. Generate it twice with two different camera moves and nothing else changed. The difference between the two clips is what the panel is doing. Prompt: A woman sits at a café window on a rainy morning, both hands around a cup of coffee, steam rising. She looks up and smiles at someone off-screen. Soft grey daylight, warm light from inside the café. One continuous shot, no cuts. Starter: Read the price before you click Before your first generation, change the model, the length and the quality one at a time and watch the credit figure on the Generate button move. The Generate button is the only place each clip's price appears. On a balance that expires at every renewal, a minute spent here tells you what your allowance actually buys. Starter: Same words, two models If your plan includes both, run this once on Kling and once on Veo 3.1 — two different companies' models sitting behind the same Higgsfield balance. The scene has water, light and a fall, which is where models disagree most, so you see the gap between them rather than reading about it. Check the Generate button before each run: the same prompt does not cost the same on every model. Prompt: A paper boat floats down a rain-filled gutter past glowing shop signs at night, then tips over a small waterfall into a drain. Starter: Make a character, not a copy of a real person Build a recurring character from an image you generated yourself, then put it in two different scenes. Higgsfield's help centre describes Soul ID training a character from photos so the same face carries across images and video. Given the consent rule in Security, a character you invented is the clean way to learn how it works. Pitfalls: - Using a credit early ends your refund. A single test clip turns a refundable first purchase into a non-refundable one. If you are unsure, explore on the free tier before you pay. - A failed clip usually refunds itself — not always. The help centre says credits for a failed generation are normally returned within minutes, but that some models may not refund them. You can check under Manage Account → Usage. - Its help pages and its marketing describe different versions. Model and studio version numbers change fast, and some of Higgsfield's own guide pages still describe older releases. Trust what the generator shows you over a guide's screenshots. Where to next: - Video Creation & Editing: https://www.launchpadhq.io/discover?category=video-creation-editing - Image generators: https://www.launchpadhq.io/discover?category=image-generation-editing ### Getting Started with HubSpot AI (Breeze) Canonical: https://www.launchpadhq.io/platform/hubspot-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/hubspot-ai.json Last reviewed: 2026-09-09 Tool: HubSpot AI (https://www.launchpadhq.io/platform/hubspot-ai): starting price See pricing; free tier yes; data as of not dated HubSpot's AI sits on top of your customer records rather than beside them — and it's billed per job, not per month. You work inside HubSpot as usual — a list of people, a company record, a half-written email. A star icon in the top bar opens a chat panel beside whatever you are looking at. You type a request there, it answers using the records already in the account, and you accept the result, edit it, or ask again. What it is: - HubSpot is a CRM — the system a sales team keeps its customer records in — and its AI is built into that, not bolted on beside it. Ask it something and it is reading your contacts, your deals and your logged calls, not a blank chat window. - The name has moved twice. The AI was ChatSpot, then Breeze; HubSpot's own AI product page is now headed Agent Hub, while its help centre still calls the underlying AI Breeze. "HubSpot AI" survives mostly in help-centre URLs and in the account setting that governs model training. - What you actually meet day to day is Breeze Assistant: a panel that drafts emails, blog posts and images, summarises a record before a call, researches a company, and builds workflows — with citations back to the record or app an answer came from. - Sitting behind it are agents — AI that takes actions on its own rather than only answering. HubSpot ships named ones (prospecting, customer, data, content) and a builder for making your own. Agents are the part that costs money per job. Before you start: - HubSpot's free plan is a real product rather than a trial: up to two users, 1,000 contacts, no card, no expiry date. Open that account first even if the AI is what you came for — the assistant is close to useless until there is something in the records for it to read. - The AI layer is priced apart from the plan, in credits — the platform's unit of spend. HubSpot's product catalogue lists included monthly credits only from the paid tiers upward, with more on each higher tier from Starter to Professional to Enterprise. They reset monthly and unused ones expire. Starter is often discounted for new customers; current prices are on its Launchpad page. - HubSpot's own pages disagree about whether Breeze Assistant is on the free plan. Its free-tools page lists the assistant among the free features and the assistant's product page says "get started free"; the pricing comparison and the legal product catalogue place it in the paid editions, while the help centre says "all products and plans". Sign up free, look for the star icon, and be ready for Starter to be the honest answer. - Building your own agent is not a free-tier activity. Agent Hub and its agent builder entered public beta in July 2026 for Professional and Enterprise customers only — tiers that cost many times more than Starter, and HubSpot lists a one-time onboarding fee as a condition of purchase. Read the sections below as: assistant now, agents later or never. Security: - Most AI tools only see what you paste into them. This one is already sitting on every customer record, email thread and call note in the account, which means its usefulness and its exposure are the same fact — and the exposure is other people's contact details, not yours. The setting that matters most is on by default: HubSpot states it "may use customer data ... to train and improve HubSpot's own AI models," and a Super Admin turns that off under Settings > Account Management > AI, on the Access tab. Security: Three things to check on that same settings page:: - The data sources are three separate switches — CRM data, customer conversation data, and files. You can also turn generative AI off in the account entirely, so "all of it" and "none of it" are both one toggle away. - Outside model providers are held to a stricter rule than HubSpot itself. HubSpot says it does "not permit HubSpot AI service providers that we engage to provide the Subscription Service to use your data for model training," and describes minimising retention "including zero-day data retention where possible." That bar covers the third parties it sends work to — it is not the same promise as the one about HubSpot's own models above. - Sensitive data is handled separately. Accounts using HubSpot's Sensitive Data feature — for things like government ID and limited financial data — are opted out of AI model training automatically and cannot opt back in. That feature is a paid Smart CRM tier, so on a free account the ordinary toggle is your only control. Best at: - Prepping for a specific conversation — summarising everything logged against one contact, then drafting the follow-up in the same panel. - Writing that has to sound like the rest of your marketing, because it draws on the brand kit, company profile and buyer profiles you set up once in AI settings. - Answering "what is actually happening in my pipeline" without you building a report first. Okay at: - Research on companies that aren't in your records. It will do it, but on a paid tier you are spending credits on something a general chatbot does for nothing. - Reaching your other software. Breeze Assistant can search apps you have installed from HubSpot's marketplace, but the help centre marks that as a beta and does not publish which apps qualify. Avoid: - Judging HubSpot's AI by what the free plan does. With no included credits below Starter, the parts that act on their own are precisely the parts you would not be trying. - Turning an agent loose on a list you have not read. HubSpot moved the prospecting agent to outcome pricing in April 2026, so you pay for each lead it recommends for outreach, and a careless enrolment list is a bill rather than a mistake. - Assuming an agent you built is cheap to leave running. HubSpot's answer to that is a monthly agent run limit, documented as a toggle you switch on per agent rather than something configured for you. Starter: Walk into a call knowing the account When a meeting with a contact is an hour away → then open their record and ask the assistant to summarise every email, call and note logged against it. HubSpot's AI-adoption guide tells you to get teams sending or logging emails in HubSpot and recording calls, and to merge duplicates with the duplicate management tool, before expecting anything from the AI. Running this on day one is the fastest way to find out how thin your records really are. Ask the same question about the company rather than the person — it reads the whole associated history, not just the thread you happen to remember. Starter: Turn a scrappy call note into the follow-up When you have just come off a call and logged a note → then ask the assistant to draft the follow-up email from that note, inside the record it belongs to. The draft lands attached to the customer instead of in a separate window you paste out of, and it inherits the brand kit, company profile and buyer profiles from your AI settings. Save the instruction as a reusable prompt. The assistant also keeps "memories" — your name, role and writing style — so you stop restating them every time. Starter: Describe an automation instead of building one When you notice you do the same three-step chore after every form submission → then describe it to the assistant in plain English and let it assemble the workflow. The automation ends up inside the same system as the records it touches, so nothing is being shuttled between two products and re-authenticated. Note that this is HubSpot's documentation describing the capability — whether the workflow builder is on your plan is one of the things you will see once you are signed in. Before switching anything on, be precise about the event that should set it off. A vague starting condition is what makes an automation fire fifty times in a row. Starter: Spend on exactly one agent job, deliberately When you are on a paid plan and want to know what an agent actually produces → then enrol one contact — one — in the prospecting agent and read what comes back. HubSpot's April 2026 announcement prices this agent per outcome, while its help centre says credits are consumed when a lead is enrolled. Those are not the same moment, and that gap is the entire reason to start with a single contact rather than a list. On Professional or Enterprise, HubSpot documents a Test feature that runs an agent as a simulation — reading CRM context and showing estimated credit usage without consuming credits or touching records. Pitfalls: - Credits do not roll over. Hoarding a quiet month to fund a busy one does not work. - Running out has two different endings. By default, usage-based features pause until your next reset or until you buy a capacity pack of extra credits. Switch on Pay-as-You-Go and they keep running past your limit instead. Know which of those two your account is set to before an agent is doing anything at volume. - The naming will trip up your own research. Most tutorials and screenshots you find will be describing an interface one or two names old. - The beta labels are doing real work here. Agent Hub, the agent builder and connected-app search all carry one — worth knowing before you build a process on top of one, given that the agent pricing itself was rewritten in April 2026. Where to next: - More AI inside the business software you already use: https://www.launchpadhq.io/discover?category=ai-plugins-business-software - Sales, marketing and SEO tools: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai - Automation platforms that connect your apps: https://www.launchpadhq.io/discover?category=workflow-automation ### Getting Started with Hugging Face Canonical: https://www.launchpadhq.io/platform/hugging-face/guide Data: https://www.launchpadhq.io/data/v1/guides/hugging-face.json Last reviewed: 2026-07-27 Tool: Hugging Face (https://www.launchpadhq.io/platform/hugging-face): starting price Free, from $9/mo; free tier yes; data as of not dated The place every open model actually lives — and an orientation to it before you try to use it. You arrive at a model's page — usually because a tutorial or another tool sent you there — read its documentation, and either download the files to use in a different tool or try a small hosted demo in your browser. There is no chat box; you browse, read, and fetch. What it is: - A public repository of openly available AI models and datasets, free to browse and download without an account. - The place other tools point at when they fetch a model — including local runners like Ollama and LM Studio. - A host for small interactive demos, so you can try a model in a browser before installing anything. - Not a chat assistant. If you want something to talk to, this is the wrong page — this is where the components live. Before you start: - This platform is marked Advanced in our catalogue and that is honest. Getting real value from it generally means writing some Python. Browsing and trying demos does not. - A free account costs nothing and needs no card. It covers unrestricted access to public models and datasets, private storage, basic demo hosting, and a small daily allowance of shared GPU time. - The individual paid tier is a low monthly fee and mostly buys a larger GPU allowance and more storage. Team and enterprise tiers are per seat. - The subscription is not the bill. Compute — hosted demos on upgraded hardware, and dedicated model deployments — is billed separately by the hour on top of any plan. - Most beginners should arrive here to fetch a model for a different tool, not to run one here. Getting set up safely (Hugging Face documentation): - Browse without an account first. You do not need to sign up to look at a model page or read its documentation. - When you do create an account, generate an access token only if a tool asks for one, and scope it to read access unless you are uploading. - Before spinning up any paid hardware, find where the pause control is. Knowing how to stop the meter matters more than knowing how to start it. - Read a model's licence before using its output for anything real. "Openly available" and "free to use commercially" are not the same claim. Security: - The risk here is not a vendor training on your conversations — it is that this is a public repository anyone can upload to. Models and datasets arrive from companies, research labs, and anonymous individuals alike, and the platform hosts them rather than vetting each one. Popularity and download counts are social signals, not safety guarantees. Security: What to check before you trust a model: - Who published it — an established organisation, or an account with no history. - Whether the model card actually documents training data and intended use, or is empty boilerplate. - The licence, which governs commercial use and is frequently more restrictive than people assume. - Whether you are downloading something that executes code on load, as opposed to plain weights. Security: - The billing risk deserves equal weight. Idle compute still bills. A dedicated deployment left running charges by the hour regardless of whether it served a single request, and the most common expensive mistake here is forgetting something is switched on rather than using it too much. Best at: - Finding and downloading open models for use in other tools - Trying a model in a browser demo before committing to a local install - Reading documentation about what a model was trained on and intended for - Finding datasets for a specific task Okay at: - Serving as a general-purpose entry point, since it assumes more background than most beginners have - Running things directly, which works but costs real money at any scale Avoid: - Treating any uploaded model as vetted or safe by default - Leaving paid hardware running between sessions - Expecting a chat interface — that is not what this is Starter: Try a model in a browser demo before installing anything Runs a small hosted demo of a model on the platform's own hardware. You find out whether a model does what you need without a download, an install, or a bill. If the demo is slow or queued, that is shared free capacity, not the model's real speed. Starter: Read one model card end to end Opens a model's page and reads its card — the description the publisher wrote — from top to bottom, including the licence section. Model cards are where training data, limitations, and licence live. Learning to read one is the actual skill this platform teaches. Compare a well-documented card against a sparse one. The difference tells you a lot about what you are about to run. Starter: Find the model a tutorial told you to download Types the exact model name from the tutorial into the site's search, opens the page under the matching publisher, and finds its files or download instructions. This is how most people genuinely use the site, and doing it deliberately once removes the mystery. Note the exact model name including the publisher prefix. Similarly-named models from different publishers are not interchangeable. Starter: Check a licence before you rely on an output Finds the licence named on the model's page and reads what it says about commercial use before using anything the model produced. Open weights routinely carry conditions on commercial use, and discovering that after you have shipped is expensive. If the licence is not plainly stated, treat that as a finding rather than an oversight. Pitfalls: - Anyone can upload. This is a public repository, not a curated store. Check the publisher, the model card, and the licence before running anything, and be sceptical of models with no documented provenance. - Idle compute bills anyway. Dedicated deployments and upgraded demo hardware charge by the hour whether or not they are used. Pausing is not the same as scaling down — find the pause control before you start. - The plan price is a fraction of the real cost. Subscription tiers cover your account, not the compute. Anything you actually run is billed separately. - Open does not mean commercially free. Licences vary widely and some prohibit commercial use outright. Read it before you build on it. - It is not an assistant. Arriving here expecting something to chat with means you wanted a different tool. This is the parts department. Where to next: - Run models on your own machine: https://www.launchpadhq.io/discover?category=local-open-source-ai - Hosted APIs instead: https://www.launchpadhq.io/discover?category=ai-apis-developer-services ### Getting Started with Ideogram Canonical: https://www.launchpadhq.io/platform/ideogram/guide Data: https://www.launchpadhq.io/data/v1/guides/ideogram.json Last reviewed: 2026-09-07 Tool: Ideogram (https://www.launchpadhq.io/platform/ideogram): starting price Free, from $20/month; free tier yes; data as of 2026-10-26 The AI image generator that can actually spell — built for posters, logos, and anything with words in it. You type a description of the picture you want — your prompt, the message you give the AI — into a box and press go; a few seconds later a set of images comes back. If one is close but not right, change the wording or edit the image directly and run it again. What it is: - A text-to-image generator: you describe a picture in plain words and it draws one, in your browser, with nothing to install. It's built for things that have words on them — posters, flyers, logo concepts, book covers, social posts, product mockups. - The reason to pick this over Midjourney, DALL·E or Firefly is legible text. Most image generators turn lettering into convincing-looking gibberish; Ideogram is built around spelling words correctly and setting them as real-looking type. If your image needs to say something, start here. Before you start: - Sign in with Google, Apple or Microsoft — not with a plain email address. Ideogram's pricing page states that free weekly credits go to accounts using one of those three; an email-only signup gets none. Both your login method and your username are permanent, so decide before you click. It's otherwise free to start, no card, at ideogram.ai. - You get a pool of credits — the unit of spend, each image costing some — refilling weekly. Free credits are "slow" credits: your job runs in the lower-priority queue, one generation at a time. Ideogram doesn't publish the exact number and says it may vary. - To try it and make the odd graphic, you won't pay. You'd want a paid plan if you generate regularly, or if you need images kept private — private generation is a paid feature. Paid plans run from personal tiers up to larger ones for heavier use; current prices are on its Launchpad page. Security: - On the free plan your images are public by default. Ideogram's FAQ is plain: images are public unless you choose private generation, which requires a paid plan. Public means the image appears on the site next to your handle, where others can find and remix it. You can unpublish afterwards — but keep anything confidential or client-sensitive out of a free-tier prompt. Security: If you upload a reference image: - Ideogram's docs say uploaded images are private and live in your Library, not published. - Anything you generate from that upload still follows the public-by-default rule above. - Its terms give Ideogram a broad licence to store and process your content to run and improve the service, and neither the terms nor the privacy policy say whether prompts and uploads train its models — so don't upload other people's photos or copyrighted artwork. Security: - Commercial use is allowed on every plan, including free. Ideogram says it doesn't claim ownership of your images and you may use them for any purpose, with no separate licence to buy. The risk stays yours: what you make must not infringe someone's trademark or copyright. Best at: - Short text rendered legibly inside an image - Posters, flyers and social graphics - Logo and wordmark concepts - Text on real objects — signage, packaging, book covers, apparel Okay at: - Photorealistic people and scenes - General illustration and painting styles - Matching the look of a reference image you supply - Enlarging an image, or removing and replacing a background Avoid: - Text-heavy layouts — its own docs say it isn't designed for complete documents - Non-Latin alphabets and accented characters, which the docs flag as unreliable - Exact brand fidelity — a real logo, a specific licensed font, an exact hex colour - Anything you need kept private while you're on the free tier Starter: Turn a business name into a logo concept A wordmark is pure spelling, exactly what Ideogram is built for. Its prompting guide says to put the words you want in quotation marks and place them early, and to describe a font's properties — soft and rounded here — rather than name a typeface. Two short words is the least you can ask it to spell, so a wrong letter there tells you something a paragraph of prompt never would. Ask instead for narrow, upright capitals with small strokes on the ends. Same name, different feel. Prompt: A simple flat logo for a coffee shop called "Third Hour", rounded sans-serif wordmark with a coffee bean above the text, black on white, generous empty space Starter: Switch Magic Prompt off when the wording matters Magic Prompt is a setting, on by default, that quietly rewrites and expands your description — often helpful, but it can wander from the words you asked for. Longer text is where Ideogram is most fragile, so run this once with it on and once off, and compare. Prompt: A hand-lettered chalkboard sign outside a bakery that reads "OPEN EARLY, SOLD OUT BY NOON", warm morning light, soft blurred background Starter: Fix a misspelled word instead of starting over Open a generated image in the Editor, paint over just the broken word, and regenerate that patch. This is Ideogram's own recommended fix for text errors, and it matters here: when 90% of a poster is right and one letter is mangled, re-rolling wastes credits and changes the parts you liked. Pitfalls: - Cramming a paragraph into the picture. Spelling accuracy falls off as text gets longer. Keep it to a few words and add long copy afterwards in any editor. - Assuming your images are private. On the free plan they're public by default, next to your username. Check before you generate anything you wouldn't post. - Naming a specific font or brand. "Make it in Helvetica" or "use the Nike logo" won't work reliably and may not be legally yours. Describe the look instead — bold, narrow, curly handwriting. Where to next: - More image generation and editing tools: https://www.launchpadhq.io/discover?category=image-generation-editing - Turn your graphics into marketing: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai ### Getting Started with Image Generation in ChatGPT Canonical: https://www.launchpadhq.io/platform/gpt-image-1-5-dall-e-successor/guide Data: https://www.launchpadhq.io/data/v1/guides/gpt-image-1-5-dall-e-successor.json Last reviewed: 2026-09-13 Tool: GPT Image 1.5 (DALL-E successor) (https://www.launchpadhq.io/platform/gpt-image-1-5-dall-e-successor): starting price $20/mo; free tier yes; data as of not dated Describe a picture in the same box you'd ask a question, and keep talking until it's right. You type a description of a picture into the same message box you'd use for a question, and a few moments later the image appears in the conversation. If something's off, you say so in plain words — smaller, warmer, move the sign left — and a new version appears underneath. You keep going until it's right. What it is: - This is OpenAI's image generator, the one that replaced DALL·E. It doesn't have a separate app or website: it lives inside ChatGPT. - The model (the AI "brain" that does the actual thinking) behind it has been swapped out three times in nine months — GPT Image 1.5 in December, GPT Image 2 in April, and ChatGPT Images 2.5 on 8 September 2026. You never choose between them; ChatGPT serves whichever is current. - It's a good first image generator because there is nothing to learn before you start — no settings, no codes appended to your request, no separate account. The trade is that you get less precise control than tools built only for image work. Before you start: - Ignore the model names. You can't pick "GPT Image 1.5" or ask for "DALL·E" — DALL·E was switched off on 12 May 2026 and GPT Image 1.5 is being retired from OpenAI's developer service on 1 December 2026. Any guide telling you to select one is out of date. - Go to chatgpt.com and sign up free with an email, Google or Apple account. Image generation is included on the free plan, so you can make pictures on day one without a card. ChatGPT isn't offered in every country, and OpenAI warns that getting around that can cost you the account. - You probably won't need to pay to try this, but you'll notice the ceiling quickly: OpenAI describes free image generation as limited and slower, and the cap on images is counted separately from the cap on chat messages — you can run out of pictures while text still works. The paid options are Plus and a cheaper Go plan beneath it, with prices varying by country. - One capability is genuinely behind the paywall rather than just rationed: OpenAI's help pages state that "Images with thinking" — where the model works through a complicated instruction before drawing — is available on Plus, Pro and Business. On the free plan, a request with several fussy requirements in it will come back partly wrong, and rephrasing won't fix it. Security: - This generator's distinguishing privacy fact is not how OpenAI handles your data but what it stamps into the pictures themselves. OpenAI puts two provenance marks on images made in ChatGPT: an invisible watermark (a mark on generated output identifying it as AI-made) and a data label the industry calls Content Credentials. OpenAI's own guidance is honest about the limits of both: the data label "can sometimes be removed by platforms, editing tools, or file conversions," and the invisible watermark degrades under compression, cropping and format changes. So it is evidence, not proof, in either direction. Security: - On the ownership question OpenAI is unusually generous, and the generosity has one string attached. Its Terms of Use, in force since 1 January 2026, say you "own the Output" and assign OpenAI's rights in it to you — but the same terms forbid you to "represent that Output was human-generated when it was not." You can sell the picture; you can't pass it off as hand-made. Security: Worth knowing before you upload a reference photo: - Photos you upload go to OpenAI's servers, and on the personal plans content is used to train its models by default until you turn that off under Settings → Data Controls → "Improve the model for everyone" - OpenAI's page on training doesn't state plainly whether uploaded and generated images are covered by that setting the same way conversations are — assume they are and opt out if it matters - OpenAI's usage policies, effective 29 October 2025, bar using someone's likeness without their consent in ways that could confuse authenticity — which covers most of what people want to do with a friend's photo as a joke Best at: - Getting a usable picture out of a plain-English description - Fixing what's wrong by replying in the same conversation instead of rewriting the request - Signs, posters, packaging and mock-ups where the words have to come out legible - Cut-out assets like logos and stickers, because it can return a genuinely transparent background rather than a white square - Keeping one invented character recognisable across several pictures in the same thread Okay at: - Real typography and page layout — the letters are legible now, but you can't specify a typeface or hold a grid - Exact brand colours and repeatable spec work; ask twice and you'll get two near-misses - Retouching your own photograph — it re-draws the picture rather than editing the pixels you gave it, so faces and hands drift Avoid: - Anything that must not be traceable to AI. Every image carries the two marks described above. - Recognisable living people, especially photorealistically. The refusals you'll hit are not negotiable — you cannot argue the model into it, and repeated attempts are an account-safety question, not a prompting one. Starter: Put the words on the poster Legible text inside the picture was the specific thing OpenAI rebuilt for the April 2026 model and kept improving in September's — which is why quoting the exact wording in your request is worth doing here rather than adding the words afterwards in another program. Prompt: Make a poster for a neighbourhood plant swap. The poster must read exactly: "PLANT SWAP — Saturday 10am — Corner of 4th and Pine — bring a cutting, take a cutting". Warm hand-drawn style, portrait orientation, plenty of empty space at the bottom. Starter: Ask for a cut-out, not a picture of one ChatGPT can return an actually transparent background — OpenAI documents making backgrounds transparent in the image editor and shipped transparent PNG and WebP output to its developer service on 20 August 2026 — so the fig drops onto any colour. Prompt: A single ripe fig with one leaf, flat illustration style, thick clean outlines, on a fully transparent background. No shadow, no ground, no frame. Starter: Keep the same character in shot two Holding a subject steady across a back-and-forth is what the 8 September 2026 update was released for, so the instruction that works here is a plain follow-up in the running thread rather than anything appended to a fresh request. Send this as a reply inside the same conversation, after you've already made a character you like. Prompt: Same character, exactly as-is — same face, same hair, same jacket, same art style. Now show them sitting on a bus at night, looking out of the window. Wide shot. Starter: Start from a template instead of a blank box Open the sidebar, choose Images, then click Templates. Pick a category, pick a template, fill in the details it asks for and send. ChatGPT may come back with a couple of follow-up questions before it draws anything. OpenAI's help pages describe a browsable Templates library inside the Images sidebar, with formats like posters and merchandise already set up — which means the hardest part for a first-timer, writing the request, is handed to you filled in and you edit it rather than invent it. That is the documentation's description rather than a screen we watched, so expect the labels to have moved. Starter: Look at what the file is carrying before you post it Download an image you've made, then run it through OpenAI's public verification page before you put it anywhere that matters. It will tell you whether the two marks OpenAI applies are still intact in that particular file. OpenAI publishes a tool to read those marks back, so this is the one image generator where you can check for yourself what a file discloses about its origin before you hand it to a client. Pitfalls: - Following any guide that tells you to choose DALL·E or GPT Image 1.5 — its advice about how to phrase things is aimed at a generator that has since been replaced twice. - Re-describing an edit instead of pointing at it. OpenAI's help pages describe an editor you open by clicking the image, with a Select tool that highlights only the area you mean and a slider to size the selection; find that first, then describe the change. - Resending because nothing's happening. OpenAI says a picture can take a few minutes. A second attempt doesn't speed anything up; it spends another image from your cap. - Hunting for the sketch tool on a laptop. Drawing a rough layout and having ChatGPT render it is a mobile-app feature — you type "@" in the message box, choose Sketch, draw, then tap the checkmark and describe what it should become. It isn't in the desktop browser, and people conclude it doesn't exist. Where to next: - Other image generators: https://www.launchpadhq.io/discover?category=image-generation-editing - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with invideo AI Canonical: https://www.launchpadhq.io/platform/invideo-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/invideo-ai.json Last reviewed: 2026-09-17 Tool: InVideo AI (https://www.launchpadhq.io/platform/invideo-ai): starting price $10 per seat/mo; free tier yes; data as of 2026-10-14 Describe a video in a sentence, get a finished cut back, then edit it by typing. You open its prompt page and type what you want the video to be — a prompt (the message you type) can be one line, a rough concept, or a finished script. Back comes a full cut with visuals, voiceover and music. Below it sits a second box where you type changes in plain English, and it re-cuts. What it is: - invideo is a video maker you drive by describing what you want, rather than by dragging clips along a timeline — though there is a timeline underneath if you want it. - It writes the script as well, and the footage and music come from licensed stock libraries — iStock, Storyblocks and Shutterstock — though it can also generate shots outright. - During 2026 invideo repositioned itself around "Agent Two", marketed as an agentic video editor for serious creatives, with project memory that carries characters and tone between sessions. The simple on-ramp — describe a video, get a cut — is still there and is still the right way in. Before you start: - Signing up takes an email address, a Google account or an Apple account, and invideo sends a six-digit code instead of asking for a password. Its AI video generator page says "Get started, It's free!"; we could not confirm from any public page whether a card is requested during signup, so expect to check that yourself. - There is a free plan, but you will not find it on the pricing page. That page currently lists only three paid tiers — Starter, Plus and Max. The free plan is documented only in the help centre, which says it gives you limited credits (the platform's unit of spend — each thing you make costs some) that reset weekly, on Monday at 12 a.m. UTC. - The thing to know before you invest an afternoon: invideo's help centre states that on the free plan, downloading without a watermark (a mark stamped on the output showing where it was made) requires a paid plan. Free exports are watermarked. - Credits are spent on *creating*, never on downloading, and different models cost wildly different amounts — invideo's own example runs from about half a credit for a quick clip to 10 or more credits for a 4K clip with generated audio. Unused monthly credits are forfeited at each reset; credits you buy separately as an add-on last 12 months. - Premium iStock footage is a second, separate meter from credits, and it behaves backwards from them: it is counted when you download, not when you generate. Security: - Unlike most video tools, invideo is explicit about training. Its privacy policy (effective 8 June 2026, last updated 8 July 2026) states flatly, "We do not use your User Content, Inputs, Output or Designs to train or fine-tune our proprietary artificial intelligence or machine-learning models." Security: The catch is the models it rents: - The same policy sorts third-party AI providers into two tiers. Standard Tier providers are "contractually prohibited from using your content for their own purposes, including model training." - Experimental Tier providers are defined as those "that have not committed to refraining from training on Customer Personal Data, or do not hold complete independent security certifications." - The policy directs you to invideo's sub-processor list at trust.invideo.io/subprocessors to see which provider is in which tier. It does not say the product labels that tier at the moment you choose a model, and we could not confirm from public pages whether it does — so if a shot matters, check before you generate it. Security: - Delete your account and invideo says content leaves active systems within about 90 days, with backups overwritten on normal rotation. Ordinary marketing footage is fine here; unreleased client material run through a model whose tier you have not checked is the case worth a second thought. Best at: - Turning a written idea, or a script you already have, into a complete first cut — subtitles included - Faceless social video, where licensed stock carries the visuals and the voiceover carries the story - Producing the same video in another language — invideo lists 50+ and names 11 as fully tested: Japanese, German, French, Russian, Spanish, Brazilian Portuguese, Italian, Dutch, Korean, Polish and Hindi Okay at: - Frame-level precision work. The timeline exists, but you are working against what the tool is good at - Subtitles in right-to-left and several Asian scripts — invideo's help centre lists Arabic, Urdu, Hebrew, Chinese (Simplified), Tamil, Marathi, Maori, Farsi, Thai and Hindi as languages where subtitles are not yet supported, which includes one language on its own fully-tested list - Planning a fixed cost per video, because the credit charge moves with the model, the resolution and whether audio is generated Avoid: - Client work where the chain of title on music and footage has to be unimpeachable. invideo's own support article concedes that YouTube sometimes issues incorrect Content ID claims against its royalty-free library, and that the fix is for you to dispute them. - Acting on invideo's support article that says you have "complete rights to all the videos that you make" and can use them "anywhere, anytime and for any purpose you see fit." The terms updated 8 July 2026 are narrower: you own outputs "subject always to these Terms, applicable law and third-party rights," and "Your rights to use Licensed Content are limited to the scope of those third-party licences." Where the two disagree, the terms are the binding document. - Video whose script has to be factually right on the first pass — the agent writes it from your description and will fill the gaps it does not know with plausible invention. Starter: Your first video, in one sentence The prompt page runs Agent mode by default and takes a one-line concept, a loose treatment or a finished script in the same field. Swap the topic for anything you would otherwise explain in a blog post. Prompt: Make a 60-second video explaining why sourdough bread takes so long to rise. Friendly, curious tone, aimed at someone who has never baked. Use real footage where you can, and end on a single practical tip. Starter: Steer the stock before it costs you iStock is the one capped library, and invideo documents a settings: directive inside the prompt that caps or removes it; anything removed is replaced automatically from its Storyblocks and Shutterstock libraries, which are uncapped. settings: max 3 iStock media caps it rather than cutting it. Prompt: settings: no iStock media Starter: Edit the finished cut by typing at it Magic Box sits under the rendered video rather than in a project panel, and it accepts media, audio, text and tone commands in the same sentence — including scene numbers. "Remove iStock media" and "Replace all iStock media" work here too, after the fact. Prompt: Change the narrator voice to an Australian accent, make scene #4 funnier, and increase the background music to 80%. Starter: Make the same video in another language Brazilian Portuguese is one of the 11 fully tested languages, and the language request goes in the prompt itself, so translation is a re-run of the same job rather than a separate export-and-dub pipeline. Check the subtitle list above first — several supported languages do not yet get subtitles. Prompt: Make this video in Brazilian Portuguese, keeping the same visuals and pacing. Starter: Price the generation before you run it Enter your prompt and, before confirming, read the credit figure invideo displays alongside the generate control; open the dropdown and choose "Learn more" for a per-model breakdown, then pick a cheaper model if the job does not need the expensive one. The credit cost is shown before the generation runs rather than billed after it, and the per-model gap is wide enough to change which model you pick. Pitfalls: - Parts of the help centre are out of date with the product. Several articles still talk about running out of "video minutes" while the pricing page and the newer articles meter credits. Where they disagree, believe the pricing page and the in-product credit figure. - Every "try again" spends credits; exporting does not. The expensive habit is regenerating, not downloading. - Re-exporting the same video with the same stock costs nothing extra — but using the same clip again in a *different* video counts against your allowance a second time. - It writes the script unless you give it one. If names, dates or numbers matter, paste your own script into the prompt rather than describing the topic and hoping. Where to next: - Voice and speech tools: https://www.launchpadhq.io/discover?category=voice-speech - Image generators: https://www.launchpadhq.io/discover?category=image-generation-editing - Other video tools: https://www.launchpadhq.io/discover?category=video-creation-editing ### Getting Started with Inworld AI Canonical: https://www.launchpadhq.io/platform/inworld-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/inworld-ai.json Last reviewed: 2026-09-12 Tool: Inworld AI (https://www.launchpadhq.io/platform/inworld-ai): starting price Free, from $25/month; free tier yes; data as of 2026-10-25 Type a line, hear it spoken — and invent the voice that speaks it by describing it in a sentence. You open Inworld's voice playground in a web browser, type or paste the words you want spoken, then either pick a ready-made voice or describe the voice you want in a sentence. Press play and the audio comes back in seconds. You reword the line, adjust the description, and generate again until it sounds right. What it is: - Inworld AI turns typed words into spoken audio, and it is good at the part most tools are weakest at: sounding like someone talking rather than someone reading aloud. You write a line, pick or invent a voice, and get back speech for a video, a game or a podcast intro. - If you have read about Inworld before, you probably read about something else. For years it sold a point-and-click studio for building talking game characters, and most of the articles and video tutorials still online describe that. The company now presents itself on its own homepage as "a research lab and inference provider for realtime AI at consumer scale" — voices, transcription and model routing, sold to developers. The old studio.inworld.ai address redirects to the new developer portal, and the old character guides redirect to a login screen. We found no vendor announcement retiring the character product, so what we can actually show you is the redirect rather than a shutdown notice — but treat anything you find about building Inworld characters as out of date. - What is left that a non-programmer can actually use is the voice playground. Everything else on the platform is an API (a way for programs to talk to each other without a person clicking), and needs someone who writes code. Before you start: - The barrier here is not skill, it is orientation: you are signing into a developer portal, and most of what you see is built for programmers. Go to inworld.ai, sign up free, and head straight for the TTS Playground — "TTS" is text-to-speech, the part that turns typed words into spoken audio. That one page is effectively the whole no-code product. Inworld says you can try it instantly, but we could not see whether a sign-in wall stands in front of it, so expect to make the account first. - The free plan is called On-Demand and costs nothing to start. Inworld's plan-comparison table currently lists it as including a limited amount of speech, a cap on how many requests can run at once, and community-only support. That speech allowance reads as a one-off welcome allowance rather than a monthly refill — the table says nothing about it renewing, so plan on it not coming back. - Past that allowance you move onto credits (the platform's unit of spend — each thing you make costs some), a dollar balance you top up by card. Subscriptions start at $25/month and mostly buy a discount on the per-use rate rather than a different product, so paying is a volume decision rather than a features one. - Inworld's terms of service, in force since 11 June 2025, require you to be at least 18. - If you intend to copy a real person's voice, sort the permission out first — the portal asks you to confirm you hold the rights before it will proceed. Security: - Inworld's terms say it will not train its generally available models on material that is not publicly available, which is better than the default at several free AI services. What it does do is keep your content: its documentation states that a setting called Zero Data Retention "stops us retaining content from that point forward" — and that the setting is available only on Enterprise plans, with the controls visible but disabled on everything below. Security: Because what you are handling here is a person's voice, two things follow:: - Only copy a voice you have permission to copy. Inworld's acceptable use policy, last updated 11 June 2025, prohibits creating or expanding "facial, voice, or other biometric recognition databases without consent" and prohibits "impersonating another individual or organization without consent or legal right." The portal's confirmation box is you making a promise, not the system checking one. - Whatever you paste in to be spoken is content sitting on their servers under the retention point above. A script with a real customer's name or an unreleased announcement in it is a bad first test. Security: - The good news on rights runs the other way. Inworld's terms assign you "all of our right, title, and interest in Outputs," and the free plan's own column on the pricing table lists a commercial licence — audio you generate on the free tier is yours, including for paid work, which is not true of every free AI voice tool. Those terms also incorporate service-specific terms that we did not read. Best at: - Speech that sounds performed rather than read out - Short spoken pieces — a video voiceover, a line of game dialogue, a podcast intro - Ending up with a voice that is yours rather than one picked off a shared list - Working outside English Okay at: - Long-form narration — the free allowance is a trial, not a workspace, so an audiobook is a paid project from day one - Explaining itself to a newcomer — the homepage, the documentation and the pricing table are all written for engineers Avoid: - Building AI game characters — the product most Inworld write-ups still describe, and no longer a path a non-programmer can take here - Any use where the listener must not realise they are hearing a machine. Inworld's acceptable use policy requires you to "clearly and prominently disclose to users they are interacting with AI rather than a human" unless it is obvious from the context. Starter: Type the performance into the line itself Inworld's TTS page lists six bracketed cues — [laugh], [breathe], [clear throat], [sigh], [cough], [yawn] — as "first-class" in its 2.0 voice, so you write them into the words rather than hunting for a settings panel. That vocabulary is Inworld's own, published on its own product page, and a fixed documented list is easier to learn than open-ended stage directions. Move the cue somewhere else in the line and generate again — the cues attach to the position you put them in. Prompt: [sigh] Fine. We'll do it your way. [laugh] But when this goes wrong, I want it on the record that I said so. Starter: Describe a voice instead of picking one Inworld's documentation describes a Voice Design flow — TTS Playground, then Create Voice, then Design — where you type a description like this alongside a short script and get back up to three voices to compare. Its own worked example is "A middle-aged male voice with a clear British accent speaking at a steady pace and with a neutral tone," which tells you the register it expects: physical attributes and delivery, not mood words. The documentation describes a Structured mode next to the freeform box, where you edit the voice's attributes directly — worth trying if your descriptions keep missing. We could not confirm what either screen looks like without an account. Prompt: A woman in her late twenties, warm and slightly breathy, with a light Irish accent, speaking quickly like she is telling a friend something she has just found out. Starter: Copy a voice from a short recording Upload a clean recording of someone speaking and Inworld builds a copy of that voice you can then type for. Its documentation puts this "instant" version at as little as 3 seconds of audio and no more than 30, and says a longer sample within that window gets closer to the original; the TTS product page quotes 5 to 15 seconds instead, so treat the safe answer as "a short, clean clip near the top of that range." The free On-Demand column on Inworld's own comparison table includes voice cloning, voice design and a ceiling of 100 custom voices. What is held back is "professional" cloning, trained on 30-plus minutes of audio and marked as an add-on further up the plans — test that against ElevenLabs and the free-tier line falls elsewhere. Record somewhere quiet — the sample is only seconds long, so background noise is a large share of what the system has to go on. Starter: Say the same line in another language Inworld's TTS page states that "Realtime TTS supports 200+ languages," and a vendor piece published 28 May 2026 makes the stronger claim that "one cloned voice can be localized across more than 200 languages." That same page attaches its own caution: this is "not a quality guarantee," and "every launch language needs native-speaker review using the product's actual scripts." Paste the English line, then a translation of it, and listen to both. Judge the second on whether a native speaker would wince. Prompt: Thanks so much for coming — we'll start in about five minutes, so grab a coffee while you can. Starter: Know where the no-code part stops Beyond the voice playground, Inworld sells three things a programmer wires into software: a routing layer reaching 220-plus text AIs from providers including OpenAI, Anthropic, Google, Groq, Mistral and xAI through a single connection; a realtime service that holds a spoken back-and-forth conversation; and game-engine kits for Unreal and Unity, the Unity one announced as early access in October 2025. This is the line that decides whether Inworld is a tool for you or a supplier to whoever builds you something. Convai, the name people reach for when comparing, still sells a character-building product aimed at people who do not code; Inworld's equivalent moved to the far side of it. Pitfalls: - Landing on inworld.ai and bouncing straight off. The homepage sells infrastructure and its two main buttons are "Get started" and "Contact Sales"; what you want is the TTS Playground inside the portal. - Assuming the free minutes come back next month. The comparison table lists them as included with the plan rather than as a recurring credit, so budget them as a one-off. - Reading the documentation as a map of what you can click. Large parts of Inworld's docs now redirect to a login — its release notes among them — so "the docs describe this" and "you will see this button" are two different claims. - Iterating on a long script. Every regeneration spends from the same small allowance, so settle the voice on one or two sentences first. - Trusting a price you read in an announcement. Inworld publicised a more-than-50% price cut in June 2026 with per-character figures that do not match the on-demand rates on its own current pricing table. Check the table, not the news. Where to next: - Voice & Speech: https://www.launchpadhq.io/discover?category=voice-speech - Gaming & Creative AI: https://www.launchpadhq.io/discover?category=gaming-creative-ai - AI APIs & Developer Services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services ### Getting Started with Jan Canonical: https://www.launchpadhq.io/platform/jan/guide Data: https://www.launchpadhq.io/data/v1/guides/jan.json Last reviewed: 2026-09-07 Tool: Jan (https://www.launchpadhq.io/platform/jan): starting price Free; free tier yes; data as of not dated A free AI chat app that runs on your own computer. Your conversations never leave it. You install Jan like any other desktop app, download a model from inside it — the AI "brain" that does the thinking — then chat in a window like any other assistant. Answers come from your own machine, so nothing you type is sent anywhere. What it is: - A desktop app for Mac, Windows and Linux that runs AI models on your own computer. Its code is public under the Apache 2.0 licence, so anyone can inspect it and it is genuinely free, including for work. - Ollama is built mainly as an engine other tools plug into, handing you plumbing rather than a place to chat; LM Studio is polished but closed. All three run the same engine and model files, so you are choosing an interface and a licence, not a smarter AI. Before you start: - Your hardware is the limit, not your ambition. With a hosted service you rent someone else's giant computer; here the ceiling is the machine on your desk. - RAM — your computer's short-term memory, what runs out when too much is open — is the binding constraint. Jan's documentation puts roughly 8GB behind small models, around 16GB behind mid-sized ones and about 32GB behind the largest, plus 10GB of free disk. Check yours in About This Mac, or Windows Settings → System → About. - No account, no subscription, no internet needed once a model is downloaded. What you give up is capability: a laptop model is slower and gets more wrong than a hosted assistant. Getting set up safely (Jan's official download page): - Download only from jan.ai or Jan's official GitHub releases. You are installing an application that can touch anything on your machine, and "free AI chat download" results are where tampered installers live. - Measure the machine against those figures first; processors older than about 2013 will not run it. Model files land in Jan's data folder — Mac ~/Library/Application Support/Jan/data, Windows %APPDATA%/Jan/data, Linux ~/.local/share/Jan/data — movable in Settings and worth watching; models you tried once accumulate silently. - Prove it before you trust it: ask the starter model something simple, then switch off wi-fi and ask again. An answer with the internet off proves it runs on your machine. Security: - The usual warning runs backwards. Jan is the private option: your conversations are not sent to a company, not stored on a server, not used to train anything. You lose that in three ways, none of them ordinary use. Security: The three ways to give that privacy away: - Installing from an unofficial mirror or a "cracked" build; a tampered installer defeats everything else here. - Pointing Jan at a hosted model from another company. The interface looks identical, but those chats go to that company under their policies. Check the model selector before pasting anything sensitive. - Opening Jan's local server to your network — from "only this computer" to "anything on my wi-fi." Jan's docs flag it use-with-caution. Security: - Your chats sit on disk as ordinary readable files, with no encryption Jan's documentation mentions. Device security is now the whole of your data security — anyone who opens your unlocked laptop reads every one. Use a login password and disk encryption. Best at: - Private drafting you'd never paste into a cloud service - Working with no internet at all - Machines with roughly 16GB of RAM or more, where it feels genuinely pleasant Okay at: - Everyday questions and rewriting on about 8GB — slower and noticeably less sharp - Light coding help with a code-focused model Avoid: - Machines under roughly 8GB of RAM, or processors older than about 2013 — it won't run well, or at all - Anyone who just wants the smartest answer — a hosted frontier model is better Starter: Start with whatever Jan downloads for you Jan fetches a default model on first launch, so you get a reply before choosing anything. Feel the speed on *your* machine before browsing the library. Ask it again with wi-fi off, to see the offline promise for yourself. Prompt: Explain what you are and what you can help me with, in plain language. Starter: Read the Fits / May be slow / Won't fit labels first Opens Jan's Hub — its model library — and reads the label beside each model before downloading anything. Jan's Hub checks each model against your actual hardware and labels it, removing the guesswork that makes people give up. Treat "May be slow" as "no" for your first week; "Won't fit" everywhere is your hardware answering honestly. Take the recommended file size too. Models come in versions squeezed to fit less memory at some cost to quality; if one feels sluggish, drop to the smaller version of the same model. Starter: Do a private task you wouldn't paste into the cloud Jan documents its own roughly 4-billion-parameter model — parameters being a rough measure of size, more being smarter but heavier — as running in about 8GB, in reach on an ordinary laptop. Privacy only pays off on real material. Prompt: Here are my rough notes from a difficult conversation at work. Organise them into what happened, how I feel about it, and what I want to do next. Pitfalls: - Downloading a model too big for your machine. Bigger looks better, then Jan crawls or refuses to load it. Believe the "Fits" label over your ambition. - Judging it from the first sixty seconds. The first reply is slow while the model loads, and your first pick may be the wrong size. Give it one deliberate second attempt. Where to next: - More local AI tools: https://www.launchpadhq.io/discover?category=local-open-source-ai - Cloud assistants to compare against: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Jasper Canonical: https://www.launchpadhq.io/platform/jasper/guide Data: https://www.launchpadhq.io/data/v1/guides/jasper.json Last reviewed: 2026-09-17 Tool: Jasper (https://www.launchpadhq.io/platform/jasper): starting price $59/month billed yearly or $69/month billed monthly; free tier no; data as of 2026-10-14 A marketing-team writing platform that learns your brand's voice — and costs money from day one. You open a project, type a prompt (the message you type) into the chat panel at the bottom, and Jasper writes the piece into a document beside it. You then highlight any sentence and ask for a rewrite, a different length, or a new angle, and it replaces the text in place. What it is: - Jasper is a writing and content platform for marketing teams. You get a chat box and a document editor, like most AI writing tools; the difference is stored brand settings Jasper applies to everything it writes, so you are not re-explaining your company in every conversation. - It does not build its own AI. Jasper's own architecture page says it routes each task to outside systems from OpenAI, Google, Anthropic and Meta and "layers in your brand, context, and guardrails" on top. You are paying for the brand layer and the workflow, not a smarter underlying AI. - It has changed shape. Jasper started as a copywriting tool for individuals and freelancers; today its homepage headline is "Put AI agents to work for marketing" and it calls itself "the agent workspace built for modern marketing teams." Older reviews describing cheap word-count plans describe a product that no longer exists. - Agents (AI that takes actions on its own rather than only answering) are the current centre of the product — pre-built ones for common marketing jobs on Pro, and more complex ones plus a no-code builder reserved for the quote-only Business plan. Before you start: - There is no free tier. The only self-serve plan is Pro ($59/month billed yearly or $69/month billed monthly), for one login. Above that is the Business plan, quote-only — Jasper publishes no price, so if Pro does not cover you your next step is a sales conversation, not a checkout page. - The 7-day trial is not a look-around. Jasper's help centre is explicit: "We do require a credit card during our trial sign-ups right now," and "Once the 7-day free trial has ended, you will be moved to and charged for the plan you selected when starting your trial." Cancelling during the trial is self-serve in the app, per the same page. - Refunds are documented two ways by Jasper itself. The terms (section 5b) say flatly: "You may cancel your Customer Account at any time; however, there are no refunds for cancellation." The help centre says you can get one by emailing within seven days of the charge. The terms bind; treat the help centre route as a courtesy you may or may not get. - The annual price is a year-long commitment billed upfront, not a discount you can walk away from: Jasper's help centre says annual plans "are one-year commitments and cannot be canceled at any time within the year," with a 7-day window at the start and no refunds for unused time after. Start monthly. - Jasper works in around 34 languages at the workspace level, but its help centre says the Content Translator Agent "is limited to Business plans." If translation is the reason you are here, Pro is probably not enough. Security: - What matters about Jasper is not how it stores your work but what its terms let it do with it. Its terms of service (version 4.1, effective 30 May 2025) say in section 3(a): "We claim no ownership rights over Customer Property. The Customer Property remains yours." But the same section has you grant Jasper a "royalty-free, sublicensable, transferable, perpetual, irrevocable, non-exclusive, worldwide license" to use what you put in and what comes out, including for "modifying, improving, and enhancing artificial intelligence models." Security: - Section 7 states it plainly: "Jasper uses Customer Property, including Input and Output, to provide, maintain, and improve the Services, including for the purpose of developing and enhancing Jasper's artificial intelligence models." An objection right exists, but section 12 limits it to people "located in the European Union or United Kingdom," exercised by emailing [email address on the web page]. Elsewhere, the terms as written offer no switch. Security: - Jasper's ethics page says something that sounds like the opposite: "Data and intellectual property entered into Jasper is NOT used to train underlying LLMs." Both can be literally true — "underlying" likely means the outside systems Jasper routes to, while section 7 covers Jasper's own — but a reader taking the ethics page at face value would conclude something the terms do not say. The terms bind; ask Jasper in writing before you paste anything sensitive. Security: Keep out of it, on any plan: - Client material covered by a confidentiality agreement you signed — the licence in section 3(a) is broader than most such agreements allow - Unannounced products, pricing or campaign plans you cannot afford to have leak - Customer lists or anything with real names and contact details in it Security: - Jasper's security page states data is encrypted in transit and at rest, and that it holds an independent security audit report customers can request. Data is kept for the life of your agreement and deleted after termination on written request. Best at: - Producing on-brand copy repeatedly once you have taught it your voice — the second and tenth pieces are where it earns the money, not the first - Turning one finished piece into the smaller pieces a campaign needs: posts, emails, captions, ad variants - Keeping a team's output consistent when several people are writing Okay at: - A one-off piece of writing. It will do it well, but the monthly fee buys a setup step you would be skipping - Research and current information — it is a writing surface first - Images. Generation and editing are included on Pro, but this is not why anyone chooses Jasper Avoid: - Enforcing a written house style guide on the plan you can actually buy: Jasper's help centre says "Style Guide is available on Business plans only" - Buying it as a general assistant. It is priced against marketing output for a single login, not against answering your everyday questions - Connecting it to your own systems on Pro — the comparison table puts API access on Business only - Publishing anything factual it wrote without checking it. A well-tuned brand voice makes a wrong claim sound more authoritative, not less Starter: Teach it your voice before you write anything In Jasper IQ, choose Add Brand Voice and give it up to eight examples of your existing writing — upload .txt, .pdf or .docx, paste text, or hand it URLs for Jasper to crawl. Do this on day one of the trial. Everything Jasper charges a premium for runs through this object: once saved, the voice applies automatically in chat, in the document editor and inside agents. Tested against ChatGPT's custom instructions, which take a typed description of your tone rather than eight documents and crawled URLs that agents inherit. Tested against Copy.ai, which also has brand voice: Jasper's is capped at two on Pro, unlimited on Business — a decision you make before you subscribe. Feed it your best-performing pieces, not your most recent ones. Starter: Make it show its work on the voice Jasper applies the brand voice silently — nothing in the interface tells you whether it did anything, so this is the cheapest way to find out during a trial whether you are getting a branded draft or a generic one. Tested against ChatGPT and Copy.ai: in ChatGPT there is no stored voice object to audit, so the question has no referent. Prompt: Write three LinkedIn posts announcing [product or feature]. Under each post, list the specific words, phrasings or structural habits you took from my brand voice. If you could not find any, say so instead of guessing. Starter: Turn one piece into a campaign in a single instruction Jasper's Canvas documentation says chat can "update one or more assets in bulk" — the output lands as separate, individually editable documents on the project board rather than one block you cut apart. Tested against ChatGPT, which returns eighteen pieces stacked in one reply: the board holds each piece as its own asset you can revise without re-running the request. Add "and flag which three you think are strongest" to get a starting point rather than a pile. Prompt: Using the blog post in this project, create separate assets for: five LinkedIn posts, three emails for a nurture sequence, and ten short social captions. Keep each in my brand voice and do not repeat the same opening line twice. Starter: Edit without leaving the page Highlight a paragraph, choose Ask Jasper Anything, and paste that in. Jasper's documentation describes the highlight-and-ask toolbar as rewriting the selected text in place, with a compose shortcut (Cmd+J on Mac, Ctrl+J on Windows) that continues a sentence you have started. Tested against ChatGPT: there you paste into a chat, get a new copy back, and paste it into your document. Prompt: Cut this by half, keep the specific numbers, and make the first sentence the one a reader would stop scrolling for. Starter: Set the cancellation clock before you set anything else The moment the trial starts, put a reminder in your calendar for day six; cancellation happens in the app under settings. Then decide what you are testing — a brand voice plus two real pieces through it is a fair trial; poking at features is not. The card is taken upfront and Jasper's own two documents disagree about refunds (see Before you start), so the deadline is the first thing to set rather than the last. Tested against ChatGPT and Copy.ai: both have a genuinely free tier you can sit on indefinitely, so neither creates a deadline on the day you sign up. Pitfalls: - The plagiarism checker's tick on Pro does not mean free scans. The pricing table shows it included, but the help centre says "Plagiarism checks are an optional add-on to any plan" and tells you to add credits (the platform's unit of spend — each thing you make costs some) before running one. It is powered by an outside service, Copyleaks. - Jasper documents its own plan gating inconsistently, so check in the app rather than trusting a page: the Jasper IQ overview article says IQ "is available on Business plans," while the pricing table and the Brand Voice article both give Pro two brand voices, five knowledge assets and three audiences. - The terms of service still describe themselves as applying to "our Creator and Pro plans" — and Creator is no longer sold. The legal documents lag the product. - The brand voice is only as good as what you feed it: eight samples of flat, hedging copy produce a machine that writes flat, hedging copy faster, in volume, with your logo on it. - The terms say cancellation notice "will be effective in the month following the month you provide us notice," so a cancellation late in a billing cycle can still carry a charge. Cancel early in the month. Where to next: - Sales & marketing AI tools: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai - General AI chatbots (including free ones): https://www.launchpadhq.io/discover?category=text-conversational-ai - Workflow automation: https://www.launchpadhq.io/discover?category=workflow-automation ### Getting Started with Jetpack AI Canonical: https://www.launchpadhq.io/platform/wordpress-ai-jetpack-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/wordpress-ai-jetpack-ai.json Last reviewed: 2026-09-09 Tool: WordPress AI (Jetpack AI) (https://www.launchpadhq.io/platform/wordpress-ai-jetpack-ai): starting price See pricing; free tier yes; data as of not dated Draft, rewrite and illustrate posts without leaving the WordPress editor. You write in your WordPress post editor as usual. When you want help, you open the AI panel in the sidebar, type what you want in plain English, and the draft appears in the post itself. You accept it, ask for another try, or edit it by hand, then keep writing. What it is: - Jetpack AI is a set of writing and picture-making tools that Automattic builds into the WordPress editor, so the work happens inside the post you are already writing rather than in a separate chat window. - "WordPress" means two different things, and which one you have decides how you get this. Sites hosted on WordPress.com have Jetpack built in already; WordPress.com's own support pages currently say Jetpack AI is included with all its paid plans. Sites you run yourself — self-hosted WordPress, meaning you run it on your own server rather than the company's — need the free Jetpack plugin (an add-on you install into a WordPress site) installed and connected first. - Automattic sells it as its own small subscription and also folds it into the Jetpack Complete bundle. Before you start: - You need a WordPress site you can edit as an administrator, and you need to be writing in WordPress's standard post editor — the one that builds a post out of stacked blocks. Jetpack's docs list that editor as a requirement. - On a self-hosted site you install Jetpack and connect it to a WordPress.com account. Automattic's documentation is explicit that Jetpack AI only works once Jetpack is properly connected to WordPress.com, so the account is not optional even though the site is yours. - The free allowance is smaller than it sounds, and it does not come back. Jetpack's support docs describe the free requests as a small per-site allowance that does not renew. A request is one action — a rewrite, a generated image. The allowance is an evening of experimenting, not a month of writing. - After that the standalone AI plan is a paid subscription, billed yearly and usually discounted for the first year; current prices are on its Launchpad page. Getting set up safely (Install Jetpack from inside your own dashboard at `Plugins → Add New`, or from `jetpack.com` itself. Automattic publishes the install steps; take them from there rather than from a zip file offered on a blog post or a theme marketplace.): - Connecting Jetpack links your site to Automattic's servers, and that link is the product rather than an optional extra — the AI runs there, not on your web host. If you are not willing to make that connection, this tool is not available to you at all. - Try it first on a draft nobody is reading. Jetpack AI writes directly into the post you have open, so the safety net is WordPress's own revision history rather than an undo button in the AI panel. Security: - Two separate things leave your site here. Connecting Jetpack at all copies site data to Automattic's servers: its documentation lists posts, pages, comments, categories and tags, all core WordPress and Jetpack settings, and registered users' IDs, usernames, email addresses and roles — passwords excluded — and says it copies what all of its features need whether you have switched those features on or not. Security: Then, separately, what the AI itself sees:: - What you type into Jetpack AI is passed on to an outside AI company. Automattic's AI guidelines, last updated 18 August 2026, name Anthropic, OpenAI and Google as its providers and state that each of them does not use that data to train or improve its models. - The same page still advises thinking carefully before entering confidential or sensitive information into any AI feature, even with those no-training statements. - A newer screen at Jetpack → AI in your dashboard can let an outside AI assistant such as Claude or Cursor read and change your site — it is labelled MCP there. Jetpack's docs say that access is off by default, that read and write permissions are set category by category, and that every write action needs your explicit confirmation before it runs. Leave it off until you specifically want it. Best at: - Second-draft work on a post already in front of you — tightening a paragraph, shifting its tone, writing the title and summary once the post actually exists. - Translating a finished post into another language inside the editor. - Putting a featured image on a post that would otherwise ship without one, using wording Jetpack suggests from the post's own content. Okay at: - Long-form drafting from nothing. It will do it, but every attempt spends a request, which is awkward economics for an allowance that never refills. - Illustration with a specific look. Jetpack's docs say images come back as 1024x1024 PNGs by default, and that the generator will not produce copyrighted characters, identifiable public figures, or imitations of an artist's style from 1912 onwards. Avoid: - Counting on the paid plan being unlimited. Automattic advertises "high request capacity" rather than a number, and footnotes it with a fair-usage policy — so a very heavy month is a conversation with Automattic rather than a guaranteed entitlement. - Installing Jetpack purely for the AI on a site where the rest of Jetpack is unwelcome. Automattic's docs route you through Jetpack and Jetpack → My Jetpack to reach the AI; there is no AI-only Jetpack you can install instead. - Assuming the free allowance covers connecting an outside AI assistant to your site. Jetpack's docs say that for self-hosted sites, that access requires a Jetpack Complete or Jetpack AI plan. Starter: When a post is finished but the title is limp → ask for a title and a summary On a post you have already written, open the AI panel in the sidebar and use the title and summary generation, then choose from what comes back. Jetpack AI reads the post from inside the editor, so its suggestions come from your actual saved text. It also spends one request instead of the several a full draft burns, which makes it the sensible first thing to try on the small free allowance. Jetpack's product page describes a one-click optimisation of post title and featured image at publishing time — the same idea with fewer clicks, worth finding before you do this by hand twice. Starter: When a post has no image and you were never going to make one → generate a featured image In the Improve with AI section of the post sidebar, or by choosing Generate Image when you set a featured image, describe the picture — Jetpack offers suggested wording drawn from the post. Jetpack's docs put an exact price on this one action: one successfully generated image costs one request, charged whether or not you end up using it, while failed generations are not charged. It is the only place in this tool where clicking twice has a stated cost, which makes it the action worth understanding before the others. The default is square. Check how your theme crops a square featured image before you spend several requests building a set of them. Starter: When a paragraph is too long to read → rework it where it sits Select an existing paragraph, heading or list and use the AI option on that block to shorten it, expand it, or change its tone. The tool acts on the block you selected and writes back into it. Jetpack's docs list ten tone options on that same menu, which turns "make this less stiff" into a menu choice rather than an instruction you have to phrase well. There is also a clarity tool, Write Brief with AI, that flags long sentences and complex words as you write. Jetpack's docs say using it does not count towards your request usage — the one AI feature here you can simply leave switched on. Starter: When a post deserves a second-language version → translate it in the editor Select the post's content and choose translate from the same AI menu; Jetpack's docs list twelve languages. The translation is written back as ordinary editable content, so you end up with a post you can keep working on rather than a block of text to reformat. Jetpack counts each action as one request, so translating a long post a paragraph at a time is markedly more expensive than doing it in one pass. Automattic's guidelines ask you to review and edit AI output before publishing. For a language you do not read, that means finding somebody who does — not skipping the step. Pitfalls: - More than one thing is now called "WordPress AI". The WordPress project's own AI team publishes a separate community plugin named AI on wordpress.org, and WordPress 7.0 added AI plumbing to core. Neither is this. Jetpack AI is Automattic's commercial product — search for that name, not the generic one. - A site with several writers has a per-person step nobody warns you about. Jetpack's docs say each user who wants the AI tools must connect their own WordPress.com account, so the editor who joins next month meets a sign-in screen you have long since forgotten about. - The introductory price renews at the full one. Automattic's own footnote says the discount applies to the first year only — worth a calendar note on a subscription that bills yearly. - Requests drain from actions you would not think of as AI. Jetpack's own definition counts excerpt generation and the feedback tool alongside the obvious ones, and says the list is not exhaustive, so features you switched on for another reason can quietly spend the allowance. - If the AI panel disappears, check the connection before the plan — a host migration or a domain change is exactly what tends to break it. Where to next: - Other AI plugins and business software: https://www.launchpadhq.io/discover?category=ai-plugins-business-software - Writing and chat assistants: https://www.launchpadhq.io/discover?category=text-conversational-ai - Image generators: https://www.launchpadhq.io/discover?category=image-generation-editing ### Getting Started with Julius AI Canonical: https://www.launchpadhq.io/platform/julius-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/julius-ai.json Last reviewed: 2026-09-15 Tool: Julius AI (https://www.launchpadhq.io/platform/julius-ai): starting price Free, from $20/mo; free tier yes; data as of not dated Upload a spreadsheet, ask questions about it in plain English, and get real calculations back. You drag a spreadsheet into the chat window, then type a question about it the way you'd ask a colleague. Julius works for a moment and answers in the chat — a number, a table, a chart. You read it, ask a follow-up to narrow it down or redraw it, and keep going until it is the answer you wanted. What it is: - Julius is a chat window pointed at your own files — a spreadsheet, an export, a PDF — rather than at the open internet. - What separates it from a general chatbot is that it doesn't estimate your numbers. It writes a short computer program, runs it on your file on Julius's servers, and reports back what the program returned. The tables, charts and statistics come out of an actual calculation. - That makes it a fit for the "I have a file and a question" job — sales by region, which survey answers cluster together, what a trend actually looks like — for people who would otherwise be losing an afternoon to pivot tables. The programs it writes are in Python or R, the two languages analysts use for this work, and you can read them. Before you start: - Free to use after a sign-up at julius.ai. The free plan currently runs on a small daily allowance of credits (the platform's unit of spend — each thing you do costs some), plus a one-off welcome bonus of credits when you join. The daily allowance refills each day and does not accumulate, and when it runs out it simply stops until the next day — Julius does not sell top-up credits, and its billing documentation does not spell out what you see when you hit the end. That is enough to explore a file and answer a few questions, not enough to work through an afternoon of analysis. - The free plan runs only Julius's smallest model (the AI "brain" that does the actual thinking) — Julius 1.2 Lite, which Julius's own model documentation scopes to "quick answers to simple questions." Every larger Julius model and every outside model on the selector is paid, starting with the Plus plan. If your first answers feel shallow, that is the likeliest reason, and it isn't your question. - Your uploaded files do not stay. On a free account they are deleted after one hour of inactivity; on a paid account they last up to seven days. Re-uploading at the start of a session is normal, and nothing here is storage — keep your own copy of anything you care about. - Julius runs each analysis inside a temporary computer it spins up for your chat, with 8GB of working memory on every plan. That covers most spreadsheets, but a file with millions of rows can exhaust it. Julius publishes no row limit and no file-size cap — its FAQ says there is "no limit enforced by us" and points at that memory as the practical ceiling — so there is no number to plan against, only the ceiling. A 32GB "Memory Boost" option exists as a Pro-plan upgrade, and Julius's FAQ separately describes the sandbox as having 32GB, which its own container documentation contradicts; the two vendor pages genuinely disagree, so assume 8GB unless you are paying for the upgrade. - Connecting a real database — Postgres, Snowflake, BigQuery — is a paid feature. Free accounts can upload files directly or pull them from Google Drive, OneDrive and SharePoint. Security: - Julius's stated position is unusually clear for a tool you feed real files to. Its data-security documentation says "Your data is not used to train any models, internally or externally," and its terms say Julius "contractually restricts the AI Providers from using User Generated Content for training or otherwise improving such AI Provider's services for general use." All data is stored and processed in the United States. The privacy policy was last updated 1 July 2026 and the terms 6 March 2026; both govern today. Security: - The same terms then take a broad licence to what you upload. By making content available through the service you grant Julius "a worldwide, irrevocable, perpetual, non-exclusive, transferable, royalty free license, with the right to sublicense, to use, access, view, copy, adapt, modify, distribute, license, sell, transfer" it — and the clause is not worded as applying only to things you choose to share publicly. The no-training commitment and that licence sit awkwardly together, and the terms are the document that binds. Read them yourself before a file matters. Security: Think hard before uploading: - Customer or patient records with real names still attached - Payroll, HR or performance files - Anything you have signed a confidentiality agreement over - Financial exports that identify individual people Security: - On the other side of the ledger: the terms say "You exclusively own all right, title, and interest in and to the AI Content," so the charts, tables and reports you get out are yours. Julius's FAQ says that when you delete data in the app "it is completely erased from our servers," and deleting your account is self-serve from account settings. Best at: - Answering a question about a file you already have — totals, breakdowns, "which of these is the odd one out" - Turning a messy export into a usable table: fixing dates, splitting columns, catching duplicates - Making a chart without deciding on a chart type first — describe what you want to see and let it pick - Statistical work you know the name of but not the syntax — a regression, a correlation matrix, a comparison between two groups Okay at: - PDFs and images as data sources: they're supported, but pulling numbers out of a scanned or awkwardly laid-out document is hit and miss - Long, multi-step analyses on the free plan — it runs out of room before you run out of questions Avoid: - Treating a number as checked just because a program produced it. Julius decides which program to write from your wording — a correct calculation run on the wrong column, or on rows it quietly dropped while tidying, comes back as a confident figure with no error message anywhere. - Building anything that needs to still exist next month. Julius has already sunset its Notebooks and Custom Agents features, and its file storage is designed to expire rather than to archive. It is a tool for a working session, not a system of record. Starter: Get a file in front of it first Click the paperclip in the chat bar, or drag a spreadsheet straight onto the conversation. Julius reads it and tells you what columns it found before you've asked anything. With no file attached you are driving Julius's smallest model as a plain chatbot, which the free tiers built for that job do better — the file is the entire reason to be here. No file to hand? Ask it to invent a realistic sample dataset and analyse that instead. Starter: Make it describe the file before you trust it A misread column costs the same day's credits as a correct one, and the free allowance does not come back until tomorrow — so making Julius state what it thinks your columns hold is the cheapest run of the session, and it catches the misread before it becomes a number in someone's slide. Paste any column it misreads back to it with a one-line explanation of what that column actually holds. Prompt: Before answering anything else: list every column in this file with its data type, tell me how many rows there are, how many blank and how many duplicate values are in each column, and flag anything that looks inconsistent. Don't clean or change anything yet. Starter: The plain question, with the format named Julius's own prompting guidance tells you to name the output format — say "table" and "bar chart" and you get both objects, rather than a paragraph describing what a chart would show. Swap in your own column names; "revenue", "region" and "date" are just placeholders. Prompt: Using the attached file, show me total revenue by region for the last 12 months as a table, sorted highest to lowest. Then show the same thing as a bar chart. Finally, tell me which region changed most compared with the 12 months before that, and by how much. Starter: Read the code it actually ran Open the Code Sandbox panel next to the chat — Julius's documentation describes it as "where Julius runs all code behind your analysis" — and read the handful of lines behind the answer you just got, paying most attention to any row it filtered out. Julius documents a place of its own for the code rather than folding it into the reply, so the one thing the chat cannot tell you — what the program did to your rows before it answered — has somewhere to be read. If a line is opaque, paste it back into the chat and ask what it does and why it chose that approach. Starter: The cleanup job that leaves with you Julius can hand the result back as a real file with a download link in the chat, so a cleanup pays off once instead of being trapped inside a conversation you'll lose at the end of the session. Ask for CSV instead if the file is heading into another program. Prompt: This export is messy. Standardise the date column to YYYY-MM-DD, strip stray spaces and fix inconsistent capitalisation in the name column, and flag rows that look like duplicates rather than deleting them. Then give me the cleaned version as an Excel file to download, and list every change you made. Pitfalls: - Credits drain unevenly. Cost depends on what a request sets off — running code, searching the web, building a deck — so a handful of ambitious asks can clear a free day's allowance while a dozen simple questions wouldn't dent it. - A new chat appears to start a fresh sandbox, so a second conversation won't have your file or anything you built from it. Julius's container documentation implies one temporary computer per conversation without quite saying so, but stay in one chat while you're working on one dataset. - Asking for the conclusion instead of the calculation. "Is this trend real?" gets an opinion. "Compare these two groups with an appropriate statistical test, tell me which test you used, and say how confident the result is" gets something you can check. - Julius's plan table marks presentations, images, videos and website-building as "Limited" on the free plan without ever saying limited to what. You find out by running into it. Where to next: - Document & PDF tools: https://www.launchpadhq.io/discover?category=document-pdf-processing - Chat assistants: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Khanmigo Canonical: https://www.launchpadhq.io/platform/khan-academy-khanmigo/guide Data: https://www.launchpadhq.io/data/v1/guides/khan-academy-khanmigo.json Last reviewed: 2026-09-09 Tool: Khan Academy (Khanmigo) (https://www.launchpadhq.io/platform/khan-academy-khanmigo): starting price Free, from See pricing; free tier yes; data as of not dated Khan Academy's AI tutor, built to ask you questions instead of handing over the answer. You sign in on Khan Academy's site and open a chat panel next to the lesson you're working on. You type a question — your prompt, the message you type — and it usually replies with a question of its own rather than the answer, so you work through the problem in turns. What it is: - Khanmigo is the AI tutor and teaching assistant built by Khan Academy, the non-profit behind the free lesson library that millions of students already use for maths, science, coding and humanities. - Its defining design choice is a refusal. Khan Academy's own description is that Khanmigo "doesn't just give answers" and instead "guides learners to find the answer themselves" — so a student who pastes in a homework question gets asked what they have tried, rather than a worked solution. - That makes it deliberately slower than a general chatbot, and it is the whole reason to pick it over one. The teacher side of the product is a different thing entirely: lesson plans, practice questions, grading guides and quick classroom checks drawn against Khan Academy's own curriculum, given away free. Before you start: - If you teach, it costs nothing. Khan Academy gives the teacher tools away free to educators aged 18 or over, with no card and no trial clock. A December 2025 update to its own announcement post puts that offer in more than 180 countries and territories, in English plus a set of experimental languages — a much wider reach than the US-only pilot it started as, so if you looked at this a year or two ago, look again. Khan Academy's own pages don't all say the same thing about which countries qualify, so check the list its teacher page links to rather than assuming yours is on it. - For a learner or a parent it is a low-cost paid subscription, monthly or yearly; current prices are on its Launchpad page. Khan Academy is a donation-funded non-profit and says plainly that the payment supports its mission. The ordinary Khan Academy lesson library stays free whether you subscribe or not — Khanmigo is a layer alongside it, not a change to it. - Those paid plans are US-only. You must be 18 or over with a billing address in one of the 50 states or the District of Columbia. Anyone under 18 cannot hold a Khanmigo subscription at all — a parent or guardian subscribes and adds them, up to 10 children on one account. If you are outside the US and you are not a teacher, there is currently no individual way in. - A student cannot get the tutor through school unless the school has bought it. Khan Academy's FAQ says classroom access is "only available through school or district implementations", priced per student per year. Some US states have funded that centrally — New Hampshire's education department has published free statewide grades 5-12 access, though its post carries dates that may since have moved — so it is worth asking your school before paying yourself. Security: - This is a tool built for children, and the honest summary is that it is designed to be watched. That is a feature rather than an oversight, but a student should know it before typing. Security: What an adult on the account can see: - A parent or guardian gets the full transcript of their child's Khanmigo conversations from the parent dashboard — not summaries, the chats. - If Khan Academy's moderation system flags a message as inappropriate, harmful or unsafe, it automatically emails an adult connected to the child's account with the details. - In a school or district deployment, administrators also see progress data and Khanmigo chat summaries. Khan Academy says children are told their history is visible to parents or guardians and, where relevant, their teacher and school administrator. Security: - On whether student conversations train AI, Khan Academy's clearest published statement covers its districts programme: it says it does not allow the companies behind the AI models — the AI "brains" doing the thinking — to train on student data, and that names and personal information are not shared with those providers. That page is written about school partnerships. We could not find an equally explicit statement covering individual paid learner and parent accounts, so read the district wording as the strongest thing Khan Academy has on record rather than as a blanket promise. Security: How long things stick around, in Khan Academy's own words: - Khanmigo conversations may stay visible in chat history for up to 365 days after the conversation ends; inactive ones may then be deleted. - Conversations that were flagged, or that you gave feedback on, can be kept for up to two years. - You can delete the whole account from Settings, and Khan Academy says it will try to email you first. Security: - Teachers have one rule worth following from day one, and it is Khan Academy's own: its educator guidelines, updated January 2026, tell teachers to keep details that identify a particular student out of Khanmigo queries entirely. Best at: - Getting unstuck on a problem without being handed the answer - Turning a teacher's blank hour into a lesson tied to a named curriculum standard - Practising writing and argument, where being asked "why do you think that?" is the point rather than the delay - Homework a parent wants to supervise without sitting through it Okay at: - Subjects Khan Academy's own library covers thinly — it will still talk, but without the exercises and hints the tutoring normally leans on - Coding, where there is real-time feedback but a dedicated coding assistant goes considerably further - College-admissions and career coaching, which are on the learner plan's feature list but sit outside the exercise library the rest of it is built on Avoid: - Extracting a straight answer under time pressure. Pushing harder generally produces more questions, and a student on a deadline will find that maddening rather than helpful. - Trusting its verdict on your maths working. Khan Academy's own engineering blog has said Khanmigo sometimes judges a student right or wrong incorrectly even when it does the arithmetic correctly, and describes attaching a calculator to compensate. Treat "that's not right" as a reason to check rather than as a result. - Using anything the teacher tools produce as evidence about a student. Khan Academy's educator guidelines say the tools "should not be relied upon as the sole basis for performance assessments or decisions about students", and that they are not a replacement for a teacher's professional judgement. Starter: Make it do the thing it was built for (learner) Asking for exactly the guide-don't-tell behaviour runs with the design instead of fighting it. Ask ChatGPT the same thing and you get a polite imitation that folds the moment you say "just tell me". Once you've got it, swap the last line for "now explain the step I got wrong three different ways". Prompt: I'm stuck on this problem and I don't want the answer. Ask me one question at a time to work out where my thinking goes wrong, and stop as soon as I get there myself. Here's the problem, and here's what I already tried: [paste both]. Starter: Build a lesson from a standard (teacher) The teacher tools draw on Khan Academy's own exercise bank, so the practice questions arrive with answer keys attached rather than being invented on the spot, and you can send students into the same library afterwards. A general chatbot writes you a plausible worksheet anchored to nothing. Give it your actual standard code rather than the topic name, and say what your class covered last week. Prompt: I teach 7th grade math. Build me a 45-minute lesson on ratios and proportional relationships, aligned to my state standard. Include an opening hook, three worked examples at increasing difficulty, a practice set with an answer key, and a four-question check for the last five minutes of class. Starter: Set up a session you can review afterwards (parent) Asking for the summary only makes sense here because the parent account already holds the transcript — you get the short version now and the full chat to check later. On a consumer chatbot there is no parent account for the summary to be addressed to. Ask for the summary as bullet points if you want to compare one week to the next. Prompt: My child is 12 and working on fractions. Act as their tutor for the next 20 minutes. Do not give answers. When they get something wrong, ask what they were thinking before you correct it. At the end, write me a three-sentence summary of what they found hardest. Starter: Argue with it (learner) Guided writing and debate with immediate feedback is a named learner feature, and Writing Coach is listed as its own component in Khan Academy's district package — so the argumentative back-and-forth is a built product here, not a prompt trick you're improvising on top of a general assistant. Tell it what grade you're writing at, and ask it to point at where your evidence is thinnest. Prompt: I'm writing a persuasive essay arguing that school should start later in the morning. Here's my opening paragraph: [paste]. Take the opposite side and give me your three strongest counterarguments, one at a time. Wait for my reply to each before you give the next one. Pitfalls: - The guardrail is a behaviour, not a lock, and not an anti-cheating measure you can rely on. A determined student can still work around it — asking for a worked example of a very similar problem instead of their own — and Khan Academy describes the tutoring style as guidance rather than a hard block. - Khanmigo is not the Khan Academy you may already know. People subscribe expecting the familiar free lesson library to change once they pay, and it doesn't. - A teacher account and a parent account are different products. Khan Academy's help centre notes that a teacher who wants the full tutoring feature set signs up for the paid parent plan and switches on the teacher role in settings — the free teacher sign-up gets you the teacher tools, not the student-facing tutor. - It falls off sharply outside school subjects. Khan Academy scopes it to elementary-through-college academic work — maths, science, coding, history, humanities. It is not the thing to reach for when you want a work email drafted. Where to next: - More education and learning tools: https://www.launchpadhq.io/discover?category=education-learning-ai - Research and study tools: https://www.launchpadhq.io/discover?category=research-academic-tools - General AI chatbots: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Kimi Canonical: https://www.launchpadhq.io/platform/kimi/guide Data: https://www.launchpadhq.io/data/v1/guides/kimi.json Last reviewed: 2026-10-15 Tool: Kimi (https://www.launchpadhq.io/platform/kimi): starting price Free, from $15; free tier yes; data as of 2026-10-26 Moonshot AI's assistant for reading long documents and handing off whole tasks — and, outside China, you sign in at kimi.ai, not kimi.com. You open kimi.ai in a browser, sign in, and type what you want into one box — a question, or a whole job such as "turn this report into slides." Kimi either answers in the chat or works through the task step by step, and you reply to correct or extend what it produced. What it is: - Kimi is the AI assistant made by Moonshot AI, a company founded in Beijing. Its pitch is that one box does two jobs: ordinary chat, and agent work — an agent is AI that takes actions on its own rather than only answering — where Kimi plans a task and delivers a finished file, such as a slide deck, a document, a spreadsheet or a research report. - Its home turf is volume. Moonshot's help centre says you can attach PDFs, Word, Excel and PowerPoint files, images, text and video, many at once in one conversation, and it has a mode called Agent Swarm that splits a big search or batch job across many helpers at once. - Above the box you choose a model (the AI "brain" that does the actual thinking). Moonshot's help centre describes K2.6 as the fast one for quick questions, which answers in text only, and K3 as the most capable, the one that can produce editable .pptx, .docx, .xlsx and .pdf files from start to finish. Since then Moonshot has added a newer model, K3.1 Pro, to Kimi on the web, and free accounts get limited use of it. You also set how hard it thinks: higher settings reason more thoroughly and take longer. - Moonshot publishes the K3 model files themselves on Hugging Face under its own "Kimi K3 License", so other companies can run the same model. At 2.8 trillion parameters it is built for data-centre hardware, not a home computer. Programmers can also reach Kimi's models through a separate pay-as-you-go service at platform.kimi.ai, which has no subscription — current rates are on the Quick facts panel beside this page. Before you start: - Outside mainland China, go to kimi.ai — not kimi.com. Moonshot split Kimi in 2026: kimi.com is now the mainland China service, with phone-number and WeChat sign-in and a Beijing company behind the account, while kimi.ai is the international one. Kimi's own sign-in screen tells people outside mainland China to log in at Kimi.ai with a Google account or a non-+86 phone number; workplace accounts can use the "Sign in with SSO" option with a work email. Moonshot's help centre still says "Visit Kimi.com" in places, so expect older guides — including some of Moonshot's own — to send you to the wrong door. You must be 18, or the adult age where you live if that is higher, unless a parent or guardian agrees to the terms for you. - Kimi's pricing page lists a Free plan with limited use of its newest model and basic agent credits, but no exact message counts, so you find out the boundary by using it. Check the Quick facts panel beside this page for the prices this site has verified before you plan around paying. - Buying a membership may not be possible the day you want one. Moonshot paused new subscriptions on 20 July 2026, saying demand for K3 had pushed its servers close to their limit and that it would reopen spots in batches. In September 2026 a reply on Moonshot's own forum said the pause still applied to both kimi.com and kimi.ai. Whether it has lifted since is not something Moonshot has announced, so treat a membership as something to check for rather than count on. - Expect to be told to wait occasionally. Moonshot's help centre describes two separate brakes: a cap on how many messages you can send within a time window (a red circle by your message, which it says clears in two to three hours), and a peak-hour message, "Kimi is a bit tired", which it says usually clears in a minute or two. Security: - Which Kimi you sign into decides which company holds your chats and which country's law applies to them. The kimi.ai service is run by NOVASCENT PRIVATE LIMITED, a company registered in Singapore; its terms (effective 13 August 2026) are governed by Singapore law, with disputes going to private arbitration in Singapore rather than to a court, and no class actions. Its privacy policy (last updated 13 August 2026) does not name the country where your data is kept and says it may be stored on servers outside your country of residence. The kimi.com service is a different contract: Beijing Moonshot Technology Co., Ltd., whose policy (effective 31 August 2026) says it stores your personal information inside the People's Republic of China. Security: Three things that hold on kimi.ai today:: - Your chats can train Kimi unless you ask in writing. The terms give Moonshot a licence to use what you type and what Kimi produces to train and improve its service. The only opt-out the terms describe is emailing [email address on the web page]; it applies from that point on and does not pull back anything already used. - Moonshot's privacy policy says it shares an account identifier, such as a scrambled form of your email address, with advertising partners to target ads, and that some US state laws count this as selling your data. US readers can turn that off from the "Do Not Sell or Share My Personal Information" link in the site footer. - Uploaded something sensitive by mistake? Moonshot's help centre says to delete the conversation at once, then email [email address on the web page] with the kind of document — "ID card", "bank card", "legal document" — in the subject line so it is prioritised. Deleting removes it from your account, but Moonshot says some records are kept for a legally required period. Security: - None of this makes ordinary use of Kimi unsafe. It does mean the decision about what to paste is yours to make before the first message, not after. Best at: - Reading a stack of documents at once. - Turning a request into a finished, editable file — slides, a Word document or a spreadsheet — when you pick K3. - Big, repetitive research jobs, where Agent Swarm spreads the searching across many helpers instead of working through it one source at a time. Okay at: - Quick everyday questions on K2.6 — fast, but text only. - Making pictures, video or audio. It can, but only after you install the matching plugin from its plugin marketplace; it is not built into the chat. - Busy hours, when the "Kimi is a bit tired" message can make you wait. Avoid: - Expecting a Word or Excel file from the plain chat window. Moonshot's help centre says the chat itself does not generate downloadable Office files; you need Kimi Slides, or an agent task on K3. - Keeping the only copy of anything in Kimi. Moonshot's help centre says deleted conversations and deleted accounts cannot be recovered, and deleting an account also erases any paid benefits on it. - Signing up with a Google account you might want to swap out later. The help centre says a linked Google account cannot be unlinked or replaced, and changing a linked email needs a support request. Starter: The stack reader (many files, one question) Moonshot's help centre says Kimi takes a whole stack of files in a single conversation. On ChatGPT you would usually feed a pile like this in a few at a time; here the comparison across the whole pile happens in one answer, which is what the second half of the prompt asks for. Ask for the disagreements as a table with one column per file. Prompt: I've attached [number] documents about [topic]. For each one, give me a two-sentence summary. Then tell me where they disagree with each other, quoting the exact passage from each and naming which file it came from. Starter: The finished file (switch to K3 first) Run this prompt on K3 and then on K2.6 and you see the file-versus-text split described above for yourself — a difference set by Kimi's own model switcher, which a ChatGPT version of this exercise does not have. Swap .pptx for .xlsx and ask for a spreadsheet of the report's figures instead. Prompt: Turn the attached report into a 6-slide presentation for [audience]. One idea per slide, a short title on each, no more than four bullet points, and a final slide listing three decisions we need to make. Give me an editable .pptx file. Starter: The fresh start (when a chat gets too long) This is Moonshot's own advice. Its help centre says every conversation has a ceiling on how much it can hold, that deleting earlier messages does not free any room, and that the fix is to ask Kimi to "summarize and generate a handoff document" and paste it into a new chat. Learn it on day one and a long project never stalls. Prompt: Summarize and generate a handoff document: the key conclusions so far, decisions made, open questions, and anything I asked you to remember. Format it so I can paste it into a new conversation. Starter: The effort test (same question, two settings) Kimi lets you set how hard the model thinks, and Moonshot's help centre says the higher settings reason more fully but use more of your allowance. Ask once on the lowest setting and once on the highest, and you learn when the extra effort is worth spending. If the two answers agree, the lower setting is enough for questions of that kind. Prompt: [Your hard question — for example, should I pay off my car loan early or invest the money? Here are my numbers: ...] Lay out the reasoning step by step, and tell me which assumption, if it were wrong, would change the answer. Starter: Save a preset for prompts you reuse Store a prompt you use often — such as the stack reader above — as a preset, so it is one click away in every new chat. Moonshot's help centre lists presets as a core feature for exactly this. Because Kimi's strength is repeat work on big piles of documents, a saved instruction pays off faster here than on an assistant you use for one-off questions. Pitfalls: - A red circle by your message is a limit, not a fault — the message cap covered in Before you start. Refreshing the page or signing out and back in does nothing. - "Let's talk about something else" ends a topic. That reply means your message tripped Kimi's safety rules. Moonshot says to start a new topic, and to tap the thumbs-down button if you think the block was a mistake; rephrasing the same request rarely gets past it. - The help centre describes more than the box in front of you. Kimi Work, Kimi Code and Kimi Claw are separate Moonshot products that share its help centre with the chat. If an article describes something you cannot find, check which product it is about before assuming your account is missing it. Where to next: - International & Regional AI: https://www.launchpadhq.io/discover?category=international-regional-ai - Chat assistants: https://www.launchpadhq.io/discover?category=text-conversational-ai - AI APIs & developer services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services ### Getting Started with Kira by Litera Canonical: https://www.launchpadhq.io/platform/kira-systems/guide Data: https://www.launchpadhq.io/data/v1/guides/kira-systems.json Last reviewed: 2026-09-13 Tool: Kira (https://www.launchpadhq.io/platform/kira-systems): starting price Custom; free tier no; data as of 2026-10-20 Firm-licensed AI that reads a stack of contracts and lines up the clauses that matter in one table. You load a folder of contracts into a project, tick the clause types you want found, and wait while it reads them. What comes back is a table: one row per contract, one column per clause, with each answer linked to the exact paragraph it came from. You click through, correct anything wrong, and export. What it is: - Kira is a contract-review tool used mainly for due diligence — the part of a deal where someone has to read every agreement a company has signed and report on what is in them. - It was built by Kira Systems, a Toronto company Litera agreed to buy in August 2021; the Kira name survived that deal but the company did not, and kirasystems.com now lands on Litera's own site. - Most beginners meet it because their firm bought it rather than because they picked it. Before you start: - You cannot sign up for this yourself. Every route on Litera's site ends at "Request a Demo," and Kira is not one of the four products Litera's own online store will sell you. If you are not somewhere that already licenses it, the honest way to start learning AI-assisted contract review is to upload a single agreement to a general assistant such as ChatGPT and ask it to list the obligations and dates — less precise, available today. - Litera publishes no price for Kira anywhere on its site, so the real answer to "will I need to pay" is that someone already has: it is licensed to the firm, per subscription, not to you. Any figure you find quoted on a software-comparison site is somebody's estimate rather than a rate card, and whether there is a minimum number of seats is not something Litera publishes either — it comes out of the sales conversation. - Access does not appear the day the paperwork is signed. Litera's current Kira cloud terms (version May 2026, in force for subscriptions starting on or after 1 May 2026) say a standard shared cloud setup is made available within three business days of the subscription start, and a private one targets around two weeks. - Check the languages before you promise anyone a deadline. Litera says you can create your own clause finders in any language just by describing what you want in a prompt (the message you type), but coverage of non-English contracts by the ready-made finders varies from field to field — Litera keeps a support article on precisely this question inside its customer portal, so ask whoever administers your firm's Kira to check it against the languages your deal actually involves. Security: - Everything you put into Kira is somebody else's confidential contract — usually a client's, often covered by an NDA (a signed agreement not to share someone's confidential information) you did not personally sign. That changes the question. It is not "is this safe for me"; it is "can I tell a client exactly where their documents went and who touched them." Kira is unusual among AI tools in that the answer is written down. Security: What Litera commits to in writing: - It will not train on your work. Litera's terms for products with generative features (version April 2026) say Litera will not, and will not let its outside providers, develop, train or fine-tune generative models using customer data, prompts or outputs unless separately agreed in writing. - Your firm picks the country. The current Kira cloud terms say the data centres holding your documents sit in the hosting jurisdiction the customer chose — so "where is this contract right now" is a fact your administrator can answer, not a guess. - The outside companies involved are published. Litera's subprocessor list for Kira names Amazon Web Services for storage in that customer-selected jurisdiction, and Microsoft's Azure OpenAI service for the AI-written answers, hosted in regions matching where Kira itself is hosted, or a comparable alternative. Security: What stays on you: - The terms make the customer warrant that it has the rights to put those documents in at all, and that doing so breaks no duty of confidentiality owed to anyone. That warranty is signed at firm level and honoured — or broken — by the person at the upload screen. - The generative features can be switched off for one project while another leaves them on, so it is worth knowing which side of the switch you are on before you start. - When a subscription ends, downloading or deleting the documents is the customer's job, with limited access for thirty days afterwards to do it. Best at: - Pulling the same clause out of hundreds of agreements at once and lining the answers up side by side - Due-diligence work where the deliverable is a report on what a pile of contracts says — mergers and acquisitions, real estate, banking and finance are the areas Litera builds for - Teaching it to find a clause nobody else needs — a house-specific provision your firm keeps running into Okay at: - Contracts in languages other than English, where the ready-made finders help unevenly and you may end up building your own - Being your only tool — it reviews contracts; drafting, comparing versions and the rest sit in separate Litera products - Small jobs. It is built for volume, and three contracts do not repay the effort of setting up a project Avoid: - Heavy exploratory use of the AI-written-answer features. Litera's terms reserve the right to throttle you, move you to a higher-cost plan, or add usage charges if your usage "materially deviates" from that of similar customers — the question box is not an all-you-can-eat part of the licence. - Building a large library of custom clause finders and assuming it is yours. Under the current cloud terms those trained finders are exclusively yours during the subscription and are deleted when it expires, with neither side permitted to use them afterwards. - Leaning on a generated answer as advice. Litera's own terms disclaim every warranty on generative output and state that these features are not professional advice and create no attorney-client relationship — which is the vendor telling you, in the contract, to verify before you send. Starter: Run one clause across the whole pile Create a project, load every agreement in the deal, and tick a handful of ready-made clause finders — change of control, assignment, governing law. Kira's shape is a library: Litera puts 1,400-plus pre-trained clause finders and a 95%-or-better precision claim on target provisions on its own product page, which is why your first useful output comes from ticking boxes rather than teaching the tool what a change-of-control clause looks like. Neither Luminance nor eBrevia has published those two figures for their own products. Start with three clause types, not thirty. The grid is only readable while it is narrow. Starter: Decide the AI question before you open the project Kira lets the generative features be turned on or off for one project without affecting others. You set that when the project is created, so a matter running under a client's AI restrictions and a matter without them can live in the same firm instance. Litera scoped this control at the project level rather than account-wide, and named it in its January 2026 release material as the reason a firm can honour one client's restrictions without disabling the feature for everyone — a governance answer you can give about a single matter. Luminance documents no equivalent per-project switch. Starter: Ask the whole deal one question Grid Chat, which Litera introduced in July 2026, answers across every contract in the worksheet at once and builds the answer from Kira's already-checked extractions rather than from raw document text, and flags when it is missing data. That validated-extraction layer is the distinction Litera draws against general chat tools, and it is not how eBrevia works. This goes into the chat inside the analysis grid, not a general search box, and it only works once the extraction has finished — Litera describes it as plain-language questions over a worksheet you have already built. Exactly how that box is labelled is worth confirming once your login is set up. Prompt: Which of these agreements contain a change-of-control provision triggered by a merger, and for each one, does the counterparty get a right to terminate or only a right to be consulted? Starter: Train a finder for a clause nobody else needs When a provision matters to your firm and is not in the library, you show Kira examples of it in your own documents and it learns to find that clause in new ones. Litera's current cloud terms give these a name — "Customer Provisions" — and a lifespan. That is a defined term in Litera's own May 2026 Kira licence rather than a capability claim, and knowing the trained finder is leased rather than owned changes whether you build ten of them or a hundred. Starter: Make the memo out of the grid Kira turns the completed table into summaries and export-ready reports, so the diligence memo is generated from the extraction instead of being retyped out of it. Litera sells this as "answers, not spreadsheets," and the mechanism is that the report is built from the same checked extractions the grid holds, so a statement in the deliverable still traces back to a clause. eBrevia has no comparable extraction-to-report pipeline to point at. Read the summary against two source clauses before you send it. Generated text is where the traceability quietly stops. Pitfalls: - Assuming "Kira Systems" is still someone you can deal with. It is not — support runs through Litera's customer portal, and anything written about Kira Systems as an independent vendor predates August 2021. - Reading extraction accuracy as answer accuracy. A precision figure on target provisions means the clause was correctly found. It says nothing about whether the clause means what you assumed it meant. - Trusting a review of "Kira" written before mid-2026. The generative layer, the analysis grid and Grid Chat all arrived between July 2025 and July 2026, so older write-ups describe a genuinely different product. Where to next: - Legal AI tools: https://www.launchpadhq.io/discover?category=legal-ai - Document & PDF processing: https://www.launchpadhq.io/discover?category=document-pdf-processing - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Kiro Canonical: https://www.launchpadhq.io/platform/kiro/guide Data: https://www.launchpadhq.io/data/v1/guides/kiro.json Last reviewed: 2026-09-08 Tool: Kiro (https://www.launchpadhq.io/platform/kiro): starting price Free, from $20/user/mo; free tier yes; data as of 2026-10-14 AWS's code editor that writes the plan down before it writes any code. You open Kiro on your computer, type what you want built in a side panel, and instead of writing code straight away it writes back a short plan you read and approve. Once you approve it, Kiro edits the files in your project itself, one step at a time, and you check each change. What it is: - Kiro is a code editor from AWS, Amazon's cloud division, with an AI agent (AI that takes actions on its own rather than only answering) built into it. - Its distinguishing idea is that the agent writes the plan down first. Ask for a feature and Kiro produces three plain Markdown files — requirements, design, and a task list — inside a .kiro/specs/ folder in your project. AWS calls this spec-driven development. - It launched in preview in July 2025 and became generally available in November 2025, so there is no waitlist. It's built on Code OSS, the open code underneath VS Code, so a VS Code user can import their settings, themes and most add-ons the first time they open it. Before you start: - Kiro is for people who already write code, or are actively learning to. It works on a project folder that already exists on your computer, and what it produces is code you are responsible for reading. If you want to build something with AI without that background, a browser-based tool like Lovable is a gentler place to start. - Getting in is free and quick: download from kiro.dev, sign in with GitHub, Google or an AWS Builder ID, and no card is asked for. - The free plan gives you a small monthly allowance of credits (the platform's unit of spend — each thing you make costs some). They reset at the start of each billing month, don't roll over, and can't be topped up on the free plan. A simple question can cost a fraction of a credit; running a spec task usually costs several — so the free allowance is an evening of experimenting, not a month of work. Paid plans come with much larger allowances; current prices are on its Launchpad page. - Which AI does the work changes how fast that runs out. The free plan includes a Claude Sonnet model, with limits, plus several cheaper open-weight models (the AI "brain" that does the actual thinking). There is also a version for people who prefer typing commands, and a browser version, but the browser one needs a paid plan. Getting set up safely (Download only from kiro.dev/downloads. A code editor gets access to everything in the project folders you open with it, so this is not the software to take from a search ad, a mirror site, or a link someone sent you.): - Check there's a build for your machine: macOS (Apple Silicon and Intel), Windows x64, and Linux (Debian/Ubuntu 24 or later). - Change the data setting before your first real project, not after — see below. - Point it at a scratch project first. Kiro edits files itself, so let the first spec run somewhere nothing is lost if it goes wrong. Security: - The question to settle before your first project isn't whether Kiro sees your code; it has to. It's what AWS is allowed to keep. On the free plan the default is that your prompts, Kiro's replies and the code it generates may be used for service improvement, including model training. One setting turns all of it off: Settings → User sub-tab → Application → Telemetry and Content → Disabled. Security: Before you point it at anything real: - Kiro's privacy page words the paid exemption around "an Amazon Q Developer Pro subscription", while its FAQ words it around Pro and above signed in through AWS IAM Identity Center. If you are counting on being exempt, confirm it against your own account rather than the docs. - Agent hooks fire on their own when files are saved or created. They're useful, and they spend credits without you typing anything. Best at: - Turning a vague feature idea into a written plan you can argue with before any code exists - Working through a multi-step change one approved task at a time - Coming back to a project later — the spec files are ordinary Markdown left in your folder, so the reasoning is still there Okay at: - Small one-line edits — you can chat to it without a spec, but the ceremony is the point of the tool, and a plain chat assistant is faster for a typo - Large existing codebases — it will take them on, but the review burden scales with everything it touches Avoid: - Assuming privacy is the default. Content collection is on unless you turn it off, and the exemption the docs describe is written for paid, AWS-account sign-ins. - Treating the approval gates as a safety net you can't lose. Quick Spec generates requirements, design and tasks in one go with no review step between them, and it sits one click from the normal flow. Starter: Your first spec The "stop and wait" is the mechanism, not politeness — Kiro writes requirements.md, design.md and tasks.md into .kiro/specs/ and pauses between each, so you rewrite the plan while it still costs a credit instead of a whole implementation. Swap in a feature for a project you already have open. Prompt: Create a spec for a small to-do list app I can run from my own machine: add a task, list tasks, mark one done, saved to a local file. Stop after the requirements and wait for me to approve them. Starter: Set the model before you spend anything In the chat panel's model picker, move off the default Auto to one of the cheap open-weight options for routine work. Kiro prices every model as a multiplier of Auto — currently as low as 0.05x, against 1.3x for Claude Sonnet 4.5 — so on the free plan's small allowance the picker, not the prompt, is what decides how long you last. Starter: Write your conventions down once Kiro keeps these as three separate files — product.md, tech.md, structure.md — in .kiro/steering/, and you can keep a global set in ~/.kiro/steering/ that applies to every project, so a correction you make once outlives the chat you made it in. Correct the generated files by hand; they're plain Markdown. Prompt: Read this project and generate steering files for it — product, tech and structure — then show me what you inferred so I can correct it. Starter: Give it a check that runs on its own work Kiro's file hooks fire on the agent's saves rather than yours, so the check lands on machine-written code. Prompt: Add a hook that runs my linter whenever you save a TypeScript file, then tell me what it caught. Pitfalls: - The free credits go faster than you'd expect. Spec refinement, task execution and hook runs all come out of the same monthly pool, and they don't carry over. - Approving a task list is not reading the code. The plan can be right and the implementation still wrong. - It's a whole editor, not a plugin. Using it means switching editors; importing your VS Code settings softens that but doesn't remove it. - The docs lag the product in places — the privacy page still frames the paid exemption around Amazon Q Developer Pro, a name that predates today's plan list. Check anything you're relying on inside the app. Where to next: - AI coding tools: https://www.launchpadhq.io/discover?category=ai-coding-development - APIs and developer services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - Run AI on your own machine: https://www.launchpadhq.io/discover?category=local-open-source-ai ### Getting Started with Kling AI Canonical: https://www.launchpadhq.io/platform/kling-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/kling-ai.json Last reviewed: 2026-10-15 Tool: Kling AI (https://www.launchpadhq.io/platform/kling-ai): starting price Free, from $1; free tier yes; data as of 2026-10-17 Type a description, get a short video clip back — paid for in credits, under terms that change depending on whether you are a paying member. You open Kling in your browser, choose video or image, and type a prompt (the message you type) describing the scene: what is in it, how the camera moves, what the light is like. Kling takes a little while to make the clip, then you watch it, adjust the wording or a setting, and generate again. What it is: - Kling AI is a video and image generator from Kuaishou, a Chinese technology company; the service you sign up to is run by its Singapore company, Kling AI Pte. Ltd. You describe a scene in words, or upload a picture to start from, and it produces a short clip or an image. It runs in a web browser, and Kling also offers developer access so other software can use it. - Its current headline video model is Kling 4.0, and its image model is Kling Image 3.0. Kling's own blog, dated 30 September 2026, describes Kling 4.0 as rolling out in limited early access, with the full model "coming in October 2026" and availability that "may vary during the initial rollout" — so whether it appears in your account yet is something you will see once you are signed in. - Its distinguishing tools are about directing rather than just describing. Kling's video guide documents a Multi-Shot switch that plans cuts between several shots inside one clip, a mode that animates between a start picture and an end picture you supply, and audio generated with the video, including spoken lines you write into the prompt. Before you start: - Everything you make costs credits (the platform's unit of spend — each thing you make costs some). Kling's Credits Policy says the charge for a generation is whatever the screen shows at that moment, and its video guide prices video by the second: longer clips, higher resolution and generated sound each cost more. Credits come off when you press generate, and Kling's policy says that if a generation fails, they are refunded. - Credits do not all last the same length of time. Under Kling's Credits Policy, a membership's credits arrive each month and expire if you have not used them before their validity runs out; credits you buy separately last far longer, and are unaffected by whether you keep your membership. Kling spends whichever credits expire soonest first. None of them can be refunded or cashed out. - This page does not quote Kling's prices or say what any membership includes, because we have not yet verified them. Kling publishes them on its membership page — check there, and check whether a plan is a first-month offer that renews at a different price, before you pay. Memberships renew automatically until you cancel, and Kling's Terms of Paid Service say fees are not refunded once a paid service starts. - Whether you may use what you make depends on whether you pay. Read the security section below before you publish anything — this is the fact on this page most likely to matter later. Security: - Kling's two sets of terms say different things about using your videos commercially, and the difference turns on membership. Its Terms of Service (updated 21 April 2026) say you own what you create, but also that you may not use your output 'for any commercial purposes' without Kling's written permission, and that you must label anything you share with Kling's brand or say it was made by Kling AI. Its Terms of Paid Service (dated 10 October 2026) say paying members' commercial use 'is not restricted' — except for building a competing product — and that only members may remove Kling's watermark (a mark on generated output identifying it as AI-made). If you are not a paying member, treat everything you make as personal use only. The paid terms do not say whether that permission survives for clips made while you were a member once the membership ends. Security: - Kling also takes rights in what you upload and generate. Its Terms of Service give Kling a royalty-free licence to your inputs and outputs, which it may use for promotion and to train and improve its AI models, and say your content may be made available to other users to build on. The terms say you can withdraw that permission by emailing [email address on the web page]. Kling's Privacy Policy (updated 21 April 2026) says your data is stored on servers in Singapore and may be shared with its corporate affiliates; it says Kling does not analyse facial-recognition features in pictures, but in the same document says that for face-based features it may analyse what you upload and record feature points. Security: Before you upload anything: - Don't upload photos of other people without their permission. Kling's terms forbid impersonating anyone and make you responsible for the portrait and personality rights of anyone in your inputs. - Don't upload confidential footage, client work or brand assets you do not control, given the licence above. Best at: - Short, cinematic clips where you want to direct the camera — push-ins, tracking shots, a handheld feel — by writing it into the prompt. - Several shots cut together inside a single clip, using the Multi-Shot switch rather than stitching clips in an editor. - Animating between a start picture and an end picture you choose. - Clips with sound and spoken lines generated together with the picture. Okay at: - Getting exactly what you pictured on the first try. Expect to regenerate, and every regeneration costs credits. - Matching its newest model to the guides. Kling's detailed video guide was written for the model before Kling 4.0, so its controls may not carry over exactly. Avoid: - Anything you intend to sell, monetise or use for a client unless you are a paying member — Kling's own Terms of Service bar commercial use of output without written permission. - Uploading faces, brand material or confidential footage you would not want Kling to reuse: its terms license your inputs and outputs for promotion and model training until you email to withdraw. Starter: One shot, camera directed Kling's video guide prices video per second and charges more for generated sound, so a short clip with audio switched off is the cheapest test of how Kling reads camera wording. Its guide's own examples write the camera into the prompt — 'cinematic handheld', 'one continuous long take' — which is the habit worth building first. Change only the camera phrase — 'slow orbit', 'handheld', 'static wide shot' — and generate again to see what each one does. Prompt: A ceramic coffee mug on a wooden table beside a rainy window, steam rising, soft morning light, slow push-in camera, one continuous shot. Starter: Three shots in one clip Turn on Kling's Multi-Shot switch before you paste this. Kling's video guide documents that switch and the 'Shot 1, …' labelling its own examples use; with it on, Kling plans the cuts between shots inside one clip. With it off, you get a single shot and the labels are just words. Kling's guide also describes a Custom Multi-Shot option for setting how many shots there are and how long each lasts. Prompt: Shot 1, wide shot of a cyclist at the top of a hill at sunrise, cinematic handheld. Shot 2, close-up of their hands gripping the handlebars. Shot 3, tracking shot as they ride downhill past pine trees. Starter: A line of dialogue Kling's video guide shows spoken lines written this way — the speaker's name, then tone and language in brackets — and says lines in languages it does not support are translated into English. Dialogue only works with generated sound switched on, which costs more, so try this once you have done a silent clip or two. Prompt: At a night market, a noodle vendor hands a steaming bowl to a customer under string lights. Vendor (cheerful, Spanish): '¡Buen provecho!' Customer (laughing): 'Thank you!' Starter: Start and end frames Upload one picture as the first frame and another as the last, and Kling animates the motion between them. Kling's guide documents a Start & End Frames mode, which gives you far more control over where a clip lands than a description alone. Use two of your own photos of a place or an object rather than a person — under Kling's terms, what you upload is licensed to Kling, and you carry responsibility for anyone pictured. Pitfalls: - Regenerating is where credits go. Every new attempt is charged again. Change one thing per attempt so each one teaches you something. - Check the charge on screen before you press generate. Settings you didn't notice can change it. - Waiting is normal, and Kling doesn't publish how long. Neither the terms nor the guides give generation times or say how long a queue can be. Kling's own explainer pages say to 'wait a few seconds', which you should not rely on for video. - Cancelling takes effect at the end of the billing cycle, not immediately. Kling's paid terms say you keep access until the current cycle ends, and if a renewal charge has already gone through shortly before you cancel, access may run into the next cycle. You cancel wherever you subscribed. Where to next: - Video Creation & Editing: https://www.launchpadhq.io/discover?category=video-creation-editing - Image generators: https://www.launchpadhq.io/discover?category=image-generation-editing ### Getting Started with Krea Canonical: https://www.launchpadhq.io/platform/krea/guide Data: https://www.launchpadhq.io/data/v1/guides/krea.json Last reviewed: 2026-10-15 Tool: Krea (https://www.launchpadhq.io/platform/krea): starting price $9/month; free tier yes; data as of 2026-10-24 One studio for many image and video AIs — plus a canvas that redraws your sketch the moment you move a line. You type a description of the picture you want into a box at the bottom of the screen (this is your prompt — the message you type), press Generate, and a few images come back. Pick the one closest to your idea, change a few words, and run it again until it looks right. What it is: - Krea is a browser studio for making and editing images and short videos with AI. It puts many other companies' image and video models (the AI "brain" that does the actual work) behind one account — among them Nano Banana 2 and Seedance 2.0 — and you choose between them from a menu, or let Krea choose for you. - Two tools are what set it apart from a plain image generator. Realtime is a drawing canvas with the result on the right that redraws itself as you work — no Generate button, no queue — so a rough circle and a few strokes become a finished-looking image while your hand is still moving. The Enhancer takes an existing picture and makes it larger and sharper, filling in detail as it goes. - Everything you make spends units (Krea's credits — the platform's unit of spend; each thing you make costs some). Different models cost very different amounts for a single picture, and the model menu shows each one's cost before you commit. - It runs in a web browser, and Krea also offers an iPhone app. Before you start: - Signing up is free. The free plan gives you a small allowance of units that refills every day — Krea's help pages confirm the daily allowance but do not say what time it resets — so a free account is for trying things out a little each day rather than for a big project in one sitting. - Read this before you use anything you make on the free plan for work. As of 15 October 2026, Krea's pricing page marks "Commercial license" as not included on the free plan, and its own FAQ names Pro, Max and Business among the plans that include the commercial licence. That means a logo, product shot or client image made on a free account is not covered for commercial use. The FAQ does not mention the Basic plan either way, so if commercial use is your reason for paying, confirm it on the pricing page before choosing Basic over Pro. Krea's terms (last updated 20 May 2024) also say it can change its commercial-use terms "at any time". - Who owns what you make depends on your plan. Krea's general terms (last updated 20 May 2024) give you a licence to "use (and in certain instances download, display, and re-share)" what you generate, and never say in plain words that you own it. Only the separate Business terms (last updated 1 January 2026) say "You retain all ownership rights in your Inputs and Outputs" — and even those warn that AI output may not be protected by copyright in some countries and that someone else may get a similar result. - If you do pay, Basic is $9/month, and Pro, Max and a Business plan for teams step up from there, each with a larger monthly allowance of units and a discount for paying yearly. Subscriptions renew automatically until you cancel. Security: - On anything below the Business plan, assume what you make is not fully private. Krea's terms (last updated 20 May 2024) say that if your plan does not include private mode, "some of your User Submissions may be viewable by other users" — and they do not say which plans include it. Its privacy policy, last updated the same day, says the service is hosted on US servers. Security: Before you upload a photo of a real person, or anything confidential: - The promise not to train AI on your work is in writing only for Business. Krea's Business terms (1 January 2026) state it "will NOT use your Inputs or Outputs to train its AI models." The general privacy policy (20 May 2024) does not say either way whether your prompts and uploads are used for training, and lists improving the service and product development among the reasons it uses your data. - Deleting an image from your screen does not delete it. Krea's image guide says Delete removes it from your session view but "your assets are not deleted." The privacy policy's route for having your personal data deleted is an email to [email address on the web page]. - Krea's terms prohibit impersonating people without their consent, including deepfakes. A photo of a friend fed into the Enhancer or Realtime is a photo of a real person on someone else's servers. Best at: - Trying the same idea on several companies' image models without opening several accounts. - Sketching a layout — a horizon line, a blob for a subject — and watching it become a picture as you draw, which is quicker for composition than rewriting a description. - Making a small or soft image bigger and crisper for screens or print. - Turning a still you like into a short video clip from the same session. Okay at: - Exact text inside an image, like a sign or a poster headline — some models handle it, others garble it, and which is which changes as the model list changes. Avoid: - Using free-plan output in anything you sell or deliver to a client — the pricing page excludes the commercial licence from Free. - Treating the Enhancer as a way to recover a blurry photo. Krea's own guide calls it "a creative tool, not a forensic one" that "does not recover what was originally there" — on a face or a document it is inventing the detail, not revealing it. - Running the same image through the Enhancer again and again. Krea's guide warns that repeated passes "compound the AI's interpretations", adding brushstrokes or skin texture that were never in the original. Starter: The four-part description Krea's own image guide recommends naming subject, setting, lighting and mood — this prompt does exactly those four. Leave the model on its default automatic setting for the first try, then press Retry: most runs return two or four images, and the differences between them show you what the description left open. Swap in your own subject and keep the other three slots filled. If one image is close, use Reuse parameters to load the same prompt and settings back into the box and change one word at a time. Prompt: A small corner bakery at dawn, warm light spilling from the window onto a wet cobblestone street, a cat asleep on the step, calm and quiet mood, photographic Starter: Draw a rough shape and watch it become a picture Open Realtime, type a short description such as "mountain lake at sunset", then drag a few shapes onto the canvas — a triangle for a mountain, a circle for the sun, a blue streak for water — and move them around. Krea's Realtime guide says there is no queue and no render button — the output reflects the canvas at every moment, so moving the sun from left to right is a one-second experiment rather than another generation. Press the seed button for a fresh variation of the same layout. When one is worth keeping, use the options under the result to download it or send it to be enlarged. Starter: Enlarge a small image — gently Upload a small picture of your own (a landscape or an object, not a person) to the Enhancer, choose 2x rather than the largest size, and compare before and after with the slider. Krea's Enhancer guide advises starting with conservative settings and stepping up 2x then 4x rather than jumping to 8x, because higher values give the AI more room to invent texture and edges. Seeing it add detail that was never there on your own photo teaches what the tool really does faster than reading about it. If the result looks over-sharpened, lower the strength rather than running it again — Krea's guide says each extra pass compounds the changes. Starter: Same idea, two models Run it once on the default, then open the model menu bottom-left and pick a different model. Krea's picker lists each model's speed, quality and unit cost side by side, so the comparison also shows you how unevenly your daily units are spent. Prompt: A paper origami crane on a wooden desk beside a cup of coffee, soft morning light from a window, shallow depth of field Pitfalls: - The model menu is the price list. Krea's guide suggests trying ideas on a cheaper model and switching to a pricier one once the description is right. - Delete does not mean gone (see the privacy section above). Publish, by contrast, makes a public page for an image — know which button you are pressing. - The model list changes often. Krea's changelog shows models being added and older ones retired month by month. A model a tutorial or video recommended may have moved or gone; pick from what the menu shows today. - Video and enlarged video use units fast. Krea's Enhancer guide warns that upscaling videos "can consume your compute budget quickly." Try an idea as a still first and turn only the keeper into a clip. Where to next: - Image generators: https://www.launchpadhq.io/discover?category=image-generation-editing - Video tools: https://www.launchpadhq.io/discover?category=video-creation-editing ### Getting Started with Leonardo.Ai Canonical: https://www.launchpadhq.io/platform/leonardo-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/leonardo-ai.json Last reviewed: 2026-09-07 Tool: Leonardo AI (https://www.launchpadhq.io/platform/leonardo-ai): starting price Free, from $28/mo; free tier yes; data as of not dated An image generator with a free allowance that refills every day — built for characters, worlds and concept art. You type a description of the picture you want — that description is a prompt, the message you type — and Leonardo draws it in seconds. You look at what came back, change a few words, and go again, each attempt spending from a balance on screen. What it is: - A browser tool at app.leonardo.ai — nothing to install, no card to start. - The free plan is why beginners land here: a daily allowance of credits (Leonardo calls them tokens), with a standard image among the cheapest things to make. Midjourney has no free tier; Firefly and Ideogram give a monthly allowance that runs dry and makes you wait. Leonardo's tomorrow brings more. - It leans toward game art and character work, and the controls show it — reference-image slots, character consistency, trainable styles. There are also Blueprints: prebuilt setups where you fill in a field or two instead of writing a prompt. Before you start: - You need an email address and a browser, plus a reference image if you already have a character in mind. - Free credits reset every 24 hours and do not carry over, so you cannot save up for something big. Paid plans swap that for a monthly allowance that banks what you don't spend. Costs vary by feature and are shown before you commit. - Learning costs nothing. What pushes people to pay is privacy — free work is public — and video, where one clip can eat a day's allowance. Paid plans come in several monthly tiers, cheaper billed annually; current prices are on its Launchpad page. - First step: sign up, open the image generator, type one plain sentence describing a scene, and generate. Don't touch a setting yet. Security: - Understand this before your first generation: on the free plan, everything you make is public. Others can see, copy and remix your images, and Leonardo's policy states free-tier content is used to improve its models. Private generation needs a paid plan — you pay with visibility, not money. Security: What that means in practice: - Generate nothing on the free plan you'd mind seeing in a public gallery — nothing you agreed to keep confidential, no unreleased designs, no personal photos. - Reference images you upload are content submitted to the service — treat other people's work with public-site caution. - Commercial use is allowed on every tier including free — but on free that is a licence to use your images, not exclusive ownership of them. - On paid plans you keep full ownership and copyright, and that survives if you later downgrade to free. Best at: - Characters, creatures and concept art — what the platform was built around - Stylised illustration, fantasy and sci-fi, painterly game-art looks - Cheap, high-volume iteration without watching a meter - Getting a usable result with no prompt-writing skill, via Blueprints Okay at: - Photorealistic product shots — good, not always best in class - Readable text in an image — newer models are far better, but check every character - Short video clips from a still — capable, but expensive on a free allowance Avoid: - Confidential or client work on the free plan — it all lands in a public gallery - Expecting images to be exclusively yours on free: commercial use is allowed, exclusivity is not - Video on the free daily refill; one clip can swallow the lot - Layout-critical design — logos, exact typography, brand fonts Starter: Generate one image with no settings touched Leonardo picks a model — the AI "brain" doing the drawing — when you don't, so you get a competent result before you know what a model is. A plain image costs very little of a daily allowance that refills, so you can run it dozens of times for nothing. Change only the last clause — swap the warm side light for flat overcast light. One variable at a time is how you learn which words do work. Prompt: A weathered brass diving helmet on a wooden dock at sunrise, salt crust on the metal, background softly out of focus, low warm light from one side Starter: Run a Blueprint instead of writing a prompt Opens a Blueprint, fills in the one or two fields it asks for, checks the credit cost it shows, and generates. Blueprints are Leonardo's answer to the blank prompt box, and each shows its credit cost before you generate — which matters here, because some cost more than a free user's whole daily allowance. Starter: Make a character concept sheet This is the work Leonardo was shaped around: its models are built and sold for game and concept art rather than photography, which is why the concept-sheet convention — several views, neutral background, painted finish — is the thing to ask it for. Prompt: Character concept sheet for a swamp ranger, front view and side view, muted green leather armour, oversized canvas satchel, weathered boots, flat neutral grey background, painted concept art style Starter: Hold one character steady across several images Uploads a picture of your character into the character-reference slot in the generation panel, then generates new scenes with the reference in place. Character reference sits in the generation panel as a plain image upload, not behind prompt syntax you must memorise. That is what turns a pile of nice pictures into a set. Pitfalls: - Free generations are public — that is the actual price. Others can view and remix them, and free content feeds model training. - Free credits don't roll over. Unused ones vanish at reset, so a little most days beats planning a marathon. - Costs vary enormously by feature. A plain image costs very little; video and heavy workflows many times that. Read the number first. - Much of the advice online points at legacy features. Alchemy, PhotoReal and Elements still work but are no longer core — if a guide leads with those, check its date. Where to next: - Compare it with other image tools: https://www.launchpadhq.io/discover?category=image-generation-editing - Turn your images into motion: https://www.launchpadhq.io/discover?category=video-creation-editing ### Getting Started with Lindy Canonical: https://www.launchpadhq.io/platform/lindy-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/lindy-ai.json Last reviewed: 2026-10-14 Tool: Lindy AI (https://www.launchpadhq.io/platform/lindy-ai): starting price from $29.99/mo per active user (3,000 credits each); free tier no; data as of 2026-10-14 Give it your inbox and calendar, and it handles the admin around them — on a schedule, without being asked. You message Lindy the way you'd message a colleague and it comes back with the work done. For jobs you want repeated, you describe one in a box in plain English, and Lindy runs it on a schedule or on a trigger (the "when" — the event that starts an automation). What it is: - Lindy sells itself as an AI employee rather than an automation builder. Its own documentation opens: "Lindy is an AI employee. It lives in your Slack, connects to your tools, and comes back with the work done." You give it standing access to your mailbox and calendar, and it works from those continuously rather than from one form submission at a time. - The work it is built around is the admin layer of a job: labelling and archiving incoming mail, drafting replies in your phrasing, joining Meet, Zoom and Teams calls to record and write them up, prepping you before a meeting, chasing emails that got no reply, and scheduling. Those ship as built-in routines you switch on, not as flows you assemble. - It also takes one-off requests — "what's on my calendar today", "draft a reply to this one" — and Lindy's docs say it connects to 1,000+ other tools through a sign-in or an access key, so a routine can reach into Notion, HubSpot or Linear as well as your inbox. - Worth knowing before you go looking for help: Lindy used to be a visual builder where you wired steps together on a canvas, and most of the tutorial material still online describes that product. The company has since replaced it with this teammate model — its founder announced the change publicly as retiring the word "agent" — and the current documentation no longer covers the old builder. Before you start: - There is no ongoing free plan. You start with free credits that last 7 days, no credit card. After that, Team is billed monthly for every active user, with a pooled credit allowance, and you can buy top-ups when it runs low. Direct signups are billed right away; a 7-day seat trial still applies to teammates who join through Slack. - Setup starts by connecting your Google or Microsoft account so Lindy can, in its own words, "run your inbox, prep and take notes on meetings, and handle scheduling." That is your live mailbox and your live calendar, not a copy. If the account belongs to an employer, expect a second obstacle: Google lets Workspace administrators decide which third-party apps may reach Workspace data, so on a locked-down work account the connection can be refused until an admin allows it. Lindy publishes nothing about this either way, so check with whoever runs your email before you plan around it. - Usage is metered in credits (the platform's unit of spend — each thing you make costs some). Lindy's pricing page shows everyday asks costing the fewest credits, research-grade work more, and big builds the most. Credits pool across everyone on the workspace and do not roll over. Run out and, per the same FAQ, "Lindy pauses credit-using actions until your credits reset" — nothing breaks and you are not billed extra. - The documented starting point is Slack, and installing the Lindy app there is normally an administrator's job. There is a web app as well — its sidebar carries Home, Chat, Meetings, Files, Routines, Skills and Integrations — so you are not locked out without Slack, but the quickstart Lindy publishes assumes the Slack route, and how much of a solo setup works without it is not something we could confirm from the public docs. Security: - Lindy draws its safety line at the edge of your account, not at the edge of your inbox. Its pricing FAQ answers the question "Will Lindy do anything without my approval?" with "No. Anything with outside impact waits for approval such as sending an email, updating a ticket, posting to another channel, publishing a doc." But read what it covers. Work inside your own mailbox is not outside impact: the built-in email labelling routine, in Lindy's own description, "reads every incoming email and labels it, archiving what you don't need to see and keeping what matters visible," and it does that on its own once switched on. Nothing you did not approve leaves your account; a fair amount happens inside it. Security: Four things worth settling before you connect a mailbox:: - Decide whose mail this really is. If your inbox carries other people's confidential correspondence, that is the fact to weigh, and it is a different question from whether you trust the drafts it writes. - Treat archiving as the setting to watch. A label is visible and reversible; archiving moves mail out of your inbox before you have seen it. - Check the age of the binding document. Lindy's security page says your data is "never sold, never shared, never used to train models" and lists an independent SOC 2 Type II audit (an audit of how a company handles customer data) alongside GDPR and HIPAA. Its privacy policy — the document that actually binds — is dated 13 February 2024, before the product that reads your mail existed. It does commit that Google-sourced data will not be used "for any artificial intelligence model training or evaluation purposes." - Know both ways to revoke. Lindy stores connection secrets write-only — its docs say "You cannot read the value back: not in the app, not through the API, not by asking Lindy" — and offers a Revoke action. Access granted through a Google or Microsoft sign-in can also be withdrawn from that account's own security settings, which is the end that still works if you lose access to Lindy. Best at: - The recurring admin around email and meetings, which is what the built-in routines cover. - Turning a standing instruction into a scheduled job you describe once in ordinary English. - Work you want done on your own accounts rather than passed between two other services. Okay at: - Precise, branching logic. Routines are described rather than drawn, so when the outcome is wrong you rewrite a sentence rather than inspect a step. Make and n8n show you exactly what each step received and emitted. - Working alone. Everything is priced and documented around a shared Slack workspace. One person on their own still works; nothing about the design is aimed at it. Avoid: - Signing up just to have a look. Lindy's terms, last updated 14 March 2025, make all purchases non-refundable, so the minimum cost of deciding it is not for you is a full month's seat. - Relying on the changelog to tell you what changed. Lindy's published changelog currently stops at 21 October 2025, while the company's blog posts several times a week. Starter: Let it sort one morning's mail When a new email arrives → then Lindy labels it, and archives anything you have told it you do not need to see. This is the built-in routine that makes the product's boundary concrete: a Zapier or n8n flow fires only on the messages a rule matched, and structurally cannot work from the whole stream. Lindy's docs also note it does nothing at all until you have created at least one label. Start with tagging only. Add archiving to a label once a week of its choices has looked right. Starter: Describe a morning brief and let it run Lindy's Routines page carries a "Describe your routine" box that takes a sentence like this one and derives the schedule, the accounts and the output from it. On Zapier or n8n the same brief is a flow you assemble and a set of fields you map before it will run once. Change the three days to one and see how much noise that adds before you settle on a number. Prompt: Every weekday at 8am, summarise the emails I haven't replied to in three days and the meetings on my calendar today, and send me the list. Starter: Send it to a meeting you are in When a call starts on your calendar → then Lindy joins the Meet, Zoom or Teams call, records it, and files the notes where you can search them. Lindy is present in the call itself. Zapier and n8n have no seat in a meeting; the most either can do is move a transcript some other product already made, which means paying for that product too. Tell the people on the call before the first one — a recorder arriving unannounced is its own problem. Starter: Chase the emails that went quiet When something you sent gets no reply inside a window you choose → then a follow-up appears in your drafts. The window is a setting with fixed choices — Lindy's docs list 1, 2, 3, 5 and 7 days — and the bump stops in your drafts folder rather than going out. On Zapier or n8n, detecting "no reply" is logic you have to build yourself, and once built it sends unattended. Read the first few before sending. The phrasing is learned from your own sent mail and takes a while to sound like you. Starter: Write down how you want something done This is a Skill, and Lindy's docs say the description is "how Lindy decides when the skill applies" — so the tool chooses whether to use it. Nothing on Zapier or n8n decides whether to run: a flow fires when its trigger matches, every time, or it does not fire at all. Name the thing you always have to correct. That is the instruction worth writing down first. Prompt: When I ask for my weekly update, pull the numbers from the dashboard, lead with the top three changes, and keep it under a page. Pitfalls: - The seat is the unit, and mentioning Lindy creates one. Per Lindy's own FAQ, anyone who uses it takes a seat, including someone who @mentions it in Slack. In a shared channel, a curious colleague adds to the bill. - One person's big job is everyone's. The ceiling you are working against is the whole workspace's, so agree who runs the expensive things before anyone runs one. - Switching a routine on is not the same as configuring it. Labelling needs a label; the follow-up routine needs a window. A routine that is on and idle looks identical to one that is working. - It learns from your edits, so early corrections stick. Lindy's docs say it picks up on the changes you make to drafts over time. Fixing a draft carelessly teaches it the careless version. Where to next: - Workflow & Automation: https://www.launchpadhq.io/discover?category=workflow-automation - Meetings & Notes: https://www.launchpadhq.io/discover?category=meetings-notes - AI Plugins for Business Software: https://www.launchpadhq.io/discover?category=ai-plugins-business-software ### Getting Started with Litmaps Canonical: https://www.launchpadhq.io/platform/litmaps/guide Data: https://www.launchpadhq.io/data/v1/guides/litmaps.json Last reviewed: 2026-09-15 Tool: Litmaps (https://www.launchpadhq.io/platform/litmaps): starting price Free, from $10/mo; free tier yes; data as of not dated Start from one paper you trust and let the citations draw you a map of the field. You search for one paper you already trust, select it, and click Explore Related Articles. Litmaps draws a map of dots — papers — joined by lines showing which cites which. You click promising dots to add them, refresh, and the map redraws around your growing set. What it is: - Litmaps is a citation-mapping tool for academic literature, not a chatbot. You hand it papers you already have, and it shows you what they cite, what cites them, and what sits nearby in the web of references. - It draws on a catalogue Litmaps puts at over 270 million articles, assembled from Crossref, Semantic Scholar and OpenAlex — which between them cover PubMed, arXiv, bioRxiv and medRxiv — and refreshed weekly. - It is one of three tools beginners usually weigh against each other. Connected Papers builds one fixed graph around one origin paper and caps how many graphs a free account can make each month. ResearchRabbit is closest in spirit — a collection you keep growing, with alerts — and its free tier is roomier. Litmaps sits between them: a map you keep, re-run and re-sort, with emailed alerts, but on a tighter free allowance than ResearchRabbit's. - Two of those three are now related. Litmaps acquired ResearchRabbit in May 2025, and ResearchRabbit's own site describes its paid tier as powered by Litmaps' technology. They still run as separate products with separate accounts today, so choosing one does not get you the other — though with the two converging, that split may not stay where it is. Before you start: - A free account gets you in at app.litmaps.com — email sign-up, no card. That is the whole barrier; there is nothing to install. - The free tier is small, and its shape is the thing to understand before you start building. Each map is capped at 100 articles, a search will take up to 20 articles as inputs, and you get basic search only — filtering by publication date, author, journal or journal ranking is a paid feature. Litmaps' own pages disagree on how many maps you get: the main plan card on its pricing page and its help centre both say two, while a second comparison table lower on that same page says one. Plan for one map until you can see your own account. - Paying is per person and priced by country, so the figure you are quoted may be lower than the advertised list rates. Litmaps Pro has a cheaper education rate, which requires an academic email address, and a higher commercial rate; both cost less per month billed annually than billed monthly. - You need a starting paper. Litmaps works outward from something you already hold — a title you can search for, or an article from a reading list — so it is the wrong tool for the very first hour on a brand-new topic, when you have nothing to seed it with. Security: - Litmaps only ever receives the titles of published papers, never your own writing. That makes its privacy policy, rather than its handling of anything you create, the document worth reading — and Litmaps' is unusually clear. Security: What the policy actually says: - Litmaps is run from New Zealand, and its privacy policy — last modified 23 June 2026, and the version in force today — commits to both the New Zealand Privacy Act 2020 and GDPR (European privacy law governing what companies may do with your data) - It states plainly: "We do not sell your personal data" - It names the outside companies that handle your data on its behalf — hosting, email, analytics, support, payments — on a public page, rather than leaving them unlisted - It offers the standard rights: see your data, delete it, object to its use, take it elsewhere Security: - Two gaps worth naming rather than glossing over. The privacy policy says nothing either way about whether your saved articles and searches are used to improve Litmaps' own systems, so treat that as unanswered rather than settled. And the terms and conditions carry a last-modified date of 10 June 2024 — two years older than the privacy policy — so the two documents are plainly not revised together. Best at: - Working outward from a paper you already trust — the "I have one good reference, what else is out there" problem - Seeing the shape of a field at a glance: which papers everything else hangs off, and roughly when the work happened - Spotting the gap between two clusters of papers that barely cite each other - Keeping a map alive so new work comes to you by email instead of you going back to look Okay at: - Judging quality — it surfaces citation counts and journal rankings; it does not read the papers for you - Being your reference manager — it is a discovery layer, not a place to keep PDFs and reading notes Avoid: - Fields whose literature lives in books, reports or venues that publishers don't index openly. Litmaps says it is "legally only able to index Open Access metadata" — the descriptive record a publisher releases about an article — so anything without one is simply absent from the map, and nothing on the map tells you it is missing. - Counting on it for group work on the plans a beginner will meet. Litmaps' pricing table lists collaboration as "None" on both Free and Pro — team-wide collaboration sits on the Team plan, which is sales-priced with no figure published. Starter: Seed a map from one paper you trust Type the paper's title into the Litmaps search bar and hit enter, select it in the results, and click Explore Related Articles. Your starting article shows up as a dark dot; suggested articles appear as white ones, joined by lines that stand for citations. Litmaps builds outward from a seed you nominate rather than from a query, and a free account has very few maps to spend — so unlike ResearchRabbit's roomier free tier, a poorly chosen first paper costs you most of what you get. If your starting paper is very recent, try an older well-cited paper on the same topic instead — it has more citation links for the map to follow. Starter: Grow the map instead of starting over On any white dot that looks relevant, click More Like This to save it into your map, then re-run the search. The papers you just added become inputs too, so the next round of suggestions is drawn from your whole collection rather than the single paper you began with. Litmaps redraws the map each round, so the shape of the field visibly changes as you add to it — where ResearchRabbit's equivalent grows the collection without showing you that movement. Tag papers as you save them — tags are what an export or a Zotero sync works from later. Starter: Change what the axes mean Open a map's axis controls and set X and Y to any of Citation Count, Reference Count, Publication Date, Momentum or Map Connectivity. The same papers rearrange themselves — date against citations surfaces the old landmarks, connectivity surfaces the hubs. Re-sortable axes are the thing a Litmap does that a fixed similarity graph cannot. Connected Papers positions its circles by similarity and you cannot re-sort them, so "which of these is actually central" is a question only a map with adjustable axes can answer. Momentum is worth trying in a fast-moving field — though Litmaps' help page lists the axis options without defining them, so check what it measures in the app before you read much into it. Starter: Turn a map into a standing alert Save the map and switch on Monitor. Litmaps then emails you when newly published articles connect to the papers you have collected. Every account can use it, including free ones — but only paid accounts can change the settings. A free account runs on Litmaps' own default and, per the pricing table, gets a monthly summary. Monitor re-runs against the articles in your map rather than against a phrase, so a new paper reaches you because it cites what you collected, not because it matched your keywords. Connected Papers builds its graph once and has nothing to switch on. A monthly digest being too slow for your field is the clearest single reason to consider Pro. Don't count on a particular frequency: the two tables on Litmaps' own pricing page disagree — one says weekly, the other daily — and its Monitor help page offers only weekly or monthly. Starter: Get your papers back out Export the articles you have collected as BibTeX (the format most reference managers read), RIS (for EndNote) or CSV. Litmaps' help page describes export without naming any plan restriction, so it appears to be open to free accounts — though the pricing table does not list export either way, so that is a silence rather than a guarantee. The live two-way Zotero sync — where a tag in Litmaps mirrors a collection in Zotero, a popular free reference manager — is documented as a Pro feature, so on a free account the export file is the way your papers leave. Pitfalls: - The map shows citation links, not relevance. A heavily cited paper sits prominently whether or not it answers your question, and the perfect-fit paper with three citations barely registers. - Suggestions come from what your current papers cite and are cited by, so a narrow start gives you a narrow map. If everything it surfaces looks the same, add a paper from a different corner of the topic and refresh, rather than clicking deeper into the cluster you already have. - Brand-new papers are what a citation map is worst at. Citation links accumulate only after publication, and Litmaps notes a delay before new articles connect up — keep an ordinary keyword alert running somewhere else for this week's work. - Cancelling a paid plan does not delete your work, but Litmaps' help centre is clear that your access drops back to the free limits — so maps beyond that allowance stop being available to you, even though the data is still there. - Two dots sitting close together are not necessarily related. Position comes entirely from whichever axes are currently set, so a layout you changed ten minutes ago is telling you something different from the one you started with. Where to next: - More research tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Document & PDF tools: https://www.launchpadhq.io/discover?category=document-pdf-processing - AI chatbots: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with llama.cpp Canonical: https://www.launchpadhq.io/platform/llama-cpp/guide Data: https://www.launchpadhq.io/data/v1/guides/llama-cpp.json Last reviewed: 2026-09-13 Tool: llama.cpp (https://www.launchpadhq.io/platform/llama-cpp): starting price Free; free tier yes; data as of not dated The engine underneath most local AI apps — free, private, and yours to run directly if you're willing to type a command. You run one command to install it and a second to fetch a model (the AI "brain" that does the actual thinking) and start it up. Then you either type at it in a text window or open a chat page in your browser, ask a question, read the answer, and keep going. What it is: - llama.cpp is the engine that runs an AI model on your own hardware — a C++ program (that's where the ".cpp" comes from) that loads a model file and produces the words, with no company's server anywhere in the loop. Most people meet it without realising: Ollama's own README lists llama.cpp among the engines it runs on, and LM Studio runs the same model files, so both of those apps are graphical front doors onto this engine. Going direct means support for brand-new models as soon as the project ships it, and you decide exactly what runs and where the files live. What you give up is the app — there is nothing to double-click. Before you start: - Technical comfort is the real barrier here, not intelligence. Every official way to install this starts with a command in a terminal (a text window where you type instructions instead of clicking). There is no installer to double-click and no desktop app. Ollama is this same engine with a desktop app built around it, so if the terminal is a wall rather than a speed bump, start there and come back when you want the engine itself. - Then hardware, and the project makes this easy to check. Its curated model list prints a memory bracket beside every model size — the smallest entries read "requires 8 GB+ memory," the largest "requires 96 GB+ memory." Around 8 GB of free memory is the realistic floor, and the bracket, not the model's reputation, decides what you can run. On Apple Silicon Macs the graphics acceleration is switched on by default in the Mac builds; for Windows and Linux the project publishes ready-made versions for common NVIDIA, AMD and Intel graphics, so you should not need to compile anything. - You will not need to pay. There is no account, no card and no paid tier — the software is published under the MIT licence, which attaches no conditions to what you do with it, including commercial work. The project states its position plainly on its own home page: no API keys, no telemetry, no limits. What it costs you is disk space and attention. - The model files are licensed separately from the program, and they do not all match. Every model you download carries its own licence set by whoever published it. Several on the project's own list are MIT or Apache-2.0 and free to use commercially; Meta's Llama models, by contrast, sit behind a request-and-approval step and a custom community licence with conditions attached. The program being MIT tells you nothing about the model — check the model's own page before you build anything you intend to sell. - Expect to keep up with it. The project publishes numbered releases alongside new builds most days, and support for a just-released model often arrives in a build only days old. "Install it once and forget it" is not how this one behaves. Getting set up safely (There is one official source: **llama.app**, the project's own site, which carries the install command and the links to its GitHub releases under the same organisation (ggml-org) that publishes the code. Do not take a llama.cpp build from a search result, a mirror, or a link someone posted. You are installing a program that will then load large files off the internet, and the project's own published security advisories include several about maliciously crafted model files — so the chain of trust starts here and it matters.): - The program and the model are two separate downloads, and this catches nearly everyone. Installing llama.cpp gets you the program, which can do nothing on its own. The intelligence arrives as a GGUF file (one file holding an entire model — weights, vocabulary and all). Those files run from under a gigabyte to tens of gigabytes each, and the project says they are stored in the standard Hugging Face cache folder — shared with other AI tools on your machine, so they accumulate quietly and are not removed when you uninstall something else. - Confirm it worked in two steps, not one. The project's own check is to ask the installed program for its version; if it prints one, the install is fine. The real checkpoint is the next one — run one small model and get a reply back. Until that happens, you have not tested the part that depends on your hardware. - Use the project's own install and quickstart pages for the literal commands. They are short and they are kept current. Security: - There is no company in this picture at all — nobody to sign up with, nobody holding an account, nothing being reported back. The project states its own position as no API keys, no telemetry, no limits, and with the defaults left alone what you type stays on your machine. There are exactly two ways to lose that, and both are things you do rather than things done to you. Security: Two things you'd have to do yourself:: - Changing where the server listens. The project's docs say the built-in server listens on http://127.0.0.1:8080 by default — your own machine only, unreachable from anywhere else — and that --host 0.0.0.0 makes it listen on every network connection instead. Plenty of guides tell you to make that change without saying what it means: it puts a chat window onto your computer's model in front of anything else on the same network, with no password in the way. Recognise the setting and leave it alone unless you know exactly why you are changing it. - Taking model files from unknown uploaders. Anyone can publish a GGUF file, and nothing vets them. The project's own security advisory list includes several about exactly this — "Buffer Overflow via Malicious GGUF Model" and "Heap Buffer Overflow via Integer Overflow in GGUF Tensor Parsing" among them. A model file is software you are running, not a document you are opening. Stick to the repositories the project's own model list links to, and keep your build reasonably current, because those advisories are fixed in later builds. Best at: - Running on hardware other local tools skip, across Mac, Windows and Linux - Being first to a new model, because this is the engine the friendlier apps are built on top of - Being something other programs can sit on top of, using your machine instead of a company's Okay at: - Telling you where to begin. The curated model list is genuinely useful, but nothing on it is marked "start here" - Hand-holding after the install. The official documentation is a handful of pages covering the install, the two commands and the interface; there is no beginner track - Being a settled product. The friendly front door — the llama.app site and the single llama command — arrived in 2026 and is far younger than the engine behind it, so guides you find elsewhere often describe a rougher, older setup Avoid: - Reaching for the headline models on the project's own front page without reading the bracket first — the top of that list is server hardware, not a laptop - Expecting a phone app or something you can send someone a link to. The Android and iOS pieces the project publishes are building blocks for developers, not an app you install and open Starter: Read the memory bracket before you read the model name Before downloading anything, open the project's curated model list and find the size you are considering. Each entry shows two numbers: the file's size on disk, and a memory bracket. The bracket is the number that decides whether it runs at all. The project prints that bracket itself, next to every size, which turns the hardware question into a two-minute check instead of a download-and-find-out. Ollama's library pages list a download size and a context window per version but not the memory you need free, so the same check there comes down to guesswork. A file that downloads happily can still fail to load. When that happens, step down one size rather than changing settings. Starter: Your first model, in the command that also downloads it The project's quickstart has you hand llama cli the name of a Hugging Face repository. It fetches the GGUF file, stores it, and drops you straight into a conversation in the text window. Their own worked example is a shrunk-down Gemma 4, which they describe as using around 6 GB of memory. The project stores downloaded models in the shared Hugging Face cache; Ollama keeps its models in its own store under its own naming, so the same model pulled there is a second copy on your disk that nothing else can see. Starter: Turn it into a chat page in your browser The second command, llama serve, starts the model as a small server on your own machine, and the project's docs describe opening its local address in a browser to reach a chat interface built into it — one they say streams replies as they arrive, renders formatting and maths, and accepts images, PDFs and audio by drag-and-drop when the model can handle them. The chat page and the model are served by the same program you started, at an address on your own machine — which is why the same docs show other software calling that address as though it were a cloud service. In LM Studio the chat lives inside its own desktop application and the server is a separate thing you switch on beside it. If the page is blank on first load, the model is probably still reading itself into memory. Give it a minute before assuming something is wrong. Starter: Ask the same question two ways without losing the first answer The project describes its built-in chat page as letting you edit any earlier message or regenerate any reply, and says doing so forks the conversation at that point rather than overwriting it — so both versions stay reachable. In Ollama's terminal chat there is no earlier message sitting there to go back and change. Edit your original question — "explain it for someone who has never studied this" — instead of asking again underneath. The fork keeps both answers side by side. Prompt: Explain [something you're trying to understand] in plain language, then give me one concrete example. Starter: Let another program use your machine as its AI Once the server is running, other software can send its work there instead of to a company. The project walks through one pairing on its own home page: start the server, install a plugin, run a local coding assistant, and it finds your model with no keys and no configuration. The project puts that pairing on its front page as a headline use, with the claim that files stay on your machine and requests never leave it — this engine is built to be the thing other tools sit on top of. LM Studio does expose a local server, but as a feature inside a desktop app; llama.cpp publishes no app for the server to live inside. Pitfalls: - Assuming the model files disappear when you remove the program. Uninstalling llama.cpp leaves tens of gigabytes in that shared cache folder behind. - Deciding what local AI is worth from the smallest file on the list. The sub-gigabyte entries exist to prove the plumbing works, not to impress you. Move up a size before you form a view. Where to next: - Local & Open-Source AI: https://www.launchpadhq.io/discover?category=local-open-source-ai - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with LM Studio Canonical: https://www.launchpadhq.io/platform/lm-studio/guide Data: https://www.launchpadhq.io/data/v1/guides/lm-studio.json Last reviewed: 2026-07-27 Tool: LM Studio (https://www.launchpadhq.io/platform/lm-studio): starting price Free; free tier yes; data as of not dated Run an AI model entirely on your own computer, through an interface that does not require a terminal. LM Studio is a desktop app that downloads AI models onto your machine and runs them there. You browse a built-in catalogue, download a model file, load it, and chat with it in a window that looks like any other assistant — except the processing happens on your own hardware. What it is: - A desktop app for Windows, macOS, and Linux that runs open models locally with a graphical interface. - The point-and-click counterpart to command-line local runners — same underlying capability, no terminal required. - Free for personal and commercial use, with an optional paid tier for organisations needing central administration. - Completely offline once a model is downloaded, which is the reason most people choose it. Before you start: - Your hardware is the limit, not your plan. Memory is the binding constraint — roughly 8GB of RAM will run small models slowly, 16GB is workable, and 32GB or more with a dedicated graphics card is where it becomes comfortable. - Model files are large. Expect several gigabytes per model and plan disk space accordingly. - Since 2025-07-08 the app has been free for use at work with no licence request, no form, and no fee. Older articles saying otherwise are out of date. - Local models are genuinely less capable than the best hosted ones. You are trading capability for privacy, offline access, and zero per-use cost. - Nothing you type is sent anywhere. That is the feature. Getting set up safely (LM Studio official site): - Download the installer from the official site for your operating system. Model files are large — do this on a connection you are not paying by the gigabyte for. - Use the built-in catalogue rather than hunting for files elsewhere. It shows estimated memory requirements before you download, which is the number that decides whether a model will run at all. - Start with a small model. A small model that runs is more useful than a large one that swaps to disk and takes minutes per reply. - If a model loads but responds unusably slowly, it is too large for your machine. Unload it and step down a size rather than waiting it out. Security: - This is the strongest privacy posture available in this catalogue, and it is structural rather than contractual. Your conversations never leave your computer, so there is no retention policy to read, no training toggle to set, and no terms that can change next year and quietly alter what happens to your data. Security: What still deserves attention: - Model files come from a public repository, so the same provenance caution applies — check who published a model before running it. - Conversations are stored locally in plain form, which means your own device security is now the whole of your data security. - The optional local server exposes an endpoint on your machine; leave it off unless you specifically need it. - Model licences still govern commercial use of outputs, independently of the app being free. Security: - Worth stating plainly because it inverts the usual advice: with a local model, there is no vendor to trust and no policy to check. The trade is that you also have no vendor to fall back on — no uptime guarantee, no support, and capability bounded by the machine on your desk. Best at: - Working with material you are unwilling to send anywhere - Using an assistant with no internet connection - Unlimited use with no per-message cost after the download - Trying several open models without wiring up any code Okay at: - General assistant work, at noticeably lower quality than hosted frontier models - Longer documents, which local models handle but slowly - Coding help, where capability varies sharply by model Avoid: - Tasks needing the strongest available reasoning - Anything on a machine with limited memory, where the experience is genuinely bad - Expecting current information, since a local model has no web access by default Starter: Download a small model first and confirm it runs Picks a small model from the built-in catalogue, checks its estimated memory figure against your RAM, downloads it, loads it, and sends one message. The most common first experience is downloading something too large and concluding local models are useless. Establish a working baseline instead. Note the estimated memory figure before downloading. If it is close to your total RAM, choose smaller. Starter: Ask it something you would not paste into a hosted assistant This is the entire value proposition. Feeling the difference makes it concrete. Then disconnect from the internet and ask again. Watching it work offline is the moment it clicks. Prompt: [Paste the document or question you have been avoiding putting into a cloud tool, and ask your real question.] Starter: Compare two models on the same question Loads a second model of a different size or family, asks it the same question, and puts the two answers side by side. Teaches that "an AI model" is not one thing, and calibrates your expectations for what a machine your size can do. Use a question you already know the answer to. You are evaluating the model, not learning the topic. Starter: Attach a local document and ask about it Lets the model read a file from your machine without uploading it anywhere. Private document analysis is the strongest practical use case for a local setup. Start with a short document. Long ones are where limited hardware shows most. Pitfalls: - Outdated licensing information is everywhere. The app has been free for commercial use since July 2025, but many guides still say a licence is required. Check the official site rather than a comparison article. - Too large a model is the classic first mistake. If replies take minutes, the model does not fit in memory. Step down a size rather than concluding local AI is unusable. - Local means genuinely less capable. Even good open models trail the best hosted ones on hard reasoning and long context. The trade is real; know which side of it you want. - Your device is now your security perimeter. Conversations sit on your disk in plain form. Full-disk encryption and a locked screen are now part of your privacy setup. - No internet means no current information. A local model knows only what it was trained on. It cannot check anything, and it will not tell you that. Where to next: - Other local and open-source tools: https://www.launchpadhq.io/discover?category=local-open-source-ai - Hosted assistants instead: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with LocalAI Canonical: https://www.launchpadhq.io/platform/localai/guide Data: https://www.launchpadhq.io/data/v1/guides/localai.json Last reviewed: 2026-09-13 Tool: LocalAI (https://www.launchpadhq.io/platform/localai): starting price Free; free tier yes; data as of not dated Run AI on hardware you own — and put it where an app expects a paid cloud service. You install LocalAI, open a web page it serves on your own computer, pick a model (the AI "brain" that does the actual thinking) from its built-in list, and wait while it downloads. Then you type in the chat box on that same page and it answers — with nothing leaving your machine. What it is: - LocalAI runs AI models on hardware you own, and nothing you type is sent anywhere. What separates it from Ollama, the usual first stop for local AI, is what it is built to impersonate: it answers on the same API (a way for programs to talk to each other without a person clicking) shapes that OpenAI's and Anthropic's paid services use, so software already written against one of those can be pointed at your own machine instead. It also does pictures, speech and transcription from the same server, where Ollama stays with text — which is why people reach for LocalAI when they want one private stand-in for a whole cloud subscription, rather than just a private chat window. Before you start: - How well this runs is a question about the model you pick and the machine you pick it for, not about LocalAI itself. The project's own answer on hardware is that requirements "vary based on" the model's size, how much it has been shrunk to fit, and which runner it uses — they publish no minimum, which is honest but unhelpful when you're deciding. Plan for roughly the same ballpark as any other local AI tool: currently around 8 GB of spare memory to run a small model comfortably, and check what your machine has before you start (Task Manager on Windows, About This Mac on a Mac). "No GPU required" is the project's headline claim and it is true — CPU-only is a supported, tested path — but true is not the same as fast. - Which computer you have decides how much work this is. On a Mac there is a proper signed installer you double-click. On Linux there is a plain downloadable program. On Windows there is no installer at all — the project publishes no Windows version, so the route is Docker (a way of running an app in a self-contained box, so you don't install its parts by hand), which is the project's own recommended path on every platform anyway. That is a genuine extra thing to learn. If you're on Windows and you mainly want private AI chat rather than a stand-in for a cloud service, Ollama is a shorter road to the same feeling — come back here when you have something you actually want to point at it. - It is free in the way that matters: MIT-licensed, code public, no account, no card, no usage cap, and — unusually for this category — no paid cloud tier sitting next to it trying to catch you. You will never be asked to pay LocalAI anything. What it does cost you is disk and electricity, and the models are separate downloads of several gigabytes each. - It arrives empty. Installing LocalAI gets you a server and a web page with nothing to talk to; the chat page only lists models you have installed yourself. The project's quickstart names a small, CPU-friendly starter model (currently qwen3-4b) as the first thing to install, which is the sensible place to begin. Getting set up safely (Get it from **localai.io** and nowhere else. That single page links the Mac installer, the container command and the official release files — which matters more here than for a web app, because you are running a server on your own machine, and a lookalike download has the run of it. Don't take an install command out of a blog post, a video description or a search ad; copy it from the project's own install page.): - Confirm you have several gigabytes of memory free and a good deal more disk while you are still deciding. Doing this first is what separates "local AI is slow and disappointing" from "local AI is fine" — the tool cannot fix a machine that was never going to keep up. - Pick the right flavour for the computer you actually have. The project publishes a plain version and separate accelerated versions for NVIDIA, AMD, Intel and Vulkan hardware. Take the plain one unless you know which graphics card is in the machine. The wrong one either won't start or will quietly ignore the hardware you chose it for. - Know where the gigabytes go. LocalAI fetches a separate runner for each kind of model you install, so the model file is not the whole cost. Everything lands in a models folder alongside a backends folder; on the container route, follow the docs' full command rather than the short one-liner, or those folders vanish with the container. - Confirm it worked before you do anything else. Open http://localhost:8080 in a browser on the same computer, install the small starter model the quickstart names, and send it one message. A reply is the proof that the download, the runner and your hardware all lined up. If that works, everything else on this page is available to you. Security: - Privacy is not the thing to worry about with LocalAI — it runs on your own computer, so your prompts, your documents and your recordings stay on your disk, and there is no company account holding any of it. The "never paste anything sensitive into a chatbot" rule mostly stops applying, and that is the whole reason to accept the setup work. What you have to get right instead is the server, because LocalAI's out-of-the-box configuration is more open than most local tools'. Security: Two ways to lose it:: - It is reachable from your network the moment it starts, and it asks nobody for a password. LocalAI's server listens on every network connection your computer has, not just the computer itself — its default listening setting is a bare port with no address in front of it, which means "anyone who can reach this machine." And authentication is off unless you turn it on: the project's own documentation says that if you configure neither authentication mode, "the authentication middleware does not restrict requests," and describes LocalAI as private by default only once you have set a key. On your own home network behind a router that is usually survivable. On shared Wi-Fi, an office network, or anything reachable from the internet, it means a stranger can use your model, read what the server exposes, and spend your electricity. If the machine ever leaves your house, set an access key first — the project documents a single setting for it. - Where your model files come from is a decision you are making. LocalAI's own gallery is curated and open-licensed. Anything you pull from outside it is a large binary file from a stranger, held to whatever standard that stranger keeps — worth choosing deliberately rather than by clicking the first result. Best at: - Standing in for a paid cloud service on hardware you own - Covering more than chat from one server — text, images, speech, transcription and search-style embeddings on the same address - Total privacy, with no company account involved at any point Okay at: - Raw answer quality — nothing you can fit on your own hardware competes with a hosted frontier model, and LocalAI does not change that - Speed: "no GPU required" means it runs without a graphics card, not that it feels like one - First-day friendliness — the Mac installer is easy, the Windows route is not Avoid: - Being your only Windows option if you have never touched Docker — Ollama installs on Windows in one step - Reading "no GPU required" as "you don't need a GPU" — nobody is promising you a usable speed - Using it purely as a chat window. Everything that repays the extra setup is the part where something else talks to it. If a private chat box is all you want, you are doing more work than the result needs. Starter: Start with the small CPU-friendly model the quickstart names Installs the starter model LocalAI's own getting-started guide points to — currently a small Qwen3 model chosen because it runs on an ordinary processor and still supports tool use. You install it from the Models page in the web interface, which shows you an estimated download size before you commit. Which model you install changes what LocalAI installs alongside it: the project keeps a small core and fetches the runner for a model's type only when a model needs it. Starting with the plain CPU-friendly one keeps you off the graphics-card-specific builds entirely — where Ollama ships one engine and one installer regardless of what you pull, so the choice there changes the download and not the software. Starter: Point something you already use at your own machine Takes an app, script or tool that expects to talk to a paid AI service and changes the address it calls to your own computer. LocalAI answers on the same shapes OpenAI's and Anthropic's services use, so in most cases the app doesn't notice the difference. This is the specific thing LocalAI is built for, and the Anthropic half is the part that isn't available elsewhere in local AI — the project describes itself as a small core that speaks both the OpenAI and the Anthropic API, so a tool written against either one has somewhere local to land. Ollama offers an OpenAI-shaped address and nothing for the other. Starter: Add a voice or picture model next to the text one Installs a second model of a different kind — speech from text, transcription of a recording, or image generation — from the same gallery, served from the same address as your chat model. LocalAI's feature list covers image generation, speech generation, transcription, speaker separation, vision and embeddings from one server, so "one private server for everything" is achievable without installing a second tool for each job. Ollama runs text and vision-input models only, with no image-generation or speech addresses at all. Starter: Your first real message, in the built-in chat page LocalAI starts with nothing installed, so the first reply is doing double duty — it's an answer, and it's proof that the model, its runner and your hardware all matched. Asking the model to describe its own limits also calibrates you fast: a model small enough to run locally will tell you a narrower story than a cloud service would. Prompt: Explain in plain language what you can and can't help with, then give me three examples of things I could ask you that play to your strengths. Pitfalls: - The models don't inherit LocalAI's licence. LocalAI is MIT-licensed and you can do as you like with it. The models are separate works with separate terms, and the project says so itself: text-generation models "might have a license which is not permissive for commercial use or might be questionable or without any license at all. Please check the model license before using it." If anything you make with this is going to be sold or published, that check is yours to do, per model. - Assuming the model file is the whole download. Install a chat model, an image model and a speech model and you have collected three model files and three runners. Watch the folder, not the individual download. Where to next: - Local & Open-Source AI: https://www.launchpadhq.io/discover?category=local-open-source-ai - AI APIs & Developer Services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services ### Getting Started with Lovable Canonical: https://www.launchpadhq.io/platform/lovable/guide Data: https://www.launchpadhq.io/data/v1/guides/lovable.json Last reviewed: 2026-07-13 Tool: Lovable (https://www.launchpadhq.io/platform/lovable): starting price Free, from $25/mo; free tier yes; data as of 2026-10-25 Describe the app or website you want in plain English and Lovable builds a working, full-stack version you can deploy — no coding required. You chat with Lovable the way you'd chat with any assistant, but each message builds or edits your app rather than answering a question. You describe what you want, try what it made, describe the next change, and keep going until it does what you meant. What it is: - Lovable is an AI app builder — the style of tool people call "vibe coding." You describe what you want in plain language and it generates a working web app: a clickable front end, a database behind it, a login flow, and the plumbing that connects them, then deploys it. It's aimed squarely at non-developers — founders, product people, hobbyists — who want a real, editable app without writing code. And because you can export the code, it doesn't lock you in. Before you start: - The free plan is genuinely free (no card, not a time-limited trial): a small daily grant of build credits, unlimited public projects, GitHub integration, and code export on every plan. It's enough to build a small prototype; a serious build day will exhaust it. - Paid plans start with Pro, which adds more credits, private projects, and custom domains. - Know the two cost layers (building credits vs. runtime hosting/AI) before you build something you plan to keep live — a prototype is cheap, a trafficked app is not. Security: - Good news on rights: you own what you build — the app, its code, your customer data, and the AI output — and you can export the code anytime (subject to any third-party rights in the underlying AI models). Lovable doesn't hold your project hostage behind a paywall. Security: What "the AI built it" doesn't mean: - AI-generated apps can ship with security holes — missing access checks, exposed data, weak auth. Don't put a vibe-coded app in front of real users or real customer data without a security review. - Lovable processes your prompts, project context, and generated code on its servers and connects to services like the database and code host it sets up — understand where your app's data lives before you store anything real in it. - The honest ceiling: it gets you most of the way, but complex or production-grade apps usually need a developer to finish and harden them. A common pattern is to build in Lovable, then export and finish in a code editor. Security: - Because credits scale with request complexity and runtime is billed separately, an app that gets real traffic has real ongoing cost. Price the "running it" layer before you launch, not after. Best at: - Turning a plain-English description into a working, deployable web app - Prototypes, landing pages, internal tools, and MVPs, fast - Letting non-coders ship something real without a terminal - Handing you exportable code you actually own Okay at: - Complex, production-grade SaaS (it gets ~70% there; the rest needs a dev) - Predictable cost (credits scale with complexity; runtime bills separately) - Deep custom backend logic (frontend-first is its strength) Avoid: - Shipping to real users without a security review - Storing real customer data in an unreviewed prototype - Assuming the subscription is the whole cost — running the app is extra Starter: Build something small and whole A small but complete app shows the whole loop — describe, generate, click through a real working result. Swap in an idea you actually want; keep the first version small. Prompt: Build a simple habit tracker: a page where I can add daily habits, check them off, and see a weekly streak. Clean, friendly design. Starter: Iterate by describing changes The core skill is editing by conversation — small, specific requests cost fewer credits and land more reliably than one giant one. Make one change per message so you can see what each does (and what it costs). Prompt: Add a dark mode toggle in the top corner, and let me give each habit a color. Starter: Export your code Proves the no-lock-in promise and sets you up to hand the app to a developer later if it grows. Even if you never touch the code, exporting is a free backup you own. Prompt: Connect this project to GitHub so I have the full source code. Pitfalls: - Credits burn by complexity, not by message. A single "add auth and payments" request can cost as much as a dozen small edits — think in work done, not messages sent. - The subscription isn't the whole bill. Building credits and runtime hosting/AI are separate layers; a live, trafficked app has ongoing cost. - "The AI built it" is not "it's secure." Review a vibe-coded app before real users or real data touch it. Where to next: - AI Coding & Development: https://www.launchpadhq.io/discover?category=ai-coding-development - Workflow & Automation: https://www.launchpadhq.io/discover?category=workflow-automation ### Getting Started with Ludo.ai Canonical: https://www.launchpadhq.io/platform/ludo-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/ludo-ai.json Last reviewed: 2026-09-12 Tool: Ludo AI (https://www.launchpadhq.io/platform/ludo-ai): starting price Free, from $16/mo; free tier yes; data as of not dated Turn a rough game idea into concepts, characters and animated art — a tool built only for making games. You type a description of the game you have in mind into a box, and Ludo sends back concept cards — a title, a short pitch, and a piece of art beside each one. You keep the ones you like, ask for variations, and turn a favourite into a project you then add art, sound and animation to. What it is: - Ludo.ai is an AI tool made for one job: the planning stage of a video game, before anyone writes code. It generates game ideas, drafts the written design document that goes with them, and makes the art, sound, animation and 3D models the game would need. - The reason to reach for it rather than a general chatbot or art generator is that everything it produces comes out shaped for a game. Characters arrive as sprites — a sprite is a single character or object image that the game draws on screen — cut out on a clear background, ready to drop into the program you build the game in. - One thing to be clear about before you sign up, because the marketing can read otherwise: it does not build a playable game. Ludo's own FAQ asks "Does Ludo.ai create playable games or prototypes?" and answers "No, it does not." It takes you to the point where a game could be built, and stops there. Before you start: - There is no free tier here — there is a one-off trial, and it is small. You get a small one-off batch of credits (the platform's unit of spend — each thing you make costs some) when you create an account, no card asked for — by Ludo's own documentation, enough for a few dozen images, or a handful of short sprite animations or 3D models. Ludo's docs call it a free trial; its FAQ calls it the free plan; the pricing comparison table has no free column at all, and none of the three says the trial credits come back. Assume they don't. Start at ludo.ai and spend a few credits on ideas and images before anything expensive. - Producing an actual set of characters means paying. The entry plan (Indie) comes with a monthly credit allowance — granted as a full year's worth up front on the annual plan, which is cheaper per month, with unused credits rolling over while the subscription is active. The next tier up (Pro) costs more; current prices are on its Launchpad page. - Anything you download on the free account is stamped. Ludo's FAQ and its sprite documentation both say sprite sheet downloads on the free plan carry a Ludo watermark on every single frame, and that subscribing to any paid plan removes it — the FAQ says without needing to generate the art again. So the trial is for deciding, not for producing. - You have to be an adult. The Terms of Service and End User Licence Agreement in force today (Jet Play, Inc., dated 26 August 2026) states plainly: "You must be at least 18 years of age to use the Service." - It works in English. Ludo says its interface and the content it generates are in English, with parts of the website and several tool guides also available in German, Spanish, Korean, Portuguese and Chinese. No vendor page states a regional restriction — though not saying so is not the same as confirming it is available everywhere. Security: - Ludo.ai runs in the cloud, so what you type reaches its servers. The question that actually matters for a game project is a different one: who owns what comes out. On that, Ludo's friendly pages and its binding contract do not say the same thing. Security: What the two documents actually say: - Ludo's FAQ says: "The individual who initiates the generation of content using the AI platform is recognized as the full owner of that content," with the freedom to use, share or monetise it. - The Terms of Service and End User Licence Agreement (Jet Play, Inc., 26 August 2026) instead grants you "a nonexclusive worldwide license and right to use, modify, distribute and create derivative works of the Materials," for personal and commercial purposes. Commercial use is clearly permitted — but a non-exclusive licence is a permission, not ownership, and it does not stop similar output reaching someone else. - The same terms have you grant Jet Play "a perpetual, worldwide, non-exclusive, royalty-free, sublicenseable and transferable license" over what you submit to the service. Security: - Where a company's terms and its help pages disagree, the terms are the document that binds. Commercial use is explicitly allowed either way — but if you need something to be exclusively yours, read the licence rather than the FAQ and take the narrower of the two. Security: - On training: Jet Play's privacy policy, last updated 10 March 2025, says "We do not use your game concepts, designs, or other content to train AI models that would make your creative work accessible to others." Read the ending carefully — it is a promise about your work not surfacing for other people, which is narrower than a promise never to train on it. Best at: - Getting from a vague itch to a screenful of concrete game concepts, each with a genre, platform, art style and camera angle attached - Art in the shapes a game needs — sprite animations export as a PNG sheet, a sheet plus a JSON file describing where each frame sits, a GIF, a ZIP of frames or an MP4; 3D models as GLB with textures and skeleton, or plain mesh files - Grounding an idea in what is really selling, through its Top Charts view - Holding one project together — concept, art, audio and 3D in one workspace Okay at: - Writing — the design document it drafts is a good scaffold with the sections named for you, but the prose inside is generic and you will rewrite most of it - Keeping a set of characters looking alike — Ludo's advice is to pin an art style filter, reuse the same phrases and supply reference images every time, which is discipline on your part - Longer animation — Ludo's documentation warns that frame counts of 49 and above are "more prone to looping issues," and that its Transfer Motion mode caps sprite sheets at 36 frames Avoid: - Treating the charts as a reason to build something — Top Charts cannot tell you why those games sell or whether a newcomer has a route in, and a chart of incumbents is the easiest thing in the world to mistake for a market gap - Using it as evidence that your idea is original — its ideation blends one to three games you pick from a chart, so "new" means recombined from existing titles - General-purpose work — every surface is game-shaped, so a non-game task gets a game-flavoured answer and still charges you for it Starter: Describe the game you half-have in mind Ludo's idea box sits above a filter panel — platform, genre, art style and camera angle are chips you tick, so they get applied rather than interpreted, and every concept comes back with art beside it and a button that turns it into a project. That is how Ludo's documentation describes the screen; we could not watch it operate without an account. You can leave the box empty and set only the filters — Ludo's documentation says it will generate open-ended from those alone. Its guidance is to start with two or three inputs, because "too many inputs dilute the result." Prompt: A cosy farming game set on an abandoned space station, where the crops are alien and slightly dangerous. Mobile, top-down, pixel art. Starter: Make a character you can actually use Ludo's sprite documentation tells you to describe only what the character looks like and leave movement out — motion is a separate step with a library of animation presets. Midjourney will happily draw you a mushroom merchant, but it has no idea it is making a game character, so you get a picture rather than something already cut out on a clear background. Pick the art style from Ludo's style list instead of writing it into the sentence, then reuse that exact setting for every character. Prompt: A stout mushroom-person merchant with a woven backpack and a lopsided hat, facing the camera, full body, cel-shaded. Starter: Turn the idea into a written plan Two sentences produce a document already broken into the sections a game design plan needs — gameplay, the game world, art and mockups — with a version history and an assistant that can see whichever section you are editing. A first-time designer's real problem is not knowing what a design document should contain. If you already have notes, you don't have to start from a sentence — Ludo's documentation says you can import an existing PDF design document and have it build the project from that. Prompt: Mushroom merchants trade rare spores between floating islands. The player runs a market stall, sets prices, and works out which spores each island secretly wants. Mobile, top-down, cel-shaded. Starter: Read the charts before you commit Pick a store — App Store, Google Play or Steam — then a country and a genre, and read three live rankings: Top Free, Top Paid, and Top Grossing, which counts in-app spending rather than downloads. Ludo says the data is refreshed daily. Click the lightbulb icon on one to three games and press Start New Generation, and the idea panel builds fresh concepts out of that combination. This is the one part of Ludo that generates nothing — it reads live store charts, filtered by country and genre, which no chatbot can do for you. No vendor page states which plans include it, and it appears nowhere in Ludo's credit-cost table, so we cannot tell you whether it spends from your trial. Change only the country and watch the same genre rearrange itself. Starter: Ask it the question you'd ask a designer friend Opened inside a project, Ludo's assistant can see that project's sections, so "my game" means the document you already wrote. It is metered: each message costs a fixed amount of credits, more per image or asset it produces, capped at three images per request. It will also answer questions about the tool itself — "What is the Game Ideator and how do I use it?" is one of Ludo's own documented examples — a cheaper way to learn the interface than clicking through it. Prompt: I want a reason for players to come back to my mushroom merchant game every day that isn't an energy timer. Give me five options, and say which is cheapest to build. Pitfalls: - Spending your trial credits in the wrong order. Images are the cheapest thing the tool makes; video and sprite animation are billed by the second, with a minimum charge per animation. Ideas and images first, motion last. - Buying the entry plan because you saw the word unlimited. Ludo's own plan comparison lists unlimited ideation, game concept and image generation against Pro and Studio — not against Indie, where the docs say everything draws on one shared pool of credits. The same table gives Indie five active projects where Pro is unlimited, and one generation running at a time where Pro gets two and Studio five. - Re-rolling when you should be editing. Every generation returns two results by default, and the tool's real strength is the editing modes that come after — edit the image, mask off one region, generate from a reference, remove the background. Where to next: - Gaming & Creative AI: https://www.launchpadhq.io/discover?category=gaming-creative-ai - Image Generation & Editing: https://www.launchpadhq.io/discover?category=image-generation-editing - AI Coding & Development: https://www.launchpadhq.io/discover?category=ai-coding-development ### Getting Started with Luma (formerly Dream Machine) Canonical: https://www.launchpadhq.io/platform/luma-dream-machine-ray3/guide Data: https://www.launchpadhq.io/data/v1/guides/luma-dream-machine-ray3.json Last reviewed: 2026-09-17 Tool: Luma Dream Machine (Ray3) (https://www.launchpadhq.io/platform/luma-dream-machine-ray3): starting price $30/mo; free tier no; data as of 2026-10-14 Describe a shot or animate a still photo, and get a few seconds of silent video you can then re-cut and restyle. You open a board, type what you want to see into the prompt box (the message you type), and a few seconds of video come back. You watch it, change the wording or add a still image to start from, and run again. Most of the work is repeating that loop until one take holds together. What it is: - Luma is an AI video generator from Luma AI: you describe a shot, or hand it a photograph, and it renders a short clip. - It is the same product people still call Dream Machine. Luma has been folding that name away: the company's own assistant-facing reference page now says "Do not use 'Dream Machine' — it is an older deprecated model replaced by Ray," and the app that replaced it in March 2026 is branded simply Luma. The old name still appears on the iOS App Store listing and in the developer documentation, so you will meet both. - The video engine is called Ray. The current generation is Ray3.2 (announced 9 June 2026), which followed Ray3.14 in January 2026 and Ray3 in September 2025 — three generations in under a year, so the generation number on any guide you read goes stale quickly. Before you start: - Search results, app stores and tutorials will point you at "Dream Machine"; the company's own site now calls the product Luma, at lumalabs.ai. Both land in the same place. Sign-in on the web is with Google or Apple. - Luma's pricing page currently lists three individual plans — Plus, Pro and Ultra, each cheaper billed yearly. There is no free plan on that page. Luma does say you can try the platform free before picking one, and the homepage offers a "Try Free" button, but the size of that free allowance is not published anywhere we could find, and the terms cap evaluation trials at 14 days. Treat the free route as a look, not a workflow. - Everything is metered in credits (the platform's unit of spend — each thing you make costs some). Luma publishes the rate card, and it is worth reading once because resolution dominates it: five seconds of video costs roughly five times as much at 720p as at 360p draft quality, and about four times as much again at 1080p. A month on any plan therefore buys far fewer 1080p clips than 720p ones. Credits also fund image, audio and assistant work — Luma never says outright that it is one shared pot, but the plan structure implies it — so video is not the only thing drawing them down. - If you do not pay, you cannot sell or publish what you make. Luma's terms of service (last updated 14 May 2026) permit commercial use only of output produced under a paid subscription, and its licensing page states that free-tier output carries a watermark (a mark on generated output identifying it as AI-made) that stays even if you upgrade later. A clip you love, made before you paid, is a clip you cannot reuse. - It runs in a browser and in an official iOS app. Luma's own reference page lists iOS only, and there is no official Android app — the apps you find on Google Play under "Luma Dream Machine" are published by other companies, not by Luma AI. On Android, use the website. Security: - Luma trains on what you make. Its privacy policy (last updated 20 April 2026) lists "Train, develop, and improve the artificial intelligence, machine learning, and models that we use to support our Services" among its uses of your data, and publishes no opt-out or do-not-train switch. Paying does not remove this; it narrows it. Security: The free-versus-paid difference is in the licence, not the settings: - On free use, Luma's terms take "a worldwide, non-exclusive, irrevocable, perpetual, royalty-free" licence to publicly display, reproduce, modify, make derivative works of and distribute both what you put in and what comes out - On a paid subscription, that licence narrows to providing the service, generating usage data, compiling aggregated data and improving Luma's products - The asymmetry runs on the *input* as well as the output — the photograph you upload to animate is covered by the same clause as the video it produces Security: Do not upload: - Photographs of other people who have not agreed to be animated - Client footage or unreleased work you agreed to keep confidential - Anything you would be unable to explain having handed to a US company for model training Security: - Deleting a generation is described by Luma as permanent with no recovery — but deleting it from your account is not the same as deleting it from Luma's records. The privacy policy says personal information is retained for legal reasons "including after your account has been deleted," with no fixed period given. If you are in the EU or UK you have access, deletion and portability rights you can exercise directly. Best at: - Animating a still photograph you already have — one subject, one motion - Pinning several specific moments across a single clip with keyframes - Restyling footage you shot yourself while keeping the original movement and timing - Re-cropping a finished clip to a different aspect ratio without reshooting Okay at: - Storyboarding a sequence before rendering it — Luma shipped a feature for this in August 2026, though holding continuity across cuts remains the hard part in AI video generally - Readable text inside the frame — Luma's own guide encourages asking for it, but it is not reliable - Matching a specific reference style closely Avoid: - Following any step-by-step you find for "Dream Machine," including Luma's own. The Learning Hub pages are dated 14 November 2024 and describe boards, buttons and a sign-in flow from the app that was replaced in March 2026 — the vendor has not updated its own instructions across two product generations. - Turning HDR or EXR export on because it sounds like the better setting. Luma's published table charges 2× credits for HDR and 3× for HDR plus EXR, on top of the resolution price — a five-second 1080p clip at 3× eats a sizeable share of a month on the entry plan for one take. - Asking for a long or multi-shot piece in one go. Native generations are five and ten seconds; restyling your own footage tops out around twenty seconds depending on frame rate. Anything longer is several clips and an edit. Starter: Fix two moments and let it solve the middle Start a clip from one image and set a second image as a later keyframe, so the model has to get from the first to the second. Ray3.2 accepts up to 16 keyframes inside a single clip, so you are not steering with words alone. Start with just two keyframes. Add a third only once the first pair is landing. Starter: One shot, written at draft quality first The twentyfold spread between draft and 1080p on Luma's rate card is a spread you can iterate inside — settle the wording at 360p before you render anything you intend to keep. Change one thing per run — the light, or the motion, or the lens — so you can see which word did the work. Prompt: A ceramic mug on a wooden windowsill, steam rising. Morning light moves slowly across the surface as a cloud passes. The camera pushes in gently. One continuous shot, no cuts. Starter: Restyle footage you filmed, without losing the performance Upload a short clip of your own first, then paste this as the instruction. Restyling is a separate line on Luma's table — roughly three and a half times the cost of generating the same length at 720p from nothing. Shoot ten seconds on a phone with one clear movement in it. Busy footage restyles badly. Prompt: Keep the motion and timing exactly as they are. Restyle the scene as an overcast winter afternoon — bare trees, grey sky, muted colour, soft flat light. Do not change how the person moves. Starter: Re-crop a clip you already like Take a finished landscape clip and reframe it vertically, positioning the subject inside the new frame rather than cropping the centre. Reframe is the one thing on Luma's rate card billed per second rather than per clip, so a twelve-second reframe costs more than generating a fresh five-second shot. Reframe the take you were going to re-render anyway, and compare the two costs. Pitfalls: - The video comes back silent. Luma's own Ray3.2 page describes preserved audio only for restyling and reframing, and says nothing about audio when you generate from a description or a still — so expect to add sound in a separate tool. - The plan names on Luma's own surfaces do not agree with each other. The web pricing page currently lists Plus, Pro and Ultra in rising order of price; Luma's own March 2026 FAQ lists the same three prices with Plus and Pro swapped; and the iOS App Store listing still offers in-app purchases named Lite, Plus and Unlimited at different prices again. Read what you are actually buying at the point of purchase rather than trusting any one page. - The first take rarely holds — re-run at draft resolution and change the wording rather than paying for the same failure at higher quality. Where to next: - Other video generators: https://www.launchpadhq.io/discover?category=video-creation-editing - Music and sound tools: https://www.launchpadhq.io/discover?category=music-generation - Image generators: https://www.launchpadhq.io/discover?category=image-generation-editing ### Getting Started with Luminance Canonical: https://www.launchpadhq.io/platform/luminance/guide Data: https://www.launchpadhq.io/data/v1/guides/luminance.json Last reviewed: 2026-09-13 Tool: Luminance (https://www.launchpadhq.io/platform/luminance): starting price Custom; free tier no; data as of not dated Enterprise contract AI — what it actually does, and the only three ways a beginner gets near it. You open a contract, usually inside Microsoft Word, and it comes back marked up: wording that differs from your side's standard is flagged, each with a suggested edit. You accept, change or reject them one at a time, and you can type a question about the document and get an answer. What it is: - Luminance is contract AI built by a Cambridge, UK company and sold to law firms and to the legal, procurement and compliance teams inside larger businesses. - It is a suite rather than one tool: separate products for generating contracts from templates, marking up incoming ones, searching everything the company has already signed, running compliance checks on counterparties, and sifting documents for disputes and investigations. - If you ever use Luminance, it will almost certainly be because your employer bought it. Before you start: - Nothing about Luminance is hard to use; the hard part is getting a login at all. Luminance publishes no prices anywhere — every product page ends in the same "Request a demo" form asking for your name, work email, company and role — so the honest answer to whether you will need to pay is that someone will, in a negotiated order form, and it will not be you personally. Luminance's terms do contemplate evaluation orders alongside commercial ones, so an assessment period may be available, but nothing on the public site offers a trial and you would have to ask. - Licensing is counted in documents rather than people. The order form sets a maximum number of documents you can hold, and the terms add a yearly cap of twice that number in uploads, counted even for documents you later delete. The limit you actually run into is how much you feed it, not how many colleagues have logins. - The cross-contract question feature, Ask Lumi Pro, is metered in credits (Luminance calls them LumiTokens) — the platform's unit of spend, where one credit buys one answer about one document. Ask a single question across fifty contracts and you have spent fifty. The pool refills on a rolling seven-day window and is normally the same size as your document allowance. - If you want contract AI you can open today rather than next quarter, Genie AI currently has a genuinely free tier you sign up for yourself, with limited AI usage and exports locked until you upgrade. It is a reasonable place to learn what these tools do while you wait for a rollout. - Students have one other door: Luminance runs a certification programme that is free of charge to universities and law schools, with partners including King's College London, Vanderbilt Law School and UC Law SF. Worth asking whether yours is on the list, because it is the only hands-on route that does not run through a procurement team. Security: - The architecture is the answer Luminance leads with: each customer gets a dedicated, single-tenant instance on AWS rather than sharing a database with other firms, and Luminance says its own staff cannot open your documents without you granting that through the interface. It holds ISO 27001:2022 certification and has completed a SOC 2 Type 2 examination — an audit of how a company handles customer data. Instances are backed up nightly to a second AWS data centre in the same region, retained at least 14 days. Where your documents physically sit is fixed in the order form, not chosen by you: UK or EU hosting if that is what was bought, otherwise an international location. Security: The clause to read before you type into the chat: - Luminance's Master Hosted Terms, updated 10 October 2025 and in force, cover features that route through outside AI models — Ask Lumi among them, with OpenAI, Microsoft Azure and Google Cloud named as processors - For those features the terms state that redaction "may not be possible" for the actual words entered as a query or a requested redraft, and that Luminance is not responsible for confidentiality or intellectual-property problems arising as a direct result - The mitigation is real and worth knowing: your administrator can switch those features off by raising a support ticket, and the terms say so explicitly - The practical rule: the document you upload is handled under the single-tenant arrangement above; the sentence you type into the chat box may not be Security: - Two exits worth knowing before you rely on it. Luminance describes deployment inside a customer's own environment as an alternative to its cloud, though it publishes no detail on what that involves — a question for whoever runs the evaluation. And when a subscription ends, the retention period is thirty days — from the end of the term or from notice of termination — during which you can download or delete everything; after that Luminance deletes it and says it cannot be recovered. Anything you keep only in Luminance is on that clock. Best at: - Marking up an incoming contract against language your side has already approved, without leaving Microsoft Word - Answering "which of our contracts say X" across an entire signed archive, rather than one document at a time - Generating a standard contract from a pre-approved template so Sales or Procurement can produce it without routing through Legal - Reading documents in languages it was never specifically configured for — Luminance says its AI reads any language rather than working from a configured list, which is the durable claim; the specific language count it once published is several years old, so treat any number you see as unverified - High-volume, low-variation paperwork — confidentiality agreements are the case the vendor leads with Okay at: - Grounding answers in actual law. The LexisNexis alliance that would put case law, statutes and citation checks inside Luminance was announced in April 2026 and is currently a waitlist for mutual customers, so today the system reasons about your contracts, not about the law around them - Bespoke, heavily negotiated one-off agreements, where the "what have we agreed before" signal it runs on is thinnest Avoid: - Bulk-uploading everything you can find to see what turns up — every upload counts against the yearly cap whether or not you delete it afterwards - Reading a clean pass as a safe contract. It reports deviation from the standards it was given, so a risk nobody wrote into those standards is invisible to it Starter: Find out what's actually different about this one Ask Lumi answers against your organisation's own prior negotiations, templates and agreed standards — Luminance describes its suggested alternatives as "based on context of prior negotiations, templates and playbooks" — which is why asking it to rank the pushbacks returns something usable. Kira (Litera) returns the provisions it has been trained or taught to extract, into a diligence report; there is no store of your side's negotiating history for it to order a pushback list from. Prompt: Compare this agreement against our standard position and give me a table of every clause where the wording differs, what each difference means for us commercially, and which three you would push back on first. Starter: Ask for the redraft and the fallback in one go The answer arrives as a mark-up inserted into the Word document rather than as text you copy across — Luminance's phrase is "surgically inserting a mark-up" — and the fallback is drawn from language your organisation has previously accepted in the same system. Robin AI will also draft you a revision; what it cannot do is source "one fallback we could live with" from your own executed contracts, because that history lives in Luminance's repository, not in the drafting assistant. Prompt: Our position is that the liability cap should be twelve months of fees. Redraft clause 9 to that position, keep the clause's existing structure, and give me one fallback wording we could live with if they refuse. Starter: Ask one question of everything you've ever signed This is the Ask Lumi Pro path, and it is worth pricing before you run it: this question across a 400-contract repository is a 400-credit question. Kira runs diligence across a project set you define and bills nothing per answer, so the "is this question worth asking" arithmetic simply does not exist there — it is specific to how Luminance meters Ask Lumi Pro. Prompt: Across all our supplier contracts, which give the other side a right to terminate for convenience on less than 60 days' notice? List the contract, the clause reference and the notice period. Starter: Let it negotiate the boilerplate without you in the loop Hand a high-volume, low-variation agreement to Luminance's autonomous negotiation and it reviews the document, applies your standards, redlines, sends revisions to the counterparty and reacts to their replies in a machine-to-machine exchange — surfacing the reasoning and the prior contracts behind each decision. Your job is setting the standards up front and reading the log afterwards. In March 2026 Luminance opened this to designated users across the business rather than Legal alone, while Legal keeps governance — so it is one of the few places you will meet an agent (AI that takes actions on its own rather than only answering) pointed at a counterparty instead of at your own document. Robin AI's assistant produces the mark-up for a person to send; the send-and-respond loop itself is what differs here. Starter: Generate the contract instead of reviewing one In Draft you work through a step-by-step walkthrough of conditions and it assembles a finished contract from a pre-approved template, so someone in Sales, Finance or Procurement can produce a compliant agreement without waiting on Legal to write it. The comparison here is not review quality but who is allowed to press the button — the product's own listed use case is "Self-Serve Contract Generation." Kira is a review and extraction tool with no generation side at all. Pitfalls: - On Ask Lumi Pro, credits are spent as you scroll, not when you ask. The terms say tokens are consumed as you page through results — so idly scrolling to the bottom of a fifty-document answer set costs fifty where reading the top five would have cost five. Read from the top and stop. - The Word experience and the browser experience are different products. Contract mark-up lives in Negotiate, inside Word; repository search and cross-contract questions live in Analyze. Which of Luminance's products you get is set in the order form your organisation signed, so what you can see once you log in may be narrower than what the marketing pages show — assuming you have both is the standard first-week disappointment. - Ask Lumi is opt-in, not on by default — the terms describe it as an interface the customer selects within the product. If you cannot find a chat box, the likeliest explanation is that nobody switched it on for your instance, not that you are looking in the wrong place. - "Legal-Grade" is Luminance's own trademarked phrase, not an external standard. It describes the company's multi-model architecture and its own benchmarking; no third party certifies it. Treat it as a product name rather than an assurance. - Acting on an answer as though it were advice. Luminance's public Lumi Go terms state that outputs "are not intended to be, do not constitute and must not be used as, legal advice or a substitute for obtaining independent legal advice" and that no lawyer-client relationship is created. Those public terms carry no date on the page, so it is worth checking them again at the point you are asked to accept them. Where to next: - Legal AI tools: https://www.launchpadhq.io/discover?category=legal-ai - Document & PDF processing: https://www.launchpadhq.io/discover?category=document-pdf-processing - Research tools: https://www.launchpadhq.io/discover?category=research-academic-tools ### Getting Started with MagicSchool Canonical: https://www.launchpadhq.io/platform/magicschool/guide Data: https://www.launchpadhq.io/data/v1/guides/magicschool.json Last reviewed: 2026-10-23 Tool: MagicSchool (https://www.launchpadhq.io/platform/magicschool): starting price Free, from $8.33; free tier yes; data as of 2026-10-14 A library of AI tools built for teachers — lesson plans, quizzes, rubrics, rewritten texts — plus safe, teacher-controlled AI rooms for students. Instead of one blank chat box, MagicSchool gives you a menu of single-purpose tools. You pick one — say, the Lesson Plan Generator — fill in a short form with your grade, topic and goals, and it drafts the material. You then refine it with Raina, MagicSchool's built-in AI assistant, and export it to Google Docs, Slides, Word or PowerPoint. What it is: - MagicSchool is an AI platform made for schools. Its site lists more than 80 tools for teachers and more than 50 for students. Each tool does one job — writing a lesson plan, building a rubric, making a multiple-choice quiz, generating a worksheet, rewriting a text at a different reading level — and asks for the details that job needs, so you don't have to work out how to prompt a general chatbot. - Raina is the assistant that runs through it. MagicSchool describes it as working inside every tool to rewrite, differentiate, add questions or make variations of what you just generated, and its September 2026 update says it now uses the grade and subject you saved at sign-up to tailor answers. - The student side is called MagicStudent. Teachers create a Student Room, choose which AI tools students can use, and monitor the activity. Students join with a link, code or QR code — or through Google Classroom — and MagicSchool's help centre says no student account is needed to join a room. Students can't chat with each other. - Finished work can go where you already teach: presentations export to Google Slides or PowerPoint, formatting carries over into Google Docs and Word, and multiple-choice quizzes can export to Google Forms. Deeper connections to Canvas, Schoology, Clever and ClassLink are, per the help centre, for Enterprise (school or district) customers only. Before you start: - It's for educators, free to start. You sign up as "An Educator" with an email and password, or with Google or Microsoft. MagicSchool's terms say educators use it as a professional aid for their work at a school, not for personal use, and must not let students or other minors use their teacher account. - Free means a weekly cap. MagicSchool's help centre says the free plan includes a set number of AI generations per week; every tool output, refinement and Raina reply counts as one, while student use doesn't count. The count resets every Sunday at midnight UTC. Some features, such as Class Writing Feedback, have their own separate limits. - Plus removes the weekly cap. Plus is a paid plan billed monthly or, more cheaply, annually; current prices are on its Launchpad page. Plus adds unlimited generations, output history, and the full set of student tools. Schools and districts can request Enterprise pricing, which adds a data privacy agreement, single sign-on and LMS integrations. - Refunds have a short window. MagicSchool's FAQ says an annual subscription can be refunded only if you ask within 14 calendar days of your most recent annual payment; after that it's non-refundable. Monthly plans have the same 14-day window from the most recent monthly payment. Security: - The rule that matters most: don't type students' personal details into MagicSchool — names, ID numbers, health or behaviour records — unless your school has a data privacy agreement with MagicSchool that covers it. Its terms ask users not to provide students' personally identifiable information and say breaking that can get an account suspended. Write "Student A" instead of a name; the tools work just as well. Security: What MagicSchool says about data and safety:: - Its homepage says it doesn't use student or teacher data to train AI, and its help centre says the AI providers it uses are contractually barred from training on user data. - It describes itself as SOC 2 certified and FERPA/COPPA compliant, with iKeepSafe certification for both; details are on its Trust Center. - In Student Rooms, unsafe language is flagged to the teacher's dashboard, and teachers can pause or lock a room or remove a student. Since September 2026, student chat threads close after about 50 messages and students are reminded they are talking to an AI. Security: - MagicSchool's terms are blunt that output can be incorrect or "fictitious" and must be reviewed. Check facts, dates and answer keys yourself, and watch for bias, before anything reaches students. Best at: - Turning a topic and a grade level into a usable first draft of a lesson plan, worksheet or quiz in a couple of minutes. - Differentiation — rewriting the same text or task at several reading levels, or making variations for different groups. - Giving students AI access in a room you control, with tools you've chosen and activity you can see. Okay at: - Accuracy in specialist subjects. Drafts are a starting point; answer keys and facts need checking. - Heavy daily use on the free plan — the weekly allowance goes quickly once you start refining. Avoid: - Entering real student names or records without a school data agreement. - Handing generated material to students without reading it first. Starter: Plan tomorrow's lesson Paste this into the Lesson Plan Generator, one of MagicSchool's most-used tools. Giving a time, a learning goal and the parts you want gets a plan you can actually teach from, rather than a generic outline. Swap in your own grade, subject and standard. Prompt: Grade 7 science, 45-minute lesson on the water cycle. Students should be able to explain evaporation, condensation and precipitation in their own words. Include a 5-minute warm-up and a quick exit ticket. Starter: Level a reading passage Use the Text Rewriter — one of MagicSchool's most-used teacher tools in 2025. It's the fastest way to give the same content to students reading at different levels. Prompt: Rewrite this passage for a 4th-grade reading level. Keep every key fact, use shorter sentences, and define any word a 9-year-old might not know: [paste your text] Starter: Build a quick check for understanding Use the Multiple Choice Quiz tool, then export to Google Forms if you use it. Check every answer in the key — this is where AI mistakes cost you most. Prompt: 10 multiple-choice questions on the causes of World War I for 10th graders, with one clearly correct answer each and plausible wrong answers. Include an answer key. Starter: Refine instead of starting over Ask Raina this after any tool gives you a draft. Refining keeps what already works — though on the free plan each refinement counts against your weekly generation allowance. Prompt: Make this easier for students who are still learning English, and add a short word bank at the top. Starter: Test a Student Room yourself first Create a Student Room with one tool, then join it yourself and type this as a student would. MagicSchool's help centre recommends testing rooms before students use them, so you see exactly what they'll see. Prompt: Explain photosynthesis to me like I'm in 6th grade, then ask me three questions to check I understood. Pitfalls: - Burning free generations on small tweaks. Every Raina refinement counts; batch your changes into one request. - Assuming integrations come with a teacher account. Canvas, Schoology, Clever and ClassLink connections are Enterprise-only. - Using your own account for students. Students should join through a Student Room, not log in as you. - Forgetting a yearly plan renews. Note the date — refunds are only possible within 14 days of a payment. Where to next: - More education & learning AI: https://www.launchpadhq.io/discover?category=education-learning-ai - Research tools for building lessons: https://www.launchpadhq.io/discover?category=research-academic-tools - Image tools for classroom visuals: https://www.launchpadhq.io/discover?category=image-generation-editing ### Getting Started with Magnific (formerly Freepik) Canonical: https://www.launchpadhq.io/platform/freepik-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/freepik-ai.json Last reviewed: 2026-10-15 Tool: Magnific (formerly Freepik) (https://www.launchpadhq.io/platform/freepik-ai): starting price $20/month; free tier yes; data as of 2026-10-17 Freepik's library of ready-made stock pictures and its AI tools for making and enlarging images and video, now under a new name — with one licence for downloaded stock and a different one for what the AI makes. You open the image or video generator in your browser, type a prompt (the message you type) describing what you want, and pick which AI engine to use from a list. A few seconds later your results appear. You keep the ones you like, adjust the wording or settings, and run it again. What it is: - Magnific is the new name for Freepik, the Spanish company that started as a library of ready-made stock images — photos, illustrations, icons and templates you can download and use. It announced the rename in April 2026. freepik.com now forwards you to magnific.com, an old Freepik login still works, and many guides online still use the old name, so if you see "Freepik" in a tutorial it means this site. - Alongside the stock library sits a suite of AI tools: an image generator, a video generator, an upscaler (which enlarges a picture and sharpens its detail), a background remover, an image editor, plus sound-effect and music generators. The upscaler is where the name comes from — Magnific was an upscaling startup Freepik bought in 2024. - The thing that sets it apart is choice of model (the AI "brain" that does the actual thinking). Rather than one house engine, Magnific puts several behind one account: its own Magnific One, plus third-party models such as Google's Nano Banana 2 for images and Kling 2.5 for video. You pick per job, and all of them draw on the same account. - It runs in the web browser, and Magnific also lists a desktop app and a mobile app. Before you start: - Sign-up is self-serve on magnific.com — you register with an email address, or use a Google login. If you already had a Freepik account, it carries over with the same email and password; you do not need a new one. - The AI tools run on credits (the platform's unit of spend — each thing you make costs some). Paid plans are currently Premium at $20/month, Premium+ at $45/month and Pro at $110/month, each cheaper per month if billed annually. Magnific's terms let it change how many credits a tool costs "at its sole discretion and without prior notice," so check the cost shown next to the button before you run something expensive, especially video. - The licence on what the AI makes depends on whether you are paying at the moment you make it. Magnific's AI Products Terms (June 2026) say that a subscriber owns an image or video it generates, provided the subscription "remains active at the time of generation." Without a paid subscription you get only a personal, non-commercial licence to use it, which Magnific can revoke. So anything you might use for a business, a client or a product should be generated while subscribed — upgrading later does not, on the terms' wording, convert things you made earlier. - Downloaded stock follows a different rule. Magnific's Terms of Use (June 2026) let you use stock content without paying on condition that you credit it; its help centre gives the credit line as "Designed by Magnific," linked to magnific.com, placed next to the image or, failing that, in the footer or credits. A Premium, Premium+ or Pro subscription removes the credit requirement for stock you download while subscribed. Crediting the stock does not change the AI rule above: a credit line is not what makes an AI image usable commercially. Security: - Magnific is a Spanish company working under EU privacy law, and its AI terms are firmer than most on one point: Magnific's AI Products Terms (June 2026) say it 'will not use Inputs (including the User's image captured when using certain AI Products) or Outputs to train its own AI models.' What you make is private by default — the same terms say your prompts and results stay confidential unless you choose to share them publicly, at which point other users may use them under the platform's terms. Security: What the April 2026 privacy policy adds: - Photos and audio you upload as a starting point are, per the policy, deleted immediately after the AI has used them. Your prompts and results are kept for as long as you have an account. - When you pick a third-party model, your upload is processed by that company, which the policy says keeps it for up to 30 days. The policy also says some models are hosted in countries without EU-level privacy protection, and that choosing one is your decision — so the model menu is also a choice about where your data goes. - Magnific still studies how prompts, results and your feedback relate in order to improve its tools. The policy says you have the right to object to this; it does not say this covers uploaded faces or voices, which it says it never uses for training. - Uploading another person's face or voice is on you. The AI terms require you to have permission from anyone whose image, voice or identity is in what you upload. Best at: - Trying the same idea on several image and video models without opening a separate account for each. - Making a small or old image bigger and sharper — upscaling is the company's original product. - Projects that mix stock and AI: find a licensed photo or icon, then generate or edit around it, in one place. Okay at: - Getting one exact picture you already have in mind. Results come back as several attempts, and getting it right usually takes a few rounds of rewording. - Video. It works, but each try spends credits whether you keep it or not. Avoid: - Using anything you generated without a paid subscription in a business, client job or product. The AI terms limit those results to personal, non-commercial use. - Counting on Magnific's legal cover for AI images. Its Terms of Use say the cover it offers when someone claims a stock download infringes their rights does not apply to anything made with its AI tools. - Printing AI images on merchandise for sale under the Pro merchandise licence. Magnific's help centre says AI-generated resources are excluded from it; that licence covers stock files only. Starter: One prompt, two engines Run it once on Magnific One and once on Nano Banana 2 and compare. Choosing the model is not only a look: Magnific's AI terms fold each third-party provider's own terms into yours, and its privacy policy says the provider you pick processes your upload. On Midjourney there is one engine and one set of terms; here the menu decides both. Change one thing at a time — the lighting, then the style — so you can see what each word is doing. Prompt: A ceramic coffee mug on a wooden kitchen table by a window, soft morning light, shallow depth of field, photorealistic, no text. Starter: Rescue a small photo with the upscaler Upload a low-resolution image — an old phone photo, a thumbnail — to the image upscaler and get back a larger, sharper version. This is the tool the brand is named after. Unlike a standalone upscaler such as Topaz, it sits in the same account as the generator, so you can enlarge something you just made without downloading and re-uploading it. Try it on an image you generated in step 1, then on a real photo, and compare how much detail it invents in each. Starter: Turn a still into a short clip Take the image you kept from the first exercise into the video generator with Kling 2.5 selected. Going to Kling directly gets you Kling only; here the still and the clip come from one account and one credit balance, so you can see how far a single image goes before you pay for many video attempts. Keep the movement to one simple action. Asking for several things to happen at once is where clips fall apart. Prompt: Slow push-in toward the mug, steam rising gently, morning light shifting across the table. Calm, no camera shake. Starter: Search the stock library before you generate Type what you need into stock search first — 'coffee mug on table' — and only generate if nothing fits. On Magnific the two kinds of picture carry different rights. A stock photo has a known author licensed through Magnific, and it is what Magnific's legal cover applies to; an AI result has no such cover, and on the AI terms it may not be unique. Midjourney has no stock library to fall back on, so it cannot offer you that choice at all. Pitfalls: - Old 'Freepik' tutorials may be out of date. Screenshots, menu names and plan details in them may have changed since the rename — trust what magnific.com shows you. - Your AI result is not exclusive. Magnific's AI terms say it does not guarantee that the same output cannot be generated again by other users. Fine for a social post; a problem if you wanted a logo or a character nobody else could have. - Credits do not all carry over. The terms say plan credits do not roll into the next billing period. Extra credit packs you buy on top do roll over, but are cancelled with no refund if you cancel your subscription and go back to an unpaid account. - Subscriptions renew automatically. You can cancel renewal at any time and keep access until the end of the period you paid for. The Terms of Use say no right of withdrawal applies to digital content beyond what consumer law guarantees, and the pricing page describes refunds differently — so check the current refund wording before you count on getting money back. - Some uses are banned outright. The AI terms prohibit political content for election campaigns, sexual or shocking material, passing an AI result off as human-made to deceive people, and using results to train another AI. From August 2026 they also require you to clearly label deepfakes and AI content meant to inform the public. Where to next: - Image generators: https://www.launchpadhq.io/discover?category=image-generation-editing - Video tools: https://www.launchpadhq.io/discover?category=video-creation-editing ### Getting Started with Mailchimp AI Canonical: https://www.launchpadhq.io/platform/mailchimp-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/mailchimp-ai.json Last reviewed: 2026-09-09 Tool: Mailchimp AI (https://www.launchpadhq.io/platform/mailchimp-ai): starting price See pricing; free tier yes; data as of not dated Email marketing you run from a browser — with AI features scattered through it, most of which begin at Mailchimp's paid plans. You work in a browser dashboard. You collect email addresses into an audience (Mailchimp's word for your contact list), build a message by dragging blocks of text and images into place, then either send it now or set it to go out on its own when something happens. Afterwards you see who opened it. What it is: - Mailchimp is an email and text-message marketing platform. It has been the default first choice for small businesses and newsletter writers for well over a decade, and it is owned by Intuit, the company behind QuickBooks and TurboTax. - There is no single product called "Mailchimp AI". What exists is roughly a dozen separate features scattered through the app under their own names — Write with AI for copy, Creative Assistant for on-brand layouts, Send Time Optimization and Content Optimizer for tuning sends, Analytics AI for asking questions about your results — several of them badged Intuit Assist, the assistant Intuit puts across its products. - "Does Mailchimp have AI" is the wrong question to ask: each feature carries its own plan requirement, and several carry a country list and a beta label on top of that. Before you start: - Signing up is free and needs no card. The free plan gives you a small contact limit, a monthly send allowance with a lower daily ceiling, one audience and one seat. That is enough to send a real newsletter to a small list, and Mailchimp's help centre notes that going over the contact limit stops live sending until you upgrade or cut the list. - The generative AI is not on the free plan, and not on Essentials either. Mailchimp's plan comparison carries a single row called "Generative AI Features" and marks it not included on both Free and Essentials, reaching Standard and Premium as a no-additional-cost add-on. Send Time Optimization is a separate row and is likewise not included on Free or Essentials. So Creative Assistant, Write with AI, Content Optimizer, Analytics AI and the ChatGPT and Claude connectors all effectively begin at Standard — the "Essentials plan or higher" line you will find on Mailchimp's Creative Assistant help pages governs branded templates through Brand Kit, not Creative Assistant itself. Paid marketing plans are priced by how many contacts you hold rather than how many emails you send, so the cheapest tier is the one for the smallest list. - There is a 14-day trial of Standard, advertised without a card on some versions of the offer. Mailchimp's promotional terms restrict it to new customers, or to free-plan accounts created within the last 90 days, in select countries, and it rolls into a paid month at the then-current rate unless you downgrade. That is the cheapest honest way to see whether the AI is worth the tier. - Write with AI is available only to accounts whose primary mailing address is in Australia, Canada, the United Kingdom or the United States, and Mailchimp describes it as a beta released to a small percentage of users. If you are outside those four countries, plan the page's copy recipes around writing your own text. Security: - The difference from a chatbot: the sensitive material is not your prompts, it is a file of names and email addresses belonging to your customers. Mailchimp's data processing addendum puts you in the controller's seat and Mailchimp in the processor's, which means the duty to have collected those addresses properly — and to answer for them under GDPR (European privacy law governing what companies may do with your data) if any of them are European — sits with you, not with Mailchimp. Security: Three things follow from that:: - Only upload people who asked. Mailchimp's acceptable use policy says you must be able to point to an opt-in form or other evidence of consent for any marketing email, and it bans purchased, rented, third-party, co-registration and publicly available lists outright. Where it identifies abuse it says it may suspend, throttle or disable the account. - Treat the login as the list. One password stands between anyone and an exportable copy of every contact you hold. Turn on two-factor authentication before your first import, not after your first campaign. - Know where the data goes. The addendum says Mailchimp may transfer and process customer data in the United States and anywhere else it operates, and Mailchimp's privacy FAQ says it may share information with its parent Intuit and other Intuit companies to develop and improve other Intuit products. Security: - The beta terms Mailchimp publishes for its AI-assisted CRM say it may use your inputs and outputs, including customer data, for machine learning to develop and improve the product, with an opt-out in account settings. The Write with AI help page points instead to a separate "Content Generator Beta" terms document that is not published alongside it, so whether that same machine-learning wording covers the email copy generator is worth checking in your account settings when you switch the feature on. Best at: - Sending one well-designed email to a list of people who asked for it, and seeing afterwards who opened and clicked. - Switching on a standard sequence from a named template rather than drawing the flow yourself. - Sitting between an online shop and its customers: connect Shopify, WooCommerce, Squarespace Commerce, Stripe, BigCommerce, Magento or Square and the store data feeds both the automations and the reporting. Okay at: - Sounding like you. Write with AI leans on whatever you have put in your Brand Kit, so it reads generic until that is filled in. - Being the place your customer records live. Mailchimp's own legal terms still carry its customer relationship management product as a beta. - Answering questions about results rather than charting them. Analytics AI does that conversationally, but it launched in May 2026 and Mailchimp frames it around accounts with store data connected. Avoid: - Emailing a list you bought, rented, or exported from somewhere else — this is the fastest way to lose the account entirely. - Sending from a gmail.com or yahoo.com address. Mailchimp says free email addresses cannot be added to your verified domains, and it rewrites your From address onto a Mailchimp subdomain so that it matches what actually sent the message. - Expecting Send Time Optimization to tune your automations. Mailchimp's help page limits it to regular, plain-text, A/B and multivariate emails, needs a send date at least 48 hours out, and falls back to a manual choice when it has too little history to work from. Starter: When you cannot settle on a subject line → get it scored as you type As you type the subject line, Mailchimp rates it on length, word count, punctuation and emoji use, drawing on patterns across the volume it sends. The subject line helper sits outside the "Generative AI Features" bucket that starts at Standard — it is its own row in Mailchimp's plan comparison — so it is the one recipe here that does not need the AI tier. Which plan it reaches is genuinely contested: Mailchimp's feature page says it is "free with every Mailchimp account", while the plan comparison table marks it not included on Free and included from Essentials up. Treat the comparison table as the one to plan around, and expect to find out which is true at the compose screen. Take the emoji advice with a pinch of salt on a business list — the scoring reflects Mailchimp senders in general, not the people on your list. Starter: When someone new joins your list → welcome them without you being there Open Automations, pick the flow template for welcoming new contacts, choose your audience, edit the email, switch it on. The signup itself is the trigger (the "when" — the event that starts an automation); Mailchimp calls it the starting point. Essentials caps a marketing automation flow at four steps, which happens to be exactly the shape of a welcome sequence — so this is the one automation you can grow into on the cheapest paid tier without paying for Standard's 200-step ceiling. Rather than building a second flow later, add a delay and a follow-up email inside this one while you still have steps spare. Starter: When you have written a rough paragraph → ask the editor to shorten it or shift its tone Inside a Paragraph, Heading or Button block in the email builder, Write with AI takes your prompt or one of its preset suggestions and returns a rewrite you accept or discard. Mailchimp also puts the same tool in the message box of the SMS builder, so one recipe covers cutting a paragraph you already like down to text-message length — the rewrite is anchored to copy you wrote rather than starting from a blank prompt. If it is missing on a plan that should have it, that is the staged beta rollout rather than a setting you got wrong. Starter: When a shopper abandons a cart → let the AI draft the whole recovery sequence With a supported store connected, Mailchimp's AI-powered flow templates generate branded emails for nine standard sequences — abandoned cart, price drop on a cart item, birthday, sign-up anniversary, win-back and more — pulling design from your Brand Kit. You pick the audience, review what it wrote, then activate. Three conditions have to be true at once — a Standard plan or higher, a US primary mailing address, and one of Mailchimp's supported store connectors — and Mailchimp's help page names all three, so you can rule this in or out before you commit to a contact tier. No store, or not in the US? The same sequences exist as ordinary flow templates in the same browser, with the timing pre-set and the writing left to you. Starter: When you want to know why last month worked → ask, instead of building a report Analytics AI answers questions about campaign performance, audience behaviour and revenue in conversation, reading your connected store data alongside your send history. Mailchimp launched it to remove the export-and-pivot step, so the answer comes back in revenue terms rather than as an open rate you then have to interpret. Mailchimp also publishes connectors for ChatGPT and Claude on the same tiers, so you can ask the same kind of question from whichever assistant you already have open. Pitfalls: - Unsubscribed contacts carry on costing you. Mailchimp counts subscribed, unsubscribed and non-subscribed contacts toward the plan limit you pay for; only cleaned, archived, pending and deleted ones stop counting. Archive people who have opted out, or you will pay a growing bill for a list you are not allowed to email. - Authenticate your domain before your first real send. This means adding a couple of records at whoever manages your web address — Mailchimp's Entri tool can do it for you — and it is not optional at any scale: Gmail and Yahoo require it of anyone sending more than 5,000 messages to their addresses in a 24-hour period, and Mailchimp says authentication only works on a domain you own. - Check what the AI wrote before it goes out, because an email cannot be recalled. Mailchimp's beta terms disclaim any warranty as to the accuracy of generated content and tell you to satisfy yourself that an output is accurate, lawful and appropriate before adding it to a campaign. A wrong chat answer is a nuisance; a wrong price in a send to your whole list is a day of apologies. - The free plan's daily ceiling catches people out. The monthly send allowance sounds generous, but a single send to a full free-plan list uses the whole day's allowance, so testing and sending on the same afternoon can stall. Where to next: - Sales, marketing and SEO tools: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai - AI built into the business software you already run: https://www.launchpadhq.io/discover?category=ai-plugins-business-software - Connecting your apps to each other: https://www.launchpadhq.io/discover?category=workflow-automation ### Getting Started with Make Canonical: https://www.launchpadhq.io/platform/make-integromat/guide Data: https://www.launchpadhq.io/data/v1/guides/make-integromat.json Last reviewed: 2026-07-27 Tool: Make (Integromat) (https://www.launchpadhq.io/platform/make-integromat): starting price Free, from See pricing; free tier yes; data as of 2026-10-25 Visual automation that connects your apps — and a meter that counts differently from its best-known rival. You build on a visual canvas, dragging modules onto it and drawing the path between them. A scenario starts with a trigger — a new form response, an incoming email — and then runs each module in turn. You run it once by hand, open the execution to see what each step received and sent, and adjust. What it is: - A no-code automation platform connecting thousands of apps through a visual builder. - Best suited to workflows with real logic — branches, filters, loops — rather than simple two-app connections. - Priced by credits consumed rather than by workflow count, with a free tier sufficient for learning. - A direct alternative to other automation platforms, with a materially different billing model that changes which one is cheaper for your specific work. Before you start: - The free plan gives a small monthly credit allowance, caps how many scenarios can be active, and enforces a long minimum gap between runs. Enough to learn on; not enough to depend on. - Paid plans start low and give a much larger credit allowance, which also removes the scenario cap and shortens the minimum gap between runs. - Every module execution costs one credit, not every workflow run. A ten-step scenario running a hundred times spends a thousand credits. - Since 2025-08-27 the meter has been called credits rather than operations, and AI features and code execution consume more per use than ordinary modules. - Sketch your workflow on paper first. Counting the modules before you build tells you the monthly cost before you commit. Security: - Automation platforms carry a different risk from assistants: you are handing over standing access to your accounts, not a single conversation. Connecting a scenario to your email, files, or customer records grants a persistent permission that keeps working while you are asleep, and keeps working after you have forgotten the scenario exists. Security: Sensible practice: - Connect the narrowest account you can. A dedicated account with access to one folder beats your primary login. - Review your connected accounts periodically and revoke anything belonging to a scenario you no longer run. - Test with harmless sample data before pointing a scenario at anything real. Automations fail in bulk, and quickly. - Be deliberate about scenarios that delete, send, or overwrite. Those are the ones where a mistake is not recoverable. Security: - Data passing through a scenario is processed on the platform's servers, so a workflow moving personal or customer information puts that data in a third party's hands. That is often perfectly acceptable — it is the same trade as any cloud tool — but it should be a decision you made rather than one you discovered. Best at: - Multi-step workflows with conditional branching and filtering - Reshaping data between apps that do not agree on format - Scheduled jobs that gather from several sources - Work where seeing the flow laid out visually helps you reason about it Okay at: - Simple two-app connections, which work but where simpler tools have less setup - Real-time responsiveness, which is constrained on lower tiers - Error handling, which is capable but needs deliberate configuration Avoid: - Anything where a silent failure would go unnoticed and matter - High-volume record processing on a small plan, where credits vanish fast - Automating a process you have not yet done manually enough times to understand Starter: Count the modules before you build Estimates monthly credit use as records times modules times runs. This one calculation prevents the most common and most expensive misunderstanding on the platform. If the number surprises you, look for modules to remove rather than a bigger plan to buy. Starter: Automate a form-to-spreadsheet flow first Builds a two-module scenario: a form-response trigger and a spreadsheet module that adds a row for each submission. Few modules, obvious success criterion, and it exercises the trigger-and-action pattern everything else builds on. Add a filter so only some submissions pass through. Filters are where Make starts to be worth its complexity. Starter: Build a scheduled digest Gathers from a source on a schedule and sends one summary rather than many alerts. Teaches scheduling and aggregation together, and produces something you will actually keep using. Lengthen the interval before shortening it — frequent runs are the fastest way to burn credits. Starter: Run a scenario once manually and read every step Runs the scenario once by hand rather than on its schedule, then clicks through each module in the execution view. The execution view shows exactly what data each module received and emitted. It is the debugging skill the platform rewards most. Do this while it works, not only when it breaks. Knowing what correct looks like makes broken obvious. Pitfalls: - Credits count modules, not runs. A ten-step scenario running a hundred times costs a thousand credits, not a hundred. Budget by multiplying steps by runs, not by counting workflows. - Comparisons with other platforms are not like for like. Rival tools meter differently, so one of their units can equal several credits here. The cheaper platform depends on your workflow shape — benchmark your own before switching. - AI modules cost more per use. Anything invoking a model or running code consumes credits at a higher rate than ordinary modules. A workflow that looked affordable can double when you add one. - Connected accounts are standing permissions. A scenario keeps its access until you revoke it, including scenarios you stopped using months ago. Audit connections periodically. - Broken automations fail quietly. Something that silently stopped running two weeks ago is worse than something that never worked. Build in a notification for failure, not just success. Where to next: - Other automation platforms: https://www.launchpadhq.io/discover?category=workflow-automation - AI inside business software: https://www.launchpadhq.io/discover?category=ai-plugins-business-software ### Getting Started with Manus Canonical: https://www.launchpadhq.io/platform/manus/guide Data: https://www.launchpadhq.io/data/v1/guides/manus.json Last reviewed: 2026-10-23 Tool: Manus (https://www.launchpadhq.io/platform/manus): starting price Free, from $20/mo; free tier yes; data as of 2026-10-13 Give Manus a goal and it works through it on its own computer in the cloud — researching, browsing, writing files, even building a website — while you do something else. You describe a task in plain language. Manus makes a plan and carries it out in its own sandbox — a private computer in the cloud with a web browser and files — showing each step. It keeps working if you close the tab, then hands back the result: a report, spreadsheet, slides or a working web page. What it is: - Manus is an AI agent — software that takes actions to finish a job rather than only answering questions. Its docs describe it as a virtual colleague with its own computer: it can search the web, open pages, write and run code, install software and create files. You watch it work, and you can step in and take over its browser at any point; Manus will ask you to when it gets stuck, for example at a CAPTCHA. - Because the work happens in the cloud, tasks keep running after you close your browser. Paid plans add Wide Research, which splits a big research job into many parallel sub-tasks, and scheduled tasks can repeat a job on a timetable — a weekday-morning briefing, say. Connectors link Manus to apps such as Gmail, Google Calendar, Notion and Slack, and an optional browser extension lets it work inside your own Chrome or Edge. - Manus runs on the web, Mac, Windows, iPhone and Android. On ownership: Manus's own blog says Meta acquired it in December 2025, and that it formally resumed independent operations in September 2026 after separating from Meta to meet regulatory requirements in certain jurisdictions — news reports said Chinese regulators had ordered the deal unwound. Its terms say the service is run by Butterfly Effect Pte. Ltd. under Singapore law, and its trust centre says data is stored in the US and Singapore. Before you start: - Everything runs on credits. Manus's credits page says a task's cost depends on how complex and long it is — the AI's reading and writing, the cloud computer's time, and any outside services used. Its examples show a long web-app build costing several times what a short chart does. Its help centre says there's currently no way to see the cost before a task starts. - The free plan is daily. Manus's help centre lists a Free plan with a daily credit allowance (reset at midnight UTC, no rollover), one task at a time and two scheduled tasks. It also caps how much of that daily allowance a free user can spend in a month. Those daily credits can only be used with the lighter Manus 2.0 Lite model. - Paid plans are monthly credit allowances. Manus's pricing help article lists several paid tiers, each with a bigger monthly credit allowance than the last (the second comes with a 7-day free trial), and a discount for annual billing. Monthly credits reset each billing date, and Manus says leftover monthly credits are wiped if you cancel. - Refunds are limited. Manus's terms say payments are non-refundable except where the law requires. Its help centre says credits are refunded when a task fails because of a technical problem on Manus's side — not if you simply don't like the result. Security: - An agent that acts for you deserves more care than a chatbot. Manus's terms say actions the agent takes can be attributed to you, and revoking access doesn't undo something it already did. Start with tasks where a mistake is harmless — research, drafts, files — before letting it send, buy or post anything. Security: Before you hand over access:: - Logins. When Manus needs a site you're signed in to, you log in yourself in its cloud browser, and it asks whether to save that session. You can review or switch off saved logins in Settings → Cloud Browser. - Your own browser. The Browser Operator extension works in your real Chrome or Edge with your existing logins. Manus says each task needs your authorisation and passwords stay in your browser; switch off the "My Browser" connector when you're done. - Skip confirmations. Scheduled tasks have an option to send, publish or post without asking you first. Leave it off until you trust a workflow. - Share links are public. Manus's help centre says tasks are private until you share them, and sharing creates a public link; its privacy policy warns shared content can be copied or picked up by search engines. Security: - Manus's trust centre says its outside AI providers are contractually barred from training on your data, but its privacy policy says Manus may use de-identified data to improve its own models. A training opt-out is advertised as a Team-plan feature; we couldn't find one for individual plans on review day. Best at: - Multi-step research that ends in a finished file — a comparison table, a report, a slide deck. - Building a simple website or web app from a description, then publishing it. - Recurring jobs on a schedule, like a daily summary of news on a topic you follow. Okay at: - Predictable costs. You find out what a task cost after it runs, so the first few are guesswork. - Accuracy. Manus's terms ask you to review outputs, and it shows a "verify before you act" notice on health, money and legal topics. Avoid: - Giving it logins to accounts where an unwanted action would be costly. - Sharing a task link that contains anything private. Starter: Compare before you buy Comparing options before a purchase is one of Manus's own beginner examples. It shows the agent researching across many pages and delivering a finished table — and the links let you check its work. Swap in anything you're shopping for, and add your must-haves. Prompt: Make a table with price, battery life, weight and the main complaint in reviews, and give me a one-paragraph recommendation with links to your sources. Starter: Plan something end to end Manus's beginner guide suggests phrases like "take ownership" for whole jobs. A complete plan with a downloadable file shows what an agent does that a chatbot doesn't. Prompt: Give me a day-by-day plan with neighbourhoods, two restaurant options per evening and estimated costs, as a document I can download. Starter: Build a one-page website Manus's credits page uses a small webpage as a typical task. You'll get a working page you can open and refine — check whether its link is temporary before sharing it. Prompt: Build a simple one-page website for a neighbourhood book club: a welcome section, the next meeting date and book, and a contact email. Keep it clean and mobile-friendly. Starter: Set up a weekly briefing Scheduled tasks are available on every plan. A weekly summary is low-risk, useful, and shows you how many credits a repeating job really uses. Prompt: Every Monday at 8am, search for the past week's news about electric bikes and give me a five-bullet summary with links. Pitfalls: - One giant conversation. Manus's credit tips say long single threads cost more; start a new task for a new job and put everything it needs in the first message. - Vague first prompts. An agent fills gaps by doing more work, which spends more credits. - Temporary links. Manus's help centre says temporary deployment links expire; use permanent deploy for anything you want to keep. - Tasks stopped by high demand. If a task ends with a high-load error, Manus says you can stop and take the credit refund, or wait and continue. Where to next: - More workflow & automation tools: https://www.launchpadhq.io/discover?category=workflow-automation - Research & academic tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Chat assistants for quick questions: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Mem Canonical: https://www.launchpadhq.io/platform/mem/guide Data: https://www.launchpadhq.io/data/v1/guides/mem.json Last reviewed: 2026-10-14 Tool: Mem (https://www.launchpadhq.io/platform/mem): starting price Free, from $9/mo; free tier yes; data as of 2026-10-14 A notes app that files everything for you, then answers questions about what you saved. You type or speak a note and it lands in one searchable pile — there are no folders to choose first. When you want something back, you open a panel beside your notes and ask in plain English; it reads across everything you have saved and answers there. You keep going from that answer. What it is: - Mem is a notes app built around the idea that filing is the part people give up on. Anything you throw at it — a typed note, a spoken thought, a forwarded email, a clipped web page — gets read, structured and grouped for you. - On top of that sits Mem Agent, an agent (AI that takes actions on its own rather than only answering) that watches for tasks and unfinished threads in what you have written and checks in about them. - The reason to pick it over Notion AI or Obsidian is narrow and worth being honest about: those two are better places to build something and expect you to decide where each thing lives, while Mem is the one that will take a half-formed thought at 11pm and put it somewhere sensible without asking you where. Before you start: - Signing up currently means signing in with Google — Mem's help centre says a Google account is the only supported login, though its privacy policy also mentions Apple, so you may find more options on the sign-in screen than the help pages describe. If you do not want Mem attached to a Google identity, this is your stopping point, and Obsidian is the obvious alternative in the same category. - The free plan is real but small, and it is metered by activity rather than by time: a small monthly number of messages to Mem and new notes, task tracking across those notes, and a few hours of Voice Mode. Free is enough to find out whether the recall actually works on your own material, which is the only question worth answering first. You would move up when the counters bite — to Mem Plus, which roughly doubles the allowances, or the pricier Mem Pro, which is where connecting Gmail, Slack or Todoist and setting up scheduled routines first become available at all. - It runs on Mac, Windows, iOS and the web. There is no Android app, and Mem's help centre says it has no timeline to share for one, so an Android phone means the web app. - English is the only officially supported language across Voice Mode, Deep Search and Chat. Mem says other languages may work but are not supported, so treat non-English results as something to test before you rely on them. Getting set up safely (Mem's own download page: https://get.mem.ai/download): - Get the Mac and Windows apps from Mem's own download page and the browser clipper from the Chrome Web Store. Search results for a popular AI notes app are exactly the kind of thing that attracts lookalike installers, so the source matters more than usual. - You do not have to install anything to start. The web app covers the note-taking and the chat; the desktop app is what adds meeting capture and the hold-a-key voice capture, and the Chrome extension is what adds web clipping. - When you first turn on Voice Mode in the desktop app it will ask for system-audio permission. That is the setting that lets it record the other side of a call, so grant it only if capturing meetings is why you installed it. Security: - Mem is not a tool you decide what to paste into — the whole product assumes you will put everything in, and it is only useful in proportion to how much you do. That inverts the usual advice. Security: - The vendor is unusually direct about the trade. Mem's security page (revised 2026-01-15) says content is encrypted in transit and at rest, and that it is unencrypted for AI processing by third-party vendors so the product can do the capture, organising and recall it advertises. Its pricing FAQ says the same thing more plainly: your data is not end-to-end encrypted, "because Mem needs to process your knowledge to provide intelligence on top of it." Mem states it is SOC 2 Type II compliant — SOC 2 being an audit of how a company handles customer data — and hosts on Google Cloud. Security: - On the question that matters most for a notes app: Mem's privacy policy, revised 2026-08-13 and in force as of 2026-09-13, states that it does not use your personal information, your notes, or information collected via third parties such as Google Workspace APIs "to develop, improve, or train any generalized AI and/or ML models," and that queries sent through its integrations with other AI platforms are not used to train models either. Read the word generalized rather than skimming it: the commitment as written is about general-purpose models, and it is not the same sentence as "nothing derived from your notes is ever used for anything." Security: Two more things a reader should know before their whole working life is in here: - Deleted is a process, not a moment. Trash empties itself after 30 days. The Terms of Service (revised 2026-09-11) say deleted content is removed from Mem's systems within 90 days, and that activity logs which may include partial copies of deleted content are kept for up to 400 days. - A Connection is a standing read on another account. Once linked, the assistant is answering from your inbox as well as your notes, which quietly changes what a shared note or a stray screen-share can reveal. The privacy policy says you can revoke access by disconnecting the service at any time; link one only when you have a specific reason. Best at: - Getting a thought out of your head before it evaporates, without naming or filing it. - Answering questions about your own past that you could not construct a keyword search for. - Turning a recorded meeting into a usable note, with the original audio and the transcript attached to it. Okay at: - Being the place you build things. The editor, templates and collections are competent, but there is no database or table-driven workspace of the kind Notion is bought for. - Team knowledge. Sharing works per note and per shared collection, but the current pricing table lists personal plans only — there is no separate team tier in it. - Large clean-up jobs. Mem's docs say Chat works through bulk operations in batches of around 50 notes and pauses to check in, so "organise everything" is a session, not a click. Avoid: - Moving a big existing library in and expecting it to survive intact. Imports accept Markdown and plain text only, around 1,000 files at a time, and Mem's own import docs say folder structure does not come across and attachments are not imported with the note. - Making Mem the only copy of anything. Its export gives you your current notes as plain-text files, and the vendor's version-history page states plainly that exports do not include a note's version history. Starter: Record the meeting, then ask for the shape you actually need Voice Mode records your microphone and the system audio locally from inside the note without joining the call as a participant, and when it stops you can hit Refine and re-run the same transcript into a different structure as many times as you like. Notion AI can summarise a transcript you give it; it has no recorder of its own to re-run from, which is what makes the second and third attempt cheap here. Prompt: Rewrite this as three sections: decisions we made, action items with an owner for each, and anything we left unresolved. Starter: Ask the question you could never have searched for Chat searches your whole workspace by default rather than matching words, and Mem's own docs recommend describing what you want in plain language instead of attaching a big collection, because pointing it at a 200-note collection fills up what the chat can hold and the answers get worse. That is the opposite of the Obsidian workflow, where finding something still starts with you knowing roughly which note it is in. Prompt: What did I decide about pricing, and what changed my mind? Starter: Tell the web clipper what you want out of the page The Chrome extension's popup has its own input field, and typing # targets a collection while @ appends the clip to an existing note — so the summarising and the filing both happen at clip time, in the browser, before you ever open the app. Obsidian's clipper saves the page into a vault folder you pick yourself and leaves the reading to you. Prompt: Pull out just the pricing claims and the launch date, then file this under #Competitors. Starter: Speak it without opening anything Hold Right Option on Mac or Right Ctrl on Windows and the desktop app sends what you say straight to the agent from on top of whatever you were doing, with the app hidden — the vendor describes this as a request the agent interprets rather than dictation into the current window. There is no equivalent global push-to-talk key in Notion AI. Note that Mem says which actions the agent can then take depend on your plan, so on the free plan expect the thought to be captured reliably and treat the follow-up reminder as the thing to verify. Double-tap the same key instead of holding it to go hands-free, then press once more to send. You can change or disable the shortcut in Settings → Mem Agent. Prompt: Remind me Thursday to send Priya the offsite budget, and keep it with the offsite note. Pitfalls: - The monthly counter counts more than you would guess. A message in the chat, a push-to-talk request, a message through a channel like WhatsApp and an email to [email address on the web page] all count as Messages to Mem, and Mem's pricing FAQ notes that a forwarded email that creates a note also spends one of your new-note allowances. On the free plan an enthusiastic first afternoon can eat the month. - Long chats get worse, not better. Mem's own docs say answers degrade as a conversation fills up, and advise starting a fresh chat rather than trying to rescue one that has gone off the rails. Beginners instinctively do the opposite and keep pushing. - Trash is on a timer. Version history cannot recover a permanently deleted note, so if you bulk-delete duplicates after an import, check the results well inside that 30-day window. - What your plan includes is not all in one table. The plan comparison lists eight metered allowances. Separately, individual feature pages and the help centre mark Briefings, Heads Up Live, choosing your own model and even Dark Mode as Pro features, and none of those four appear as rows in the comparison table. If one of them is the reason you are upgrading, confirm it before you pay rather than assuming the table is complete. Where to next: - Meetings & Notes: https://www.launchpadhq.io/discover?category=meetings-notes - Workflow & Automation: https://www.launchpadhq.io/discover?category=workflow-automation ### Getting Started with Meshy Canonical: https://www.launchpadhq.io/platform/meshy/guide Data: https://www.launchpadhq.io/data/v1/guides/meshy.json Last reviewed: 2026-09-12 Tool: Meshy (https://www.launchpadhq.io/platform/meshy): starting price Free, from $20/mo; free tier yes; data as of not dated Describe an object in a sentence and get a 3D model of it about a minute later. You type a description of an object into a box, press generate, and about a minute later a 3D shape appears that you can spin around with your mouse. If it's wrong, you change the wording and generate again. When one looks right, you add colour and surface detail in a second pass. What it is: - Meshy makes 3D models — the kind of digital object you'd drop into a game, a video, or a 3D printer — from a written description or a photo. - It's aimed at people who want a usable object without first learning 3D modelling software. It runs in a browser with nothing to install, and hands you the finished file in the formats other 3D software opens. - What comes back is good enough for a prototype, a background prop or a print; anything a customer looks at closely, a person still rebuilds by hand. Before you start: - A free account gets a small monthly allowance of credits (the platform's unit of spend — each thing you make costs some), resetting on the first rather than stacking up, so an unspent month is simply gone. Meshy's own credit table prices the shape lower on its lightest generator than on the heavier two, with colour and surface detail costing extra — so a free month covers only a handful of finished objects. Daily sign-in bonuses add a few more credits, and those ones don't expire. - Before you spend anything, know that the generator you pick decides whether you can keep the file. Meshy currently offers three: Meshy 7, Meshy 6 and Meshy 6 Lite. Its free-plan help page says free accounts get a small number of downloads a month, and only for Meshy 6 Lite models; its launch post for Meshy 7 says plainly that only Pro tier and above can download Meshy 7 models; its plan comparison table is blunter still and shows free downloads as simply unavailable. Those three vendor pages don't agree, so the honest version is: generating is free, keeping the file may not be, and you should check what your own account offers once you are signed in, before you burn credits on something you can't export. - Per Meshy's own model-comparison page the newest generator, Meshy 7, does images only — typed descriptions run on Meshy 6 or Meshy 6 Lite, so the newest model isn't the one your typed words reach at all. - You'd pay for the files and for the licence: paid plans come with a much larger credit allowance and move you from a share-alike licence to private ownership (see Privacy & security). Start at meshy.ai — signing up free asks for no card. Security: - Meshy's risk isn't mainly about what you type — it's about what you end up owning. Its Terms of Use, currently effective 7 March 2026, say that on the free plan Meshy itself owns the output and grants you a licence to it under Creative Commons Attribution 4.0. You may use a free model commercially, but you have to credit Meshy — its help centre suggests the line "Model created with Meshy – CC BY 4.0 License" in your project description — and a Creative Commons licence is one anyone else who gets hold of that model may exercise too. Paid plans replace this with private ownership, and Meshy's help centre says models made while you were paying stay private indefinitely even if you later cancel. Security: Two things in the terms worth knowing before you upload anything: - Your inputs and outputs may be used for training. Section 2.9 of the same terms lets Meshy use inputs and outputs from all non-Enterprise customers — free and paid alike — to train and improve its services. On the vendor's own account only an Enterprise agreement switches that off. - Publishing to the community gallery goes much further than sharing. The terms put anything you publish publicly under CC0, a public-domain dedication — which hands the model to everyone with no credit owed to you at all, and isn't something you can take back. Security: - One thing we couldn't establish: whether a free-plan model you haven't published is visible to anyone else. No Meshy page we could find says either way, and the terms only state that paid customers have the option to keep content private — which implies free accounts don't. Until that's clearer, treat a reference photo you upload the way you'd treat something you posted publicly. Best at: - Props and objects — a chair, a lantern, a barrel, a cartoon mushroom. Single, self-contained things are what it's built around. - Handing you a file other software opens — its text-to-3D page currently lists eight export formats, including FBX, OBJ, GLB, USDZ, STL and BLEND. Okay at: - Anything with thin parts. Handles, wires, railings and spokes tend to come back fused together or thickened into blobs. - Characters. You'll get a figure; you won't get one a game could animate without a person reworking it. Avoid: - Scenes rather than objects. Ask for "a medieval market square" and you get one lumpy mass, not a square you can move around in. Describe one stall, then another. - Anything that has to match real-world measurements — a replacement bracket, a part that must fit an existing hole. Nothing in the generator is dimensioned, and it will cheerfully produce a confident wrong size. - Uploading a photo of a copyrighted character or someone else's product and assuming the 3D version is yours. Meshy's own help centre says it does not guarantee you rights to models generated from copyrighted references you upload — that exposure stays with you. Starter: One object, described by its shape What goes in the box is your prompt — the message you type. Because Meshy builds the shape before any colour, words about silhouette, material and pose do real work in a first prompt and words about colour mostly don't. The box takes up to 800 characters, room for all three and not much more. Swap the object for anything you like, but keep "one single object" — that phrase is what stops you getting a scene. Prompt: A weathered wooden treasure chest with iron bands and a curved lid, closed, sitting flat on the ground, one single object, nothing else in frame Starter: A prop with a style, not just a thing Asking for exaggerated chunky proportions up front costs nothing; discovering that the thin version fused into a blob costs you another generation out of the month's credits. Run the same object twice, once as "stylised cartoon" and once as "realistic", and see which holds together better. Prompt: A stylised low-detail cartoon lantern with a rounded glass body and a thick chunky handle, exaggerated proportions, one single object Starter: Run one description on two generators and compare Generate the same description twice, once on Meshy 6 Lite and once on Meshy 6, and compare the two before you commit to either. The two differ in price and, per the download rules in "Before you start", possibly in whether the free plan lets you keep the result at all — a cheaper thing to learn on a test object than on the one you actually wanted. Do this on your first day, while most of the month's credits are still there. Starter: Point it at a photo instead of a sentence Upload a straight-on photo of a real object you own — a mug, a shoe, a plant pot — and let it build from the picture rather than your words. This is the route Meshy's newest generator was built for: its August 2026 launch post is entirely about how closely the 3D result lines up with the input image. Since typed descriptions don't reach that generator, a photo is the only way to see its current best work. One object, plain background, shot straight on and evenly lit. Starter: Something you could actually print Meshy exports STL and 3MF and sells a separate printability-repair step that costs extra credits — which tells you plainly that a freshly generated model often isn't printable as it stands. Asking for thick, flat-bottomed shapes from the start avoids paying to fix thin walls afterwards. "Flat underside" is the load-bearing phrase — a printer needs something to stand the model on. Prompt: A simple solid desk organiser shaped like a curled sleeping cat, thick rounded body, no thin parts, flat underside, one single object Pitfalls: - Judging it on the grey preview. The shape arrives with no colour or surface detail and looks far worse than it is — adding those is a second step you have to ask for and pay credits for separately. - Assuming the file will drop into a game engine and just work. It'll open, but what's underneath is generated geometry, not something built to be animated or made efficient. - Spending the whole month's credits on one ambitious idea. Your first two or three attempts are calibration — on a free account that's most of what the month buys you. Where to next: - Gaming and creative AI: https://www.launchpadhq.io/discover?category=gaming-creative-ai - Image generation and editing: https://www.launchpadhq.io/discover?category=image-generation-editing - Video creation and editing: https://www.launchpadhq.io/discover?category=video-creation-editing ### Getting Started with Meta AI Canonical: https://www.launchpadhq.io/platform/meta-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/meta-ai.json Last reviewed: 2026-10-15 Tool: Meta AI (https://www.launchpadhq.io/platform/meta-ai): starting price Free, from See pricing; free tier yes; data as of 2026-10-17 The free AI assistant already sitting inside WhatsApp, Instagram, Messenger and Facebook, plus its own app and website. You open a chat with Meta AI, either in its own app, at meta.ai, or inside an app you already use, and type a message (this is your prompt — the message you type). It answers in plain language, and you reply to steer it: shorter, different, try again. What it is: - Meta AI is the AI assistant made by Meta, the company behind Facebook, Instagram, WhatsApp and Messenger. It answers questions, writes and rewrites text, makes and edits images, and can look at a photo you show it. - What sets it apart is where it lives. Besides its own app (iPhone, Android, and a Mac version in beta) and the website at meta.ai, it is built into the search bars and chats of Meta's own apps, and it is the assistant inside Meta's AI glasses. If you already use WhatsApp or Instagram, you may have Meta AI on your phone without ever installing anything. - The model (the AI "brain" that does the actual thinking) behind it is called Muse Spark, built by Meta's own lab. - Its other difference is that it is wired into Meta's social world: Meta describes it searching Facebook Marketplace listings near you when you shop, and surfacing public posts from locals about a place. That is useful, and it is also why the privacy section below matters more here than on most chatbots. Before you start: - It is free. Meta's own page says Meta AI is free for everyday use, that it is testing usage limits, and that a paid subscription gives you more. - You can start in two places, and which one you pick decides which account it's tied to. Inside WhatsApp, Instagram, Messenger or Facebook, Meta AI uses the account you are already signed in to. For the standalone app or meta.ai, you sign in when you open it; Meta's public pages we could read don't spell out whether that requires a Facebook or Instagram account, so check the sign-in screen for your options. - If your Facebook and Instagram accounts are linked in Meta's account settings, Meta AI can draw on both, including your profile details and what you engage with, to tailor its answers. Meta says these personalized answers are currently available only in the US and Canada. - Features roll out unevenly. Meta's announcements repeatedly say new abilities start in the US and reach "more countries" later, and the app listing warns some features are limited to certain countries and languages, so a feature you read about may not be on your phone yet. - If you own Meta's AI glasses, the Meta AI app is also their required companion app, so you may already have it installed for that reason. Security: - Meta AI is built by an advertising company, and it says so. Since 16 December 2025, in what Meta calls "most regions", Meta uses your conversations with its AI features to personalize the posts, reels and ads you see across its apps: chat about hiking and you may be shown hiking groups and hiking-boot ads (Meta newsroom, announced 1 October 2025). Meta's announcement describes no switch to turn this off, only your usual ads and feed preferences. Meta also lists 'people's interactions with AI at Meta features' among the data it uses to train its AI (Meta Transparency Center, seen 2026-10-15). Security: What Meta says limits this:: - Conversations about religion, sexual orientation, politics, health, racial or ethnic origin, philosophical beliefs or trade-union membership are not used to show you ads, per the same October 2025 announcement. They are not excluded from everything else. - Your accounts only share this with each other if they are linked in Meta's account settings. Meta's example: a WhatsApp account you have not linked won't feed your Facebook or Instagram experience. - Incognito Chat, announced 13 May 2026 for WhatsApp and the Meta AI app and rolling out over the coming months, is a private mode Meta says it cannot see, where conversations are not saved and messages disappear by default. If you have it, use it for anything personal. Security: Never paste in:: - Passwords, card or bank numbers, ID numbers - Health details or anything you would not want shaping the ads you see - Other people's private information, including screenshots of group chats Security: - And watch the Share button. The app has a public Discover feed of other people's AI chats. Meta says nothing goes there unless you choose to post it, so read what you are sharing before you tap it. Best at: - Quick questions and writing help in the middle of a chat you are already having in WhatsApp or Messenger, without switching apps. - Local and shopping questions, where it can pull in Facebook Marketplace listings and public posts from people near a place. - Making and editing images from a description. - Looking at something through your camera or a photo and telling you what it is. Okay at: - Long, careful reasoning. There is a slower Thinking mode for harder questions; use it, and still check the answer. - Anything that depends on a feature you read about in the news; it may not have reached you yet. Avoid: - Anything you would not want feeding the ads in your Facebook and Instagram feed; the privacy section explains why. - Treating a group-chat answer as private. Mentioning Meta AI inside a WhatsApp group puts the question and its answer in front of everyone in that chat. - Assuming 'free' means unlimited. The usage limits Meta is testing are not published. Starter: The local scout Meta says Meta AI can search public posts from locals about a place, alongside the wider web. ChatGPT can search the web too, but it has no access to Facebook and Instagram posts, so the 'what are people saying' half of this answer is Meta's own. Swap the kids for a budget, a dog, or 'rainy day'. Prompt: I'm visiting [neighbourhood or town] this Saturday with two kids under 10. What are three things to do there, what are local people saying about them right now, and what should I skip? Starter: The Marketplace shopper Meta describes a shopping mode that searches Facebook Marketplace listings near you, with a map and filters for price and distance. No other assistant on this site searches Marketplace. If you don't see listings in the answer, that feature may not have reached your country or account yet. Prompt: Help me find a used [bike / desk / stroller] near me for under $[amount]. Show me a few options and tell me what to check before I buy one secondhand. Starter: What is this? Point the camera or attach a photo first, then send this. Meta's own example is comparing snack labels from a photo. The 'how sure are you' line is the important part: it makes the assistant tell you when it's guessing. Works for a plant, a strange error message on a screen, or a label in another language. Prompt: What is this, and is it safe to [eat / touch / leave in the garden]? Tell me how sure you are. Starter: The group-chat helper Meta says Meta AI is coming into group chats in its apps; where it is switched on in WhatsApp, you call it in by mentioning it, and the whole group sees the answer. ChatGPT can't join a WhatsApp group chat that way. Use it to settle a trivia argument, or to turn a long thread of suggestions into one list. Prompt: @Meta AI suggest three dinner places within 15 minutes of [area] that can take a group of 8 on Friday, with a rough price for each. Pitfalls: - Taking the first answer. Reply with 'shorter', 'friendlier', or 'you missed X'. The second and third tries are usually where it gets good. - Forgetting which account you're in. Because each place uses its own sign-in, a chat you had in one may not be in another. - Thinking a chat is private because you're in a messaging app. WhatsApp is known for keeping personal chats private, but talking to Meta AI is talking to Meta. - Confident answers about the local world. Public posts from locals can be out of date or wrong, and it can mix them into an answer that sounds certain. Check opening hours and prices yourself. Where to next: - More AI chatbots: https://www.launchpadhq.io/discover?category=text-conversational-ai - Image generators: https://www.launchpadhq.io/discover?category=image-generation-editing - Research tools: https://www.launchpadhq.io/discover?category=research-academic-tools ### Getting Started with Microsoft Copilot Canonical: https://www.launchpadhq.io/platform/microsoft-copilot/guide Data: https://www.launchpadhq.io/data/v1/guides/microsoft-copilot.json Last reviewed: 2026-09-07 Tool: Microsoft Copilot (https://www.launchpadhq.io/platform/microsoft-copilot): starting price $9.99/month; free tier yes; data as of 2026-10-26 Microsoft's free AI assistant, already sitting inside the Windows and Office world you use every day. You open Copilot in a browser, on your phone, or from the taskbar on a Windows PC, then type a question or paste in text and it answers in a chat window. You keep replying in the same thread to sharpen the answer. What it is: - The name is the confusing part. This page is about the free consumer Copilot — at copilot.com, in the Copilot app, and on the Windows taskbar, signed in with a personal Microsoft account. It is not GitHub Copilot (a paid tool that writes code in a programmer's editor) and not Microsoft 365 Copilot (a paid work add-on that reads your organisation's files and emails). Same brand, different products, sold separately. - Why pick it over ChatGPT or Gemini: it is already on a Windows 11 PC and leans into Microsoft's world — the Edge sidebar, Windows settings, Word and Excel if you subscribe. Otherwise it is in the same class as its rivals, so the deciding factor is which ecosystem you live in. Before you start: - Nothing to start. Go to copilot.com, or open the Copilot app on a Windows 11 PC, where Microsoft says it is typically installed by default. You can chat without an account. - Signing in with a free personal Microsoft, Apple or Google account is optional, but Microsoft says it unlocks chat history, longer conversations, more image creation and longer voice sessions. - The free tier is not a trial that expires but a working assistant with ceilings: ordinary chat keeps going, while image generation, voice and the heaviest thinking run down and refill. - Most beginners never pay. You would pay for two things — Copilot inside the Word, Excel and PowerPoint desktop apps, and Copilot Vision, which lets it look at your screen — both currently needing a paid Microsoft 365 Personal, Family or Premium subscription. Businesses pay for a separate per-user add-on; see current plans on the tool page. Security: - Copilot is a cloud service: whatever you type, paste or upload goes to Microsoft's servers. What matters here is less what you send than which account you send it from. Security: Personal account versus work or school account: - On a personal Microsoft account, Microsoft says it uses data from Bing, MSN and Copilot to train its models — the AI "brains" that do the thinking. A privacy setting opts your conversation activity out; find it early. - On a work or school account — the sign-in from your employer — Microsoft states plainly it does not train Copilot on that data, and that what you type and what it answers are not used to train the underlying models. Your IT admins then get controls and visibility they lack over your personal account. - You can see which one you are in: the app puts a Work or Personal label next to your profile. Glance at it before pasting anything sensitive. Security: - The five minutes worth spending on day one go into the privacy settings; the label in the corner is what to check thereafter. Best at: - Rewriting and re-toning text you have already written - Explaining an unfamiliar term or error message in plain language - Questions needing current information — it searches the live web and shows its sources - "How do I do this on my Windows PC?" questions about settings - Getting a rough first draft onto a blank page Okay at: - Summarising a long article or document you paste or upload - Casual image generation, and light spreadsheet and formula help Avoid: - Serious programming work — that is GitHub Copilot, a separate paid product - Confidential, client or regulated material in a personal account, where activity feeds model training unless you opt out - Expecting a subscription to make it cleverer; Microsoft 365 buys reach, not better answers - Anything where being confidently wrong is expensive — medical, legal, financial, tax — or any figure you cannot check Starter: Fix a real email before you send it On a Windows 11 PC Copilot is on the taskbar by default, so it reaches you while you are writing rather than making you open a tab. ChatGPT and Gemini both want that tab. Prompt: Here is an email I am about to send. Rewrite it to be clearer and about a third shorter while keeping my voice, then list what you changed. Starter: Ask your own PC how to do something Microsoft lists settings help among the things Copilot on Windows does that the browser version cannot, so it names the actual path on your machine. Prompt: My laptop screen is hard to read and the text feels too small. Walk me through fixing this in Windows, and tell me where each setting lives. Starter: Get a sourced answer on something current Copilot sits in the Edge sidebar, so a sourced answer arrives beside the page you are already reading. Perplexity and Gemini show sources too; neither is in the sidebar of the browser Windows ships with. Prompt: What are the current rules for this topic? Short answer first, then the detail, with links to the sources so I can check them. Pitfalls: - Signed out is a weaker Copilot — shorter conversations, no history, less image creation. If it feels clipped or forgetful, check you are signed in. - Work habits leak into personal accounts. People paste work documents into a personal session without thinking. The data handling is different, and so is your employer's policy. - Assuming a tip you read applies to your Copilot. Much of the advice online is about one of the other three. Where to next: - Compare the big chat assistants: https://www.launchpadhq.io/discover?category=text-conversational-ai - AI inside business software: https://www.launchpadhq.io/discover?category=ai-plugins-business-software ### Getting Started with Microsoft Designer Canonical: https://www.launchpadhq.io/platform/microsoft-designer/guide Data: https://www.launchpadhq.io/data/v1/guides/microsoft-designer.json Last reviewed: 2026-09-13 Tool: Microsoft Designer (https://www.launchpadhq.io/platform/microsoft-designer): starting price Free; free tier yes; data as of not dated Describe a card, invitation or image in one sentence and get a finished design you can edit. You sign in, type a description of the picture or card you want, and a few finished designs come back within seconds. Pick the one closest to what you meant, then drag, retype and recolour anything on it, or describe a change and let it redo that part. Download the result when it looks right. What it is: - Microsoft Designer is Microsoft's design app for people who are not designers: you describe something in ordinary words and it produces a finished-looking image, card, invitation or social post that you can then edit. - It runs in a browser at designer.microsoft.com and as a phone app, and it signs in with the Microsoft account you may already have from Outlook.com, Xbox or a Windows PC. - Compared with a general chatbot that also makes pictures, Designer hands back an editable layout — text boxes, shapes and photos you can move — rather than one flat image. Before you start: - The account is the real gate, not the money. Designer needs a personal Microsoft account. Microsoft's support page says Designer is not available to enterprise users, so if the only Microsoft login you have is a work or school one, that is not the way in. First step: go to designer.microsoft.com and sign in with a personal account. - It costs nothing to start and most people never pay. Every account gets a set of credits (the platform's unit of spend — each thing you make costs some) that refills at the start of each month, and Microsoft's plan page says each use of an AI feature in Designer or Copilot spends from it. Microsoft no longer publishes how many credits a free account gets each month — its current plan tables describe usage in words like "standard" and "extensive" rather than in numbers — so expect a monthly allowance of unspecified size rather than a figure you can plan against. You would only pay if you run the balance flat regularly, and the paid route is a Microsoft 365 subscription, which raises the allowance rather than unlocking a different Designer. - Microsoft's own pages disagree about that free tier, so be ready for it. Its Designer plan comparison shows a free column with monthly credits, and its Designer FAQ says plainly "Designer is free to use. A subscription may be required for those who want to create more frequently" — while the "Welcome to Microsoft Designer" support article says you will need a Microsoft 365 Personal or Family account. Nobody has reconciled the two; if you are bounced to an upgrade screen, that is why rather than a mistake on your part. - The AI features are age-gated and the threshold differs by country. Microsoft's FAQ says a message that Designer, or one feature such as Upscale, "isn't available for your account" usually means you do not meet the age requirement for your region — not that anything is broken. Whether every AI feature is switched on in your country is not published in one place, so you may find something unavailable to you without a stated reason. Security: - Designer is a consumer Microsoft service, and its own terms are unusually blunt about what happens to what you type. Microsoft does not claim to own your prompts (the messages you type) or the things you make, but you grant it and its suppliers a licence to copy, distribute, display, edit, translate and reformat them "in connection with the operation of their businesses" — and the Designer FAQ says prompt data is collected both for abuse monitoring and to improve the service. Nothing sinister, but it is not a private sketchpad. Security: - Everything Designer's AI makes or edits leaves a mark you cannot see. Microsoft attaches Content Credentials — an invisible tag inside the image file recording that AI was involved — to images created with Designer's AI features, and anyone can read it on the Content Credentials site. That matters if you are posting somewhere that treats AI-made images differently; it travels with the file after you download it. Security: Worth knowing before you upload a photo: - Photos you upload are stored in your own OneDrive, not in a scratch folder that clears itself. Best at: - Turning one line of description into a finished design that already has a layout, fonts and colours chosen for you. - Cleaning up a photo you already have — erasing an object out of it, blurring or removing the background, making a small image bigger without it going blurry. - Personal one-offs where the blank page is the problem: a birthday card, a party invite, a sticker, a phone wallpaper. Okay at: - Getting readable words inside a generated picture. Microsoft's own documentation warns that text in a generated image "may not be in the requested language or may have spelling errors" — a real limit, not bad luck. - Precise layout control — a simplified editor next to a full design program. Avoid: - Anything connected to a business. The Designer consumer terms — the supplement to the Microsoft Services Agreement, effective 31 May 2024 and still the version served today — say your use of Designer is "for personal use only and not for use in the course of trade or commerce." That rules out a shop logo, client work, or anything you sell, however personal it feels. - Treating a Designer image as safely yours to publish widely. The same terms disclaim any warranty that what it generates does not infringe someone else's rights, tell you to consult a lawyer about how you intend to use it, and make you indemnify Microsoft — cover its costs — if a claim arrives. Starter: An invitation from one sentence Designer's invitation feature is documented to take the occasion plus date, time, location and RSVP as separate details and return one to five vertical designs sized for sharing from a phone, so the event facts you type land as text in the design instead of as a description of one. Canva's equivalent starts you in a template you fill in yourself rather than generating the card from event details. Change only the last sentence — the style words — and generate again to see how much of the design they control. Prompt: Create a birthday party invitation for a 7-year-old who loves dinosaurs. Saturday 14 March, 2pm, at Riverside Park. RSVP to Sam by 7 March. Bright and friendly, not scary. Starter: Press Enhance prompt before you press Generate Type a short description of an image, then choose Enhance prompt instead of Generate. Designer rewrites your line into a longer, more detailed one and shows it to you; edit it, then generate. Microsoft documents Enhance prompt as turning your description into "a descriptive high-quality prompt" that you can then edit, so you can see exactly which added words changed the picture. Canva's Magic Media gives you no comparable step that returns the expanded wording for editing. Delete half of what it added and regenerate — the fastest way to learn which words are doing work. Starter: Erase the thing you wish wasn't in the shot Upload a photo, open the erase tool, click the object you want gone — or brush over the area by hand — then accept the result or start the selection over. Having both selection modes on the same photo, plus a Start over that redoes only the selection, is what makes a bad first attempt cheap — and it runs on a free Microsoft account. Adobe Photoshop's generative fill needs a paid Creative Cloud plan. Starter: Ten photos, one described frame Upload up to ten photos, type a theme into the Elements box — "autumn leaves", "beach holiday" — pick a style, and generate. Microsoft documents the collage as coming back "complete with a frame" generated from the words you type, so the border is made for your set rather than chosen from a library. Canva's collage templates give you a fixed frame you drop photos into. Keep the photos and change only the Elements words to get a different set entirely. Starter: A thank-you card with the message already written Microsoft describes the greeting-card feature as producing an animated digital card — an AI-made image plus a written message inside, with the message text still editable afterwards — from a single description, so you get something you can send rather than an image you then have to write on. Edit the message text rather than regenerating — the wording is the part it gets wrong most often. Prompt: A thank-you card for a neighbour who looked after our dog for a week. Warm and a bit funny, with a scruffy terrier on the front. Sign it from Priya and Tom. Pitfalls: - Designer quietly shares your Microsoft storage in both directions. Designs save to the cloud automatically and the photos you upload count against the same allowance as your Outlook mail — currently 5 GB on a free account. Microsoft's own warning is that exceeding it can stop you saving files or sending and receiving email. Deleting a project inside Designer does not remove the photo you uploaded; that has to go from OneDrive. - There is no recycle bin. Microsoft states plainly that a deleted Designer project cannot be recovered. Download anything you would be sorry to lose, the day you make it. - Paying for a Family plan does not buy the family anything. Microsoft's plan documentation says AI benefits are available to the subscription owner only and cannot be shared — so the larger allowance follows one account, not the household. - The counter is not where you would look for it. Microsoft points you to Services & subscriptions in your Microsoft account to see your credits and limits, so if generation stops and Designer does not tell you why, check there before assuming something is broken. - Every interface step described here comes from Microsoft's published documentation rather than a session anyone watched, so a button may have moved or been renamed since it was written. Where to next: - Image generators: https://www.launchpadhq.io/discover?category=image-generation-editing - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Microsoft Power Automate Canonical: https://www.launchpadhq.io/platform/microsoft-power-automate/guide Data: https://www.launchpadhq.io/data/v1/guides/microsoft-power-automate.json Last reviewed: 2026-09-17 Tool: Microsoft Power Automate (https://www.launchpadhq.io/platform/microsoft-power-automate): starting price $15/user/mo; free tier yes; data as of not dated Microsoft's automation tool — already inside the licence your employer bought. You sign in at make.powerautomate.com and open the flow designer. You choose a trigger — the "when", the event that starts an automation — then add the steps that should follow, picking each from a menu and filling in its boxes. Nothing runs until you test it with real data and switch it on. What it is: - Power Automate is Microsoft's tool for making the apps at work hand things to each other without you in the middle. It watches for something — an email arriving, a form being submitted, a file landing in a folder — and then does the next thing itself. - What separates it from the other automation platforms is where it comes from: it is part of Microsoft 365, so for a great many office workers it is already paid for and already sitting in the app launcher next to Outlook and Teams, rather than being something you go out and buy. - It is strongest on Microsoft's own apps and reaches everything else through connectors — Microsoft's ready-made links to other services, published as a catalogue that currently runs to well over a thousand entries. For the programs that have no connector at all, a separate Windows-only companion app drives ordinary desktop software by recording and replaying what you click. Before you start: - What decides whether you can use Power Automate is your email address, not your experience. Microsoft ended support for personal Microsoft accounts here on 26 July 2025: after that date accounts like gmail.com and outlook.com cannot sign into the Power Automate portal or create, edit or manage flows, and Microsoft's notice says the cloud flows attached to those accounts were deleted. A work or school Microsoft account is the way in; if you do not have one, Zapier and Make in this same category both take a personal address on day one. - If your employer gives you Microsoft 365, a first automation usually costs nothing extra. Those licences carry limited Power Automate rights: flows built on Microsoft's "standard" connectors — Outlook, SharePoint, Teams, OneDrive for Business, Excel Online, Forms, Approvals — metered by a daily per-user action limit. The paywall begins at the connector list, not at the number of flows. Salesforce, SQL Server, Jira, DocuSign, Amazon S3 and a plain web request are all classed premium, and reaching any of them needs Power Automate Premium, a per-user licence paid monthly. Microsoft's own note on its listed price is that it is "for marketing purposes only" and varies by country and agreement, so check what your organisation would actually pay. - Two things your IT administrator controls, and neither is visible until you hit it. Administrators set data policies that decide which connectors may be used, and Microsoft's documentation says that when a policy changes a background job scans active flows and suspends the ones that violate it — an automation that worked yesterday can be stopped without you touching it. Separately, if your organisation runs in Europe, Canada or most regions outside the US, UK, Australia and India, the Copilot box that drafts a flow from a sentence only appears once an admin has switched on data movement for generative AI features. - The desktop recorder is a separate download and a separate world. Microsoft's requirements page lists Windows 10, Windows 11 and Windows Server 2016-2025, rules out machines with ARM processors, and offers no macOS or Linux version. On Windows Home editions you can build and run recordings on your own machine but cannot trigger them from the cloud, and work or school accounts need a Dataverse database provisioned in their default environment before desktop flows can be created at all — an administrator's job, not yours, and worth asking about before you install anything. Security: - Most automation tools answer only to you. Power Automate answers to your employer as well: a flow you build on a work account lives inside your organisation's Microsoft account, alongside its Outlook and Teams, where IT administrators can see it and set the rules it runs under. Build accordingly. Security: What that means in practice:: - A flow runs as you. The connections it uses carry your work identity, so anything it sends, files, deletes or shares is attributed to you and can reach whatever you can reach. Point a new flow at a test folder and a test mailbox and watch one real run before you aim it at anything shared. - What you type into the Copilot box is handled under Microsoft's own terms, not a general AI policy. Microsoft's published position, as of September 2026, is that prompts and responses stay within its cloud and are not used to train its AI models unless your organisation opts in to sharing them. - Don't bridge your personal accounts into a work flow. A private Gmail or Dropbox connected from a work automation becomes a permanent two-way route between your own accounts and your employer's systems, and it is your work account that will be audited. Best at: - Joining up the Microsoft apps your workplace already runs — Outlook, SharePoint, Teams, OneDrive, Excel, Forms — without buying anything new. - Approvals: sending a request out and holding the rest of the flow until someone answers, with the request arriving in Teams or Outlook where your colleagues already are. - Driving old Windows software that has no web connection at all, by recording what you click and replaying it. Okay at: - Anything outside Microsoft's own apps. The connectors exist, but the ones people ask for first are classed premium. - Being picked up quickly. The designer is click-through, but the first time a flow needs a condition, a loop or a date in a different format you are writing expressions in a formula language, and the page you land on assumes you already know it. Avoid: - Treating the connector catalogue as the entitlement list. A connector's presence on Microsoft's published list tells you it exists, not that your licence may trigger it — that answer lives in the licensing plan table, and the two documents read very differently. - Building your one critical automation around the mobile app. Microsoft deprecated the Power Automate app for iOS and Android effective 31 August 2026 and removed it from both stores; push notifications sent by flows and the home-screen "run a flow" widget stopped working with it. - Planning to hand a flow to a colleague if you are on the standalone Power Automate Free licence rather than the rights that come with Microsoft 365. Microsoft's licensing documentation states flatly that you can't share flows on the Free licence, so the automation you build for the team stays yours alone. Starter: Never lose an email attachment again When an email arrives with an attachment → then save the file to a OneDrive folder and mark the message as read. Outlook, OneDrive for Business and SharePoint are all standard connectors, so this flow runs on the Microsoft 365 licence your employer already pays for. Put Zapier in that sentence and it stops being true: there the same three steps are metered against a plan you buy on top. The box on the Power Automate home page will draft this from a sentence you type — "when I get an email with an attachment, save it to OneDrive" — on an ordinary Microsoft 365 licence, and if the box isn't there it is usually the generative-AI region switch above. Starter: Turn a form into a row and a nudge When someone submits a Microsoft Form → then add a row to an Excel file and post a note in a Teams channel. Forms, Excel Online (Business) and Teams normally authorise off the same work identity you are already signed in with, so there is usually nothing to connect before you start building. On Zapier each of the three is its own app connection you grant, name and later have to remember to revoke. Starter: Send yourself one summary instead of forty alerts Every Monday at 8am → then gather last week's rows from a SharePoint list and email yourself a single digest. The Power Automate rights inside Microsoft 365 are metered daily rather than as a monthly bundle, so a weekly digest costs you nothing you would ever notice. On Zapier's free plan the same job draws down a monthly task allowance you also want for everything else you build that month. Starter: Automate the program that has no internet in it When you start it yourself → then the desktop recorder opens a Windows program, copies the figures out and types them into a spreadsheet. Zapier reaches an app through its web interface. Power Automate for desktop reaches it through the screen, clicking and typing like a person would, which is why a twenty-year-old Windows program with no web address at all is still automatable here — on the separate Windows-only download whose requirements are above. Record the task once at normal speed and then read the list of steps it produced — that list, not the recording, is the thing you edit. Pitfalls: - Saving is close to switching on. An automated flow generally starts listening as soon as you save it, so the gap between "I was just trying this" and "it fired on a live mailbox" is one click. Run the test, watch one real execution end to end, then leave it. - A trigger that fires on every item really will fire on every item. Point one at a busy shared mailbox and you get a run per message, each one drawing on your daily action allowance. Put the condition in before you switch it on, not after you've read the run history. - The desktop recorder watches the screen, not the program. When the software it drives gets a new layout, the recording stops matching what it expects — usually the morning after an update nobody told you about. Anything built this way needs a person who notices when it goes quiet. - Two Microsoft pages will tell you two different things about the free plan. The pricing page currently advertises a 30-day trial, while the licensing documentation still describes an ongoing Power Automate Free licence and even says personal accounts can use it — which the July 2025 deprecation notice contradicts. Where they disagree, the deprecation notice is the dated one and the sign-in screen you actually see is the final word. Where to next: - Workflow & Automation: https://www.launchpadhq.io/discover?category=workflow-automation - AI Plugins for Business Software: https://www.launchpadhq.io/discover?category=ai-plugins-business-software ### Getting Started with Midjourney Canonical: https://www.launchpadhq.io/platform/midjourney/guide Data: https://www.launchpadhq.io/data/v1/guides/midjourney.json Last reviewed: 2026-06-25 Tool: Midjourney (https://www.launchpadhq.io/platform/midjourney): starting price $10/mo; free tier no; data as of not dated Turn a sentence into striking AI art — no design skills required. You type a description of an image — this is your prompt — and Midjourney generates four versions of it. You pick the one you like best, then refine it: upscale it for more detail, make variations, or adjust the wording and try again. It all happens on Midjourney's website, and you don't need any design skill — just words. What it is: - Midjourney is widely regarded as the best AI image generator for sheer visual quality — it turns a plain sentence into striking, often genuinely beautiful art. It's the tool most people mean when they talk about "AI art." - A beginner reaches for it when the goal is how the image looks — mood, style, atmosphere — more than precise control or a rock-bottom price. It used to live entirely inside Discord, which scared a lot of people off; there's now a normal website, so it's far friendlier than its old reputation. Before you start: - Paid only — there's no free tier. This is the big one: unlike most tools here, you pay before you make a single image. The cheapest way in is the Basic plan (cheaper if billed yearly). First step: go to midjourney.com, sign in with a Google or Discord account, pick Basic, and you're in. - Will you need to pay? Yes — so here's how the plans work. Basic gives you a limited monthly allowance of fast generations, and when it runs out your fast generations stop. Most people who stick with it move to Standard, which adds unlimited "Relax mode" — slower to generate, but you stop counting. One quirk: you're buying GPU time (the computing power to make images), not a fixed number of images, so heavier jobs use up your allowance faster. - If free is a dealbreaker, you're not stuck. Midjourney is the quality benchmark, but it isn't the only option — the catalog has solid free image generators too, a gentler way to test whether AI image-making is for you before you pay for anything. Browse the Image Generation category (linked below) to compare. Security: - Midjourney's privacy is unusual, and it's the main reason to use it deliberately. Two things are true by default: what you make is public, and the prompts and images you create can be used to train Midjourney's models. Privacy ("Stealth Mode") is a paid feature — Pro plan and up — and even then it doesn't make images you already generated private. Security: So keep these in mind:: - Treat a prompt like a public post, not a private sketchbook. Don't put anything sensitive or identifying into a prompt or a reference image on the lower plans — it's visible to others and it's retained. - You own what you create and can use it commercially on any paid plan, even after you cancel — but Midjourney won't defend you in a dispute. Generating a real person's face, a copyrighted character, or a living artist's signature style is a legal risk that's on you. Best at: - Aesthetic quality — mood, lighting, composition, atmosphere. Give it an evocative scene and it produces something beautiful with little effort. This is its whole reason to exist. - Stylistic range — name a medium or era (watercolor, 1970s sci-fi poster, charcoal sketch) and it nails the look better than almost anything else. - Concept art and inspiration — quickly visualizing ideas and directions you can't quite put into words yet. Okay at: - Precise text inside an image — improved, but tools like Ideogram and ChatGPT's image model are more reliable when words must be spelled correctly. - The same character across many images — its reference features help, but the character still drifts; for a consistent face across a series, other tools currently do better. - Photorealism — very good now, though a couple of rivals have edged ahead specifically on looking like a real photo. Avoid: - Exact likenesses of real people, copyrighted characters, or a living artist's signature style — legally risky, against the rules, and Midjourney won't defend you in a dispute. Make "a wizard," not "a famous actor as a wizard." - Anything you need spelled out in precise, reliable text (logos with exact wording, detailed signage) — pick a text-focused generator instead. - If you specifically want a free tool — Midjourney is paid-only. Start with a free image generator from the catalog and come back when top quality is worth paying for. Starter: Atmosphere over precision (its signature move) Midjourney's edge is mood, light, and composition. Give it an evocative scene rather than exact specifications and it fills in beauty that other generators miss. Change the lighting words — "foggy morning," "neon night," "candlelit" — that's what sells the whole mood. Prompt: a cozy bookshop on a rainy evening, warm golden light spilling onto wet cobblestones, cinematic, painterly Starter: Pick a visual style by name Naming a medium or era is where Midjourney's huge stylistic range shines — most generators flatten everything toward the same look, and this one doesn't. Keep the subject, change only the style phrase, and watch how far the range goes. Prompt: a red fox curled asleep, in the style of a vintage watercolor children's book illustration Starter: Set the shape for a real use (the one setting worth learning) The --ar 16:9 at the end makes a wide image instead of the default square — the one parameter worth knowing on day one, turning a generation into a usable wallpaper or page header. Try --ar 9:16 for a phone wallpaper, or --ar 3:2 for something print-shaped. Prompt: a minimalist mountain range at dawn, muted earth tones, soft mist --ar 16:9 Starter: The portrait test (and its honest limit) Faces and hands are far better than Midjourney's early reputation suggests — a good way to see how far it's come. The honest catch: for the same character across many images, it still drifts. Add an emotion — "contemplative," "laughing" — the expression is where it tends to surprise you. Prompt: a friendly elderly fisherman with a weathered face, soft window light, photographic portrait Pitfalls: - There's no free tier — you pay before your first image. Unlike most tools here, you can't try it free; budget for at least the Basic plan before you sign up. - Everything you make is public by default. On the plans below Pro, your prompts and images are visible to others in Midjourney's gallery — don't generate anything you'd want kept private unless you're paying for Stealth Mode. - You're buying GPU time, not a set number of images. Basic's allowance runs out, then fast generation stops; Standard's unlimited "Relax mode" (slower) is why most people move up. - Overstuffing the prompt. Beginners pile in twenty details at once and get a muddle. Start with a clear, simple description and add one element at a time. - Stopping at the first four. The opening grid is a starting point, not the answer — pick the closest, use Vary, or run it again. The good results come from iterating. Where to next: - Image Generation & Editing: https://www.launchpadhq.io/discover?category=image-generation-editing - Video Creation & Editing: https://www.launchpadhq.io/discover?category=video-creation-editing ### Getting Started with Mistral Vibe (formerly Le Chat) Canonical: https://www.launchpadhq.io/platform/mistral-vibe/guide Data: https://www.launchpadhq.io/data/v1/guides/mistral-vibe.json Last reviewed: 2026-10-14 Tool: Mistral Vibe (https://www.launchpadhq.io/platform/mistral-vibe): starting price Free, from $14.99/month; free tier yes; data as of 2026-10-25 The French-built assistant that also works through multi-step jobs — from a company that publishes its AI for anyone to run. You type what you want in ordinary language and it writes back, in a browser at chat.mistral.ai or its phone and desktop apps. For anything bigger than a question you attach files or point it at tools you already use, and it works through the job a step at a time while you watch. What it is: - It drafts, summarises documents you upload, searches the web and makes images — and chains those steps toward one goal rather than answering one message at a time, which makes it an agent: AI that takes actions on its own. Chat handles quick questions; Work handles multi-step jobs across your email, calendar, Slack or Drive; Code is for developers. - Made by a French company, and for some that is the point. Mistral AI is registered in Paris and is your data controller under GDPR — European privacy law governing what companies may do with your data. It also publishes several of its models — the AI "brains" that do the actual thinking — for anyone to run, Mistral Small 4 under a permissive Apache 2.0 licence. Vibe itself is hosted like any other assistant, but that is an exit. - On everyday questions you won't tell it apart from ChatGPT, Claude or Gemini — pick it for the European story or the open models, a rival for the far bigger pile of beginner tutorials. Before you start: - Free to start: an email address and two minutes, no card. Before you type anything, open Settings and deal with the privacy toggle below — the one setting that defaults against you. - The free tier limits you by mechanism, not by a published number. Messages, web searches, coding sessions and images are capped; hit a cap and you wait for a reset. It is not a trial that expires but a permanent account that slows when you lean on it. Mistral publishes no message count, so treat any daily-limit figure as a guess. - You will feel the ceiling only running multi-step Work tasks daily. Paid plans are Pro at $14.99/month, with a discount for verified students, and Team at $24.99/user/month. Security: - Mistral's privacy policy says it prioritises EU providers that adhere to GDPR, attaching the European Commission's Standard Contractual Clauses in the exceptional cases where it uses one outside. Stronger than most assistants offer — but not a promise your data never leaves the EU, and not the data residency sold to Enterprise. Security: The setting to change first: - Mistral's documentation states Vibe users are not opted out by default — your conversations may train its models unless you say otherwise. Enterprise customers are opted out; you are not. - On web: Settings, then Vibe under Manage, then Privacy, and disable Allow your interactions to be used to train our models. - On iOS or Android: Settings, then Data & Account Controls, then deselect data sharing. - Mistral's separate developer service has its own toggle; turning one off does not turn the other off. Security: What is kept, and what to withhold: - Conversations are retained until you delete the chat or your account — deleting a chat is a real action, not tidying. - Connectors are the highest-stakes setting: linking email or Drive gives the assistant reach into that whole account. Connect one at a time; disconnect what you stopped using. Best at: - Everyday writing: drafting, rewriting, tightening - Answering specific questions about long documents you upload - Multi-step tasks across several sources, where Work mode earns its keep - European-language work, French especially Okay at: - Web research — it searches and reads pages, but check what matters - Coding help for beginners; Code mode targets developers Avoid: - Assuming your conversations are private — a personal account is opted in to model training until you change it, on web and phone - Repeating "my data stays in Europe" — the policy prioritises EU providers, it doesn't promise EU-only - Following an older tutorial's clicks — Le Chat became Vibe in May 2026 and the interface those guides describe is gone Starter: Make it explain something at your level Chat mode descends directly from Le Chat: fast single answers, nothing spent on the free plan. Prompt: Explain how AI writing tools produce text to someone with no technical background. One everyday analogy, under 200 words, ending with one thing people commonly get wrong. Starter: Upload a document and interrogate it Work mode takes documents, spreadsheets, PDFs and images directly, and Libraries holds files you question weekly so you don't re-attach them. Prompt: I am attaching a document. Give me a five-bullet summary, then list every deadline, obligation or number I would regret missing. Flag anything ambiguous rather than guessing. Starter: Test its honesty before you trust it In Work mode the steps are visible as it runs, so you see which page a claim came from and can redirect it mid-run. If you want this weekly, Vibe can re-run it on a schedule instead. Prompt: What were the three biggest changes in my industry in the last two months? Search the web, cite a link for each claim, and say what you could not verify. Pitfalls: - The training default runs the wrong way. Personal Vibe users are not opted out of model training unless they opt out — do it in the first five minutes, on web and mobile. - The name changed and the internet hasn't caught up. Le Chat became Vibe in May 2026, but most guides still describe an interface that no longer exists. Your account carried over; the tutorials did not. Where to next: - Compare it with the other big assistants: https://www.launchpadhq.io/discover?category=text-conversational-ai - Explore the open models behind it: https://www.launchpadhq.io/discover?category=local-open-source-ai ### Getting Started with Mubert Canonical: https://www.launchpadhq.io/platform/mubert/guide Data: https://www.launchpadhq.io/data/v1/guides/mubert.json Last reviewed: 2026-10-14 Tool: Mubert (https://www.launchpadhq.io/platform/mubert): starting price $14/mo; free tier yes; data as of 2026-10-14 Royalty-free background music in a couple of minutes — and the licence rules that decide what you may do with it. You open Render, type a short description of the music you want into the box (this is your prompt — the message you type), or pick a genre and mood from the lists instead, and press Generate. A finished instrumental track plays back a few moments later. Regenerate it or download it. What it is: - Mubert is an AI music generator aimed at background music — the bed that sits under a video, a podcast, a stream or a game, rather than songs you would listen to on their own. - It is built differently from the song-writing generators it gets compared with. Mubert assembles a track from a library it says holds more than 2.5 million short pieces of recorded music, contributed by artists it pays through its Mubert Studio programme, rather than composing a whole song from scratch. The output is instrumental: there are no lyrics, and where a vocal layer exists it is textures and short vocal samples rather than singing. - Mubert's site sells four things, and only one of them is this page's subject: Render, the track generator; Studio, for musicians contributing material; a developer service for putting music inside other products; and Mubert Play, a listening app. The homepage's main button, "Generate a track now", takes you to Render. - Nothing here is sold to you. Mubert keeps the copyright in every track it generates and grants you a licence to use it inside something else — royalty-free in the sense that you never pay again per play, not in the sense that it becomes yours. *Which* uses you are allowed is the entire substance of its plans, and the main thing to understand before you start. Before you start: - A free account — email, Google or Facebook — puts you on the Ambassador plan: a monthly allowance of generations, but a much smaller monthly download allowance, MP3 only, and a track length you cannot change, because the Duration field is locked behind an "Upgrade to edit" note. Downloads, not generations, are what you run out of first. Mubert's pricing page calls this plan Ambassador and its licence page calls it Free; they appear to be the same plan. Its public pages also don't make clear whether you can generate a track before signing in — you certainly need a free account to download one. Mubert's terms require account holders to be 18, or to have a parent's permission. - The free plan is not licensed for anything you earn from, and it may not be silent either. Mubert's plan-comparison table lists an audible watermark on Ambassador downloads and its Terms of Use describe free tracks as watermark-protected, although its FAQ says the watermark is removed once you download — so treat a free download as marked until you can hear otherwise. Crediting is not optional: the FAQ calls it obligatory. - Where the money line falls depends on what the music is *for*, not how much you make. Creator adds higher-quality WAV files, tracks up to 25 minutes and promoted or boosted posts. Pro is the first plan cleared for a monetised channel, digital ads and commercial content. Business, the top plan, covers client and agency work, apps, games and sub-licensing. Annual billing currently costs less per month, and current prices are on its Launchpad page. Nothing on the public pricing page suggests the free plan asks for a card, though Mubert doesn't say so outright. - Paid plans renew automatically until you cancel them in My Profile, and Mubert's terms say payments are non-refundable because the music is delivered immediately. Card only — Visa, MasterCard or American Express; PayPal isn't supported. Security: - Mubert asks for very little. An account is an email address or a Google or Facebook login; if you pay, Mubert's privacy policy says card details sit with third-party processors such as Stripe rather than with Mubert, and that it shares your data with nobody else except on a court or government request. Security: - The gap worth knowing about is what that policy doesn't cover. Mubert Render's privacy policy is marked "Last revised 31.08.2021" — years before the text-prompt and image-to-music features the site now leads with — and it accounts only for account details, payment data, contact forms, cookies and analytics. It says nothing either way about what becomes of the descriptions you type or the images you upload, so it is safest to assume they may be kept. Security: Two habits that follow from that: - Don't upload an image you don't own, or one showing someone identifiable — nothing published tells you where it goes - Delete the account from My Profile if you stop using it; the policy says profile data is stored "as long as it is necessary", which is not a period Best at: - Instrumental beds that sit under a voice — narrated video, podcasts, streams, presentations - Long, continuous audio rather than a three-minute song: paid plans generate up to 25 minutes in one piece - Starting from a library instead of a blank box — roughly 12,000 staff-picked tracks are browsable by genre, mood, tempo (labelled BPM), theme and instrument Okay at: - Matching a reference track: you can paste a link and Mubert analyses its style rather than copying the composition, but the field is marked beta on the page - Precise control: every plan can regenerate or remove individual instrument layers — Mubert calls them stems — with Quick Remix, but reshaping a track's structure, tempo or key needs the full Track Editor on Pro or Business Avoid: - Releasing what you make as music. Every plan, paid ones included, forbids putting Mubert tracks on Spotify, Apple Music, YouTube Music or a stock library, and forbids registering them in Content ID (the system that spots a track inside a video and claims it for whoever owns the music). - Anything you need to carry on making after you stop paying. Mubert's FAQ is explicit: when a subscription ends you may leave already-published projects up and keep monetising them, but you may not use those tracks in new projects. - Background music inside an app, a game or software you are building. Everything below the top Business plan excludes it, and the licence agreement tells software developers outright that they cannot use tracks within a mobile app or computer software. Starter: Describe the room, not the song Every Mubert track is arranged out of short recordings already sitting in its own licensed library, so naming textures and instruments gives the arranger something it can actually match in there. Naming a band or a chart song matches nothing, because no recording of theirs is in the library to find. The box is labelled Type text or upload image and its placeholder reads Type anything in English — write the description in English even if you work in another language. Prompt: warm lo-fi hip hop with soft rhodes piano, brushed drums and vinyl crackle, relaxed and unobtrusive, for a study vlog Starter: Choose the track type before you press Generate Next to the prompt box is a track-type selector with four settings, and it changes the shape of what comes back more than the description does. Track is standard length. Jingle makes a complete short piece of up to about 40 seconds, rather than a fade-out of a longer one. Loop makes something built to repeat seamlessly. Mix chains generated tracks one after another like a DJ set, which is how you get hours of stream or study-channel audio out of a 25-minute ceiling. Suno and Udio give you one output shape and let you trim it. Mubert asks the question up front because it is assembling material, not writing a song. Starter: Aim a track at the job it has to do Mubert's own FAQ says you cannot combine two genre categories in the browse filters and tells you to add a text description instead — which makes the prompt box the only place a mixed instruction, or a negative one like "no drums", can go at all. Swap the final clause for "building to a lift in the last fifteen seconds" when the video ends on a call to action. Prompt: calm ambient pads under a corporate explainer video, no drums, no melody that competes with a voiceover, steady and neutral all the way through Starter: Check the audio before you publish, not after Mubert runs a free copyright checker at mubert.com/tools/copyright-checker that needs no account and no plan. Export your finished audio as MP3, WAV or AIFF under 50MB, upload it, and it returns a risk estimate for whether a copyright-matching system will flag it. It is the only part of Mubert that will look at audio you did not make there. Epidemic Sound clears its own catalogue inside its own player; this reads a file you already have, which is what you need when a video mixes music from more than one place. Mubert calls the result a risk estimate and not legal advice, and says a "low risk" verdict can simply mean a track has not been indexed yet. Starter: Download the licence, not just the track Every download comes with a licence certificate file. Mubert's FAQ says that if a platform files a false-positive copyright claim against your video, that certificate — the one issued when you downloaded the track — is what you upload to dispute it. Save it beside the audio, named for the project it belongs to. Because Mubert keeps the copyright and licenses you the use, the proof that you are allowed to use a track lives in a separate file rather than in the fact that you made it. On Suno's paid plans the ownership itself transfers; here, losing the certificate leaves you with audio and no evidence. Pitfalls: - Mubert's own documents disagree about the Creator plan, on exactly the question that costs money. The pricing table and the FAQ both call Creator non-commercial and say monetising needs Pro; the Subscription License Agreement — the binding document — says the Creator Licence lets you "publish and monetize your Derivative Works on Social Media". A third page, Mubert's marketing licence page, prints "Cleared for monetization" without naming a tier at all, which reads more permissively than the plan table it links to. Neither the Terms of Use nor the Subscription License Agreement carries an effective or last-revised date, so there is no way from the public pages to tell when either was last changed; both were in force on 15 September 2026. Until Mubert reconciles them, the safe reading when money is involved is the stricter one, Pro, unless you get the answer from Mubert in writing. - A remix costs a generation. Quick Remix and the Track Editor each produce a new version of the track, and the FAQ confirms each one draws on your monthly generation allowance — so nudging a track you nearly like costs the same as starting again. - The required credit is not written the same way twice. The FAQ asks for a link to Mubert Render plus a #mubert hashtag; the Terms of Use ask you to print "This free music track provided by Mubert Inc by Mubert Render Service. All rights reserved and belong to Mubert Inc."; the Subscription License Agreement asks for a link to mubert.com on every page the music appears on. Satisfying all three costs one line of description text, which is cheaper than guessing. - Regenerating drifts. Mubert's FAQ warns that because generation carries variation, a regenerated layer "may sound different from what you expected and can slightly affect the overall feel of the track" — so a near-miss can get further away rather than closer. The original version is kept, so go back to it rather than remixing a remix. Where to next: - Video tools to put the music in: https://www.launchpadhq.io/discover?category=video-creation-editing - Other music generators: https://www.launchpadhq.io/discover?category=music-generation - Voice and speech tools: https://www.launchpadhq.io/discover?category=voice-speech ### Getting Started with Murf AI Canonical: https://www.launchpadhq.io/platform/murf-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/murf-ai.json Last reviewed: 2026-09-17 Tool: Murf AI (https://www.launchpadhq.io/platform/murf-ai): starting price Free, from $29/mo; free tier yes; data as of not dated Turning a written script into a finished voiceover — and what it actually costs to keep one. You open a project in your browser, type or paste your script into a text editor that splits it into paragraph-sized blocks, pick a voice from a list and press play. Murf reads it back in seconds. You then adjust the wording, the voice, the speed or the pauses and play it again until the read sounds right. What it is: - Murf AI is a browser-based text-to-speech studio: you give it written words, it gives back a recorded-sounding voice reading them aloud. It is made by Murf Inc., a San Francisco company, and has run since 2020. - It is aimed at people producing narration in volume — e-learning modules, explainer videos, YouTube scripts, product demos, phone-system greetings — rather than one-off novelty clips. The editor also carries a timeline for images and video, so a narrated video can be assembled without leaving it. - Murf currently advertises somewhere between 200 and 300 voices across roughly 30-plus languages and accents; its own pages give different counts in different places, so treat the headline as approximate and check the voice list for the language you need before you commit. Before you start: - Signing up is free and needs no card. You get a small one-time allowance of voice generation plus transcription minutes. It has no expiry date and does not refill monthly — Murf describes it as a one-time credit that simply ends when the minutes are used up. - The free plan cannot download anything. Murf's help centre states that exporting is available on paid plans only, and its pricing table marks the free plan "No Commercial Rights". You can listen, share a preview link and embed the result on a page; you cannot save the audio file. So you pay to keep what you make, on Creator or the pricier Business plan. Murf advertises a sizeable discount for paying yearly, so month to month costs meaningfully more; current prices are on its Launchpad page. - Paid allowances are measured in Voice Generation Time — Murf's credits (the platform's unit of spend), counted in minutes of finished audio rather than words typed. Murf sells it by the year, and Business includes several times Creator's allowance. As a rule of thumb Murf publishes, a 1,000-word English script uses roughly six minutes. - Voice cloning — making a synthetic copy of one specific real person's voice — is not part of what you can buy here. Murf's help centre says cloning is available only on its Enterprise plan and is "not a self-serve solution": you arrange it through their sales team. It currently covers English, Spanish, Portuguese, French and German. - Everything runs in the browser. There is nothing to install, though Murf offers optional add-ons for Canva, PowerPoint and Google Slides on paid tiers. Security: - Your script, any audio you upload and everything it generates sit on Murf's servers. Murf says it keeps all customer data in the United States (Amazon's Ohio region), encrypts it at rest, and holds SOC 2 Type II (an audit of how a company handles customer data) along with ISO 27001 and ISO 42001 certifications. Security: - The detail worth noticing is on the price list rather than the security page: Murf's plan-comparison table lists "No Training on your Data" as an Enterprise-tier feature, and its published privacy policy does not say either way what happens on the cheaper plans. Safest to assume scripts pasted on a Free, Creator or Business account may be used to improve Murf's systems, and to keep unreleased, confidential or client-owned material out until you have asked directly. Security: If you ever go as far as cloning a voice: - Murf's terms require explicit written consent from the person whose voice it is — submitting someone's recordings without it is prohibited outright, not merely discouraged - The same terms say Murf staff, service providers and affiliates may listen to your training audio and your synthesised audio to check quality - They also reserve the right to have the cloned voice speak a set of non-defamatory test phrases internally for quality assurance Best at: - Long narration read straight through — e-learning modules, explainer videos, audiobook-style reads — where one voice carries many paragraphs - Turning a script you have already written into audio without booking a recording session - Trying the same paragraph in a dozen different voices before committing to one - Scripts with more than one speaker, where each paragraph is assigned its own voice - Laying a voiceover over your images and video inside a single editor, rather than exporting audio and importing it elsewhere Okay at: - Emotional or performed delivery — the emphasis and pacing controls are real, but you are steering a read rather than directing an actor - Languages other than English: how many voices exist per language is uneven, so check yours specifically - Very short social clips — the project-and-blocks structure is built for scripts, and is more machinery than a fifteen-second hook needs Avoid: - Cloning a real person's voice on a whim. The written consent above is a contractual requirement, not a courtesy, and cloning is not a button you will find in the plans you can buy. - Anchoring a long-running series to one particular voice with no fallback. Murf's voices are licensed from real voice actors, and when an actor ends the collaboration Murf retires that voice from the Studio. You get notice, but not on your schedule. - Feeding Murf-generated audio into another AI system as training material, or reselling access to the voices themselves. Murf's terms forbid both by name, separately from the ordinary commercial licence. Starter: Hear the same paragraph in a dozen voices Paste it into a new project's script box, press play, then keep swapping the voice on the same text. Murf charges by the length of audio it generates, and its documentation says changing the voice actor, style, pitch, speed, pause, emphasis, pronunciation or volume on already-rendered text consumes no further voice generation time — so the twelfth voice you audition costs what the first did, which is nothing. Paste a paragraph of your own writing instead — you are listening for whether the voice suits your rhythm, not Murf's demo copy. Prompt: Every product has a story, but most of them are told badly. The trick is not to say more. It is to say less, and to mean it. Start with the one thing your customer cannot stop thinking about, and build outward from there. Starter: Give a two-person script two voices Paste it in, accept the split-by-paragraph option, then assign a different voice to alternating blocks. Murf's split screen turns each paragraph into a block and optionally each sentence into a sub-block, and voice selection is per block rather than per project — so a two-hander is paste-and-assign rather than two renders you stitch together. Tick "Don't Split Sentences" if you want each speaker's lines to flow as one continuous take instead of sentence by sentence. Prompt: So what changed? Six months ago you said this was impossible. It was impossible the way we were doing it. We stopped trying to fix the old process and threw it out instead. That sounds expensive. It was. It was also the only thing that worked. Starter: Work out whether your script fits before you type it Count the words in the script you actually want to make, divide by 1,000, and multiply by six. That is roughly how many minutes it will cost. Murf publishes that conversion itself, which means you can decide before opening an account whether your real project fits inside the one-time 10-minute allowance or blows straight through it. Add a margin for the re-renders below. Starter: Teach it a name it keeps getting wrong Render it, listen for what it mangles, then use the pronunciation control to correct each word. Corrections go into a pronunciation library Murf's comparison table marks as available from the cheapest paid tier upward, scoped to your own account — so a surname or acronym you fix once carries into your next project. Front-load it with the names, brands and jargon from your actual work — that is the test of whether Murf can handle your vocabulary. Prompt: Welcome to the Q3 review from Aoife, Nguyen and Sioban at NCAA headquarters. Our EBITDA rose 14% year over year, and the Reuters piece on Worcestershire was, frankly, premature. Starter: Get something out of a free account without paying Once you have a render you like, skip the download button and use Share project preview or Create Embed link instead. Murf's own free-trial feature list includes preview sharing and embed links alongside the note that downloads are not included — so on a free account the work leaves as a link someone can play in a browser, not a file you own. Enough to ask a client whether the voice is right before anyone spends money. Send the link, not a description — whether it sounds right is the only question a sample answers. Pitfalls: - Re-rendering costs again. Voice generation time is deducted every time Murf produces audio for a sentence, so each round of rewriting and replaying a line spends the allowance again. Editing a script you have already rendered costs more minutes than the word count suggests. - Unused time is more fragile than it looks. Murf carries leftover time forward exactly one billing cycle; carried-over time unused in the second cycle disappears rather than accumulating. Cancelling forfeits it outright, though pausing keeps it. - Murf's own background music can get your video flagged on YouTube. Its soundtrack library triggers YouTube's automatic copyright matching, sometimes with a revenue-sharing notice attached. Murf issues "YouTube codes" you add to your channel to clear these — a step you have to know about before you publish. - The refund window is genuinely narrow. Murf's terms make a purchase refundable only if it was bought within the last 24 hours and you have used under 10 minutes of voice generation. Past either line, the money is spent. - The cheapest paid plan does not include the delivery controls the demos show off. Murf's comparison table marks Emphasis, Variability and "Say It My Way" as Business tier and above, while Creator gets pronunciation editing but not those. If control over how a line lands is what sold you, check which tier the demo was running on. Where to next: - Other voice and speech tools: https://www.launchpadhq.io/discover?category=voice-speech - Video creation and editing: https://www.launchpadhq.io/discover?category=video-creation-editing - Education and learning AI: https://www.launchpadhq.io/discover?category=education-learning-ai ### Getting Started with n8n Canonical: https://www.launchpadhq.io/platform/n8n/guide Data: https://www.launchpadhq.io/data/v1/guides/n8n.json Last reviewed: 2026-10-14 Tool: n8n (https://www.launchpadhq.io/platform/n8n): starting price Free, from €20/month; free tier yes; data as of 2026-10-25 Connect your apps and let them pass work to each other — with a version you can run yourself, free, forever. You build on a canvas. You drop down a trigger — the "when," like a set time, a new email, or a form being filled in — then chain actions after it: send this, save that, look this up. You can run the whole chain with your real data before you ever turn it on. What it is: - A workflow automation tool: you build chains of steps that run without you, connecting apps you already use. It ships with hundreds of prebuilt connections and a library of ready-made workflows to adapt. - It's noticeably more capable than Zapier or Make, and noticeably harder to learn. Both are true. - The reason to pick it is the billing shape. Zapier and Make charge roughly per step; n8n's paid cloud charges per execution — one full run of a workflow, however many steps it contains — so a 3-step and a 30-step workflow cost the same. That changes what you're willing to build. Before you start: - There are two n8n's, and confusing them is the most common beginner mistake. n8n Cloud is the version n8n runs for you: sign up, build, pay monthly. Self-hosting means running the free Community edition on a machine you control: almost the full feature set, no cap on how many times your workflows run (nobody else in this category offers that), and the maintenance is yours. - Cloud needs only an email address, and the trial needs no card. Self-hosting needs a machine that stays on, and comfort typing instructions into a text window instead of clicking — n8n's own docs list technical expertise as required, and point non-technical readers to Cloud. - So if you're not technical you'll pay after the trial; if you'll self-host, possibly never — you pay in time and responsibility instead. Cloud has no free plan: Starter is €20/month and Pro €50/month, both billed annually, each with a monthly execution allowance. Those are euros, not dollars. Free means you run it yourself. - First step: start a free Cloud trial and open the template library rather than a blank canvas. Find a template near what you want, run it once, change one thing. Security: - Every app you link hands n8n a stored credential that lets it act as you — and a workflow with a mistake in it acts as you at machine speed, over and over. Connect one account at a time, only what a workflow needs, and run everything manually a few times before switching it on. Any connection can be revoked from the app's own settings, not just from n8n. Security: If you self-host, these become your job: - Keeping the software updated — security fixes don't install themselves. - Backups: your workflows and stored credentials live on your machine, and if it dies, they die. - Locking down access, so your n8n isn't sitting open on the public internet. Best at: - Multi-step workflows where the step count would get expensive elsewhere - Reaching a service with no prebuilt integration, via a plain web request - Adding an AI step to an ordinary automation, on hardware you control Okay at: - Being your first-ever automation tool — you'll hit concepts other tools hide - Team work on the free tier, where sharing and permissions are limited - Quick one-offs you'll build in five minutes and never touch again Avoid: - Choosing it to never see anything technical — Zapier is the kinder first step - Self-hosting because it's free, if you won't maintain it Starter: When a form is submitted → log it to a spreadsheet and email you n8n hosts the form itself, drops each response into a Google Sheet, and emails you the details. n8n makes you place and wire the boxes yourself, and run the chain with real data while it's switched off — so you watch data pass between two nodes (one step in a workflow, shown as a box you connect). Nothing runs until you switch the workflow on — canvas testing works fine while it's off. Starter: When an email arrives with attachments → save every file and log the sender An email trigger fires on each new message, a step saves every attachment to a folder in your file storage, and another appends the sender and subject to a spreadsheet. An email with five attachments is one execution in n8n. On a per-step tool, looping over each attachment is where your quota quietly evaporates — the pricing difference made visible. Filter to only invoices, by matching the subject line. Starter: Every weekday at 8am → gather headlines, summarise with AI, post to chat A schedule trigger runs at 8am on weekdays, a step fetches headlines from a feed, an AI step summarises them, and a chat step posts the result. AI steps are ordinary boxes here rather than a separate product you bolt on, and you can point one at your own model key instead of buying platform credits. Starter: When a row is added to a sheet → call a service n8n has never heard of Uses n8n's general-purpose web-request box to talk to a service directly, instead of waiting for an official integration. This is the honest reason people switch: on Zapier, no integration means you're stuck; here, if a service publishes a way for programs to talk to it, you have one. Start with a public service that needs no login, so you see it work first. Pitfalls: - "Free" and "open source" aren't the same thing here. n8n publishes its source but licenses it as fair-code, under its Sustainable Use License: free for personal or internal use, but you may not resell it, white-label it, or host it commercially for others. Fine for almost every beginner; worth knowing before you build a business on it. - Data moves as a list of items, not one thing. A step usually runs once per item it receives — powerful, and the source of most confusing early results. - Templates from strangers carry credentials-shaped holes. They're a great way to learn, but read what a workflow does before you connect your accounts. Where to next: - More automation tools to compare: https://www.launchpadhq.io/discover?category=workflow-automation - Connect AI models directly into your workflows: https://www.launchpadhq.io/discover?category=ai-apis-developer-services ### Getting Started with Nanonets Canonical: https://www.launchpadhq.io/platform/nanonets/guide Data: https://www.launchpadhq.io/data/v1/guides/nanonets.json Last reviewed: 2026-09-09 Tool: Nanonets (https://www.launchpadhq.io/platform/nanonets): starting price Free, from $499/mo; free tier yes; data as of not dated Feed it the same document a hundred times and it learns your fields — the first twenty are the work. You upload documents — or forward them from an email inbox — and Nanonets reads them, showing the fields it found beside the page it read them from. You check each field, fix what it got wrong, and approve. What you approve feeds back in, so the next batch arrives needing fewer corrections than the last. What it is: - Nanonets pulls structured data out of documents that arrive as files — supplier invoices, purchase orders, bank statements, delivery notes, identity documents — and hands the result to the software you keep records in, so nobody retypes it. - It is built for the back office rather than the desk. The people it is designed for handle the same kind of document hundreds of times a month and want that handling to stop being somebody's afternoon. - Beyond reading, it will act on what it read: route anything it is unsure about to a named person, then push the approved result into accounting software like QuickBooks or Xero, or a larger business system like SAP or NetSuite. - If you have one contract to understand, this is the wrong shape of tool — a document chat tool answers you in a minute. Nanonets only pays back on repetition. Before you start: - Every account currently starts with free credits (the platform's unit of spend — each thing you run costs some) and no card, and Nanonets says credits never expire. That is enough to put real documents through before you decide anything. - The prices are published, which is unusual in this category: each step has a listed rate, with a standard AI step costing more than a simple one and a data-extraction run more again. There is one self-serve paid plan; above that the buttons read Talk to sales and Book a demo. - Bring a stack, not a file. Nanonets' own documentation describes a new extraction starting at 0% confidence and rising only as you correct and approve documents, with 90% named as the point it is ready to run unattended. One document cannot move that number. - You do not need accounting or ERP software connected to try it. The free plan currently includes importing from an email address and from cloud folders such as Google Drive, Dropbox, OneDrive and SharePoint, and results can come back out as a spreadsheet. - The free plan is currently capped at three users with community support, so it is a working trial rather than a small-team licence. Security: - The stake here is different from a chatbot's. You are not typing questions — you are uploading whole documents, and the documents this tool exists to read are the sensitive ones: supplier bank details, payment amounts, staff expenses, and, if you use its pre-built identity-document extraction, driving licences. One upload can contain more personal data about other people than a year of chat. Security: What Nanonets publishes: - SOC 2 (an audit of how a company handles customer data) — Nanonets states it is SOC 2 compliant and ISO/IEC 27001 certified, with Type I and Type II reports available on request. - AES-256 encryption at rest and 256-bit encryption in transit. - All customer data stored in the USA. Data residency elsewhere, private-cloud or on-premises deployment, and single sign-on sit in the Enterprise tier. - Deletion on written request to [email address on the web page], completed within 30 days. - Where an extraction is personalised to your account, the privacy policy states it exists solely to serve you and is never used to serve another customer. Security: Three things worth reading before you upload anyone else's documents: - The terms let Nanonets "freely use Derived Data for its internal business purposes (including without limitation, for purposes of improving, testing, operating, promoting and marketing Nanonets's products and services)". That is a narrower promise than "we do not use your data", and it is worth knowing which one you are relying on. - The strongest no-training language in the privacy policy is scoped to data pulled from your Google account, not to everything you upload. Read it as a Google-integration commitment rather than a blanket one. - The privacy policy says processing may send data to third-party model providers, but the published subprocessor list carries an effective date of December 2, 2020 and names none of them — only cloud hosts and support tools. If you handle regulated material, ask who sees it before you upload, rather than inferring an answer from that page. Security: - One asymmetry beginners miss: under GDPR (European privacy law governing what companies may do with your data), if you upload an invoice containing a supplier's employee's name and bank details, you are usually the party answerable for it, not the tool. Nanonets publishes access and erasure rights for its own users; it does not carry your obligation for the people inside the documents. Best at: - The same document a hundred times over — invoices, purchase orders, bank statements, delivery notes - Documents that arrive in a mess of formats, where no two suppliers lay an invoice out the same way - Line-item tables, which it is built to return as rows rather than as one block of text - Getting the result into accounting or business software without anyone retyping it - Saying when it is unsure, rather than passing a doubtful read through silently Okay at: - Phone photographs and poor scans, which cap everything downstream - A layout it has never seen, where its documentation notes confidence can fall back - Handwriting, which its training guidance does not address either way - Anything you only need to do once, where correcting and approving costs more time than reading it yourself would Avoid: - Taking the accuracy figures on the marketing pages as your own. Percentages like 93% touchless processing and 95% extraction accuracy are presented as customer outcomes, while your own extraction starts at 0%. Both are true and they describe different months. - Uploading identity documents or regulated records before you have asked which outside model providers see them — the privacy policy says data may go to them and the published list of subprocessors does not name any. - Assuming that because Nanonets says a personalised extraction only serves you, your data is off-limits for product improvement. The Derived Data clause in the terms is the sentence that governs that, and it points the other way. - Planning around private-cloud deployment, data residency or single sign-on on the free or self-serve paid plan — Enterprise is a conversation, not a checkout. Starter: Run twenty of the same document, not one You collect the last twenty invoices from a single supplier, upload them together, and correct every wrong field before approving any of them. A one-file test measures nothing, because the mechanism the tool is built on — confidence rising only as you correct and approve — has not run yet. Pick the supplier whose invoices are ugliest. Clean ones flatter it. Starter: Time yourself correcting You time how long checking and fixing the extracted fields takes across ten documents, then divide by ten for a per-document figure. Nanonets is unusual in publishing what each run costs up front, so the machine side of the bill is knowable on day one. The minutes a person spends fixing fields before approval are the half nobody publishes, and early on they are the larger number. Take the same measurement again after fifty approvals. The gap between the two readings is the whole return. Starter: Feed it the same receipt twice You upload a phone photograph of a crumpled receipt alongside a clean scan of that same receipt, then compare the two results field by field. Nanonets reports a confidence score for the extraction, not for the image it started from, so a bad photograph and a hard document look identical in the output. A controlled pair is the only way to see which one you actually have. Then try a layout you have never sent it. Starter: Send it in the way it will really arrive Instead of uploading a file by hand, you forward a real invoice email — attachment and all — to the address Nanonets assigns your workflow. Email and cloud-folder intake are both on the free plan, and they are the only test that matches how documents actually reach a back office — as attachments nobody downloaded. Forward one email with two attachments, and one with the invoice pasted into the body instead. The second is the case that breaks. Starter: Find out who gets interrupted You raise a field's confidence requirement until documents start failing review, then follow where the review request goes. Nanonets raises what it is unsure about in Slack, Teams or email rather than inside its own screen, so the person doing the checking may never open the tool. Deciding who that person is before rollout costs less than discovering it after. Pitfalls: - The correcting is the real cost. For the first few weeks the human hours spent fixing fields before approval are larger than what the tool saves. Budget the attention, not just the credits. - Accuracy claims assume a clean scan. The training guidance asks for images larger than 1000x1000 pixels and states plainly that blurry images hinder text recognition. A phone photograph of a crumpled delivery note is a different tool's problem. - A supplier's redesign can quietly undo months of progress. A workflow that ran unattended can start needing human eyes again with no announcement, so watch the confidence figure, not just the output. - Self-serve stops at one paid plan. The capabilities a regulated or larger team is most likely to need — private-cloud or on-premises deployment, data residency options, single sign-on, role-based access — are in that sales conversation rather than on the price list. Where to next: - Other document and PDF tools: https://www.launchpadhq.io/discover?category=document-pdf-processing - Workflow automation: https://www.launchpadhq.io/discover?category=workflow-automation - Finance and accounting AI: https://www.launchpadhq.io/discover?category=finance-real-estate-ai ### Getting Started with Napkin AI Canonical: https://www.launchpadhq.io/platform/napkin-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/napkin-ai.json Last reviewed: 2026-10-15 Tool: Napkin AI (https://www.launchpadhq.io/platform/napkin-ai): starting price Free, from $12; free tier yes; data as of 2026-10-14 Paste in something you've already written and get a diagram back — no design tool, and for the most part, no prompt (the message you type). You open a blank page in your browser and paste in something you've already written. Highlight a paragraph and click the blue Generate Visual button; a few seconds later Napkin shows several diagrams built from those words. Pick one, choose a style, and it drops into your page, where you can change colours, icons and wording by hand. What it is: - Napkin AI turns writing into diagrams — flowcharts, mind maps, timelines, comparison charts, data charts and similar visuals. You give it text you already have, and it reads that text and proposes several ways of drawing it. - That is the difference from an image generator. You are not describing a picture and hoping; Napkin lays out your own points as boxes, icons and arrows, and every piece stays editable afterwards. Its homepage puts it bluntly: "Forget prompting, Napkin works directly from your text." - It suits anyone who writes things other people have to follow — a report, a blog post, a training handout, a slide — and wants a diagram without learning a design program. Finished visuals can go into PowerPoint, Google Slides, Google Docs, Word, Notion, Canva and similar tools. - Napkin also has a newer product, Napkin Slides, still labelled beta, that builds whole presentation decks. This page sticks to single visuals, which is what Napkin is known for and where a first-timer should start. Before you start: - Sign up at app.napkin.ai with a Google account or an email and password. You can sign in from a phone or tablet, but Napkin's help centre says generating visuals needs a computer — on mobile you can only view your work and work on presentations. - There is a free plan, "Free Forever", with a weekly allowance of AI credits (the platform's unit of spend — each visual you generate costs some), and Napkin's branding on your visuals. The credits refresh weekly and do not roll over. Napkin's pricing page says generation is charged by the words you highlight, so the free allowance goes much further on a short paragraph than on a whole document. Editing a visual by hand costs nothing. - If you outgrow that, the paid plans, Plus and Pro, are priced per person, each with a much larger monthly credit allowance and a discount for paying yearly. For occasional diagrams in your own documents, the free plan is the realistic starting point. - Bring your own words. Napkin is built to draw text, not to write it — you can paste text in or import a document. Its first-run tutorial can write some sample text for you from a short prompt so you have something to practise on, but the useful habit is starting from your own notes. Security: - Napkin is a cloud service: the text you paste is stored with your account and sent to outside AI companies to produce the visuals. Napkin's privacy policy, effective 31 August 2026, says its data is hosted in the United States, names OpenAI, Anthropic and Google (Vertex AI and Gemini) among the services it uses for AI generation, and says your visuals and input text are kept for as long as your account is active. Security: Two things to do before you paste anything real:: - Decide about the training setting. Napkin's help centre says your content may be used in anonymised form to improve the product, "including training and fine-tuning our models", on every plan including Free. It says you can switch this off under Account Settings → Data Privacy by turning off "Use my content for product improvements" — in a shared Teamspace, the administrator controls it instead. - Treat a draft as something that leaves your computer. The privacy policy itself does not mention training at all — the opt-out lives in the help centre — so if your text is confidential work material, check with whoever owns it before pasting, setting or no setting. Security: - To delete your data, the privacy policy says you request account deletion; most personal data is then removed from active systems straight away, with backups purged later on their normal schedule. Best at: - Turning a paragraph that describes a process, a list of steps or a comparison into a clean diagram in seconds. - Offering several layouts for the same words, so you can see which shape — timeline, flowchart, mind map — explains your point best. - Visuals for documents and slides you are already writing. - Hand-tweaking the result — colours, icons, sizes, wording. Okay at: - Getting one exact layout. Napkin's own help centre says a typed request "guides the suggestions; it does not guarantee an exact layout," so expect to pick from options rather than order a specific design. - Long documents. It will import them, but generating from a whole report at once is the fastest way through a weekly allowance. Avoid: - Anything that has to be one of a kind, such as a logo-like graphic or a visual you need to call exclusively yours. Napkin's terms say outright that it does not guarantee your output will be unique and that other users may get "substantially similar" visuals. - Diagrams where every label must match your source exactly — figures in a report, steps in a safety procedure. Napkin can summarise as well as draw (there is a "Summary" detail level), and its own AI transparency page warns results can be "incomplete" or "misleading". Check every box against what you wrote. Starter: Draw a paragraph you already wrote Paste a paragraph describing something with steps — how you onboard a new hire, how a recipe comes together, how a bill gets approved. Highlight it, click the blue Generate Visual button, and scroll through the options it offers. There is no instruction to write here: Napkin reads the words on your page and proposes several layouts from them. Gamma, by contrast, starts from a topic you type and writes the content for you — Napkin's starting point is text you have already written. Highlight one paragraph, not the whole page: smaller selections cost less and usually produce a clearer diagram. Starter: Ask for a particular shape After you click Generate Visual, choose Customize at the bottom of the preview and type this into the "Specify visual" box, then apply it. Napkin re-suggests visuals in that form. "Specify visual" is the one place Napkin takes a typed instruction, and it is deliberately short — a visual type rather than a paragraph of directions. Try "flowchart" for a process or "mind map" for a topic with branches. The same Customize panel has Orientation (Landscape suits a slide) and a Detail level of Summary or Detailed. Prompt: timeline Starter: Restyle a visual without spending credits Double-click any text, icon or shape in a finished visual to change its colour; click an icon to swap it for another, or use Spark Search to find a new icon or shape. Napkin's help centre draws a hard line here: editing colours, icons, sizes and shapes "doesn't cost credits," but re-highlighting your text and generating again does. Once a layout is close, fixing it by hand is free; asking for a fresh one is not. Starter: Turn a whole document into a set of diagrams Import a document you already have — Napkin accepts PowerPoint, Word, PDF, HTML and Markdown files — then generate a visual for the one or two sections that would most benefit from a picture. Importing keeps your original wording on the page next to each visual, so you can check every label against the source — which matters, because Napkin can condense what it draws. Pick sections; don't generate for every paragraph. Starter: Get a visual out and into a slide or doc Hover over a visual, click the export button in its top-right corner and choose a format; or use Share → Export PDF to export the whole page at once. Export PDF turns the whole Napkin page, visuals included, into one file laid out as a continuous page, A4 pages or 16:9 presentation pages, so a document with several diagrams goes out as a handout or a slide deck without placing each picture by hand. For a single visual, PNG or PDF works in most documents and slides; SVG and PowerPoint files stay editable afterwards. Check which formats your account can export once you're signed in. Pitfalls: - Effects cost credits too. Napkin's help centre says applying a visual Effect uses credits for each visual or slide it is applied to, so check your balance before decorating a whole page. - Saving credits for later. Paid-plan credits don't roll over either; they reset monthly, so there is nothing to gain by hoarding them. - Letting a paid plan renew by accident. Napkin's terms say subscriptions renew automatically, that you must cancel at least 24 hours before the renewal date, and that there are no refunds for partly used AI credits. Where to next: - Image generators: https://www.launchpadhq.io/discover?category=image-generation-editing - Document & PDF tools: https://www.launchpadhq.io/discover?category=document-pdf-processing - AI chatbots, to draft the text first: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Naver HyperCLOVA X Canonical: https://www.launchpadhq.io/platform/naver-hyperclova-x-cue/guide Data: https://www.launchpadhq.io/data/v1/guides/naver-hyperclova-x-cue.json Last reviewed: 2026-09-13 Tool: Naver (HyperCLOVA X / Cue:) (https://www.launchpadhq.io/platform/naver-hyperclova-x-cue): starting price Free, from See pricing; free tier yes; data as of not dated Korea's home-grown AI — now built into Naver Search instead of living in a chatbot of its own. You go to Naver's search page and type a question into the same box you would use for a normal search — in Korean. Instead of only a list of links, a written answer comes back, and you can reply to it in the box to narrow things down, the way you would in a conversation. What it is: - HyperCLOVA X is Naver's own family of AI models (the AI "brain" that does the actual thinking), built in South Korea by the company that runs the search engine and portal most Koreans use instead of Google. - Until spring 2026 Naver ran it as two separate products — a chatbot called CLOVA X and an AI search tool called Cue: — and closed both on 9 April 2026, folding the technology into ordinary Naver Search instead. - What it does that nothing else does as well is answer questions about Korean life — shops, neighbourhoods, places, reservations — from Korean-language blog and café posts that other search engines index thinly or not at all. - Naver also publishes smaller versions of the same model family, called HyperCLOVA X SEED, as free downloads anyone can run on their own computer. Before you start: - This is a Korean-language service, and that is the real entry barrier — not the price. The answers are written in Korean and built from Korean web content. You can type English and get something back, but you are using it against the grain. Naver's search page is currently reachable from outside Korea, though the experience is built for people searching in Korean about Korea. If what you actually want is a general-purpose assistant in English, start with a Text & Conversational AI tool from this site and come back to Naver when the question is specifically about Korea. - Almost every guide you will find describes something that no longer exists. CLOVA X, the standalone chatbot at its own web address, and Cue:, the standalone AI search page, were both switched off on 9 April 2026 (Cue: had already stopped taking new sign-ups on 2 March). If a tutorial tells you to visit a separate site to chat, it predates that date and the site is gone. - There is nothing to buy. Naver Search is free, and the conversational AI answers went from a members-only beta to all users on 26 June 2026, so you will not need to pay to try any of this. The paid product carrying the HyperCLOVA X name is CLOVA Studio on Naver Cloud Platform — a developer and business tool that bills in Korean won by how much text it reads and writes, sold to companies rather than to people learning AI. - Reading an answer is one thing; acting on it is another. The feature is built to carry you through to a booking or a purchase, and that part runs on a signed-in Naver account. Whether the AI answers themselves need you to be logged in is something you may need to check once you are on the page. - If you would rather not deal with Naver at all, the downloadable SEED models are the way in — but they are model files, not an app, so you need a program that runs them and a computer with enough memory. The smallest is around half a billion parameters (a rough measure of a model's size — more is smarter but heavier) and runs on an ordinary laptop; the 32-billion one does not. Security: - Naver is a South Korean company and this is South Korean infrastructure — your questions are handled under Korean law and Korea's privacy regulator, not your own country's. Naver's own model licence makes the point in writing: it is governed by the laws of the Republic of Korea and disputes go to arbitration in Seoul. Nothing sinister in that; it just means the usual escalation route you have with a US or EU service is not the route here. Security: What changes once you sign in: - Naver's launch material describes the AI answers reaching into shopping, Places, maps and real-time reservations - The payment side sits behind Korean identity verification, which is a much heavier disclosure than an email-and-password sign-up - Treat a Naver account like the portal account it is — mail, payments, maps and search history under one login, not a throwaway for one AI feature Security: - Download the SEED models instead and the picture inverts completely: the files run on your computer and nothing you type leaves it. The risk moves to where the file came from. The official source is the naver-hyperclovax organisation on Hugging Face; the ready-made builds circulating under individual usernames in local-model catalogues are convenient re-uploads by other people, not Naver. Best at: - Questions about Korean places, shops and reservations - Reading and writing Korean naturally, including the difference between formal and casual register - Getting a written answer at the top of an ordinary search instead of working through a page of links Okay at: - English. Answers come back, but you are using it against the grain - General world knowledge that has nothing to do with Korea — you are paying a language tax for no benefit - The smallest SEED downloads — genuinely usable and genuinely simple; expect help, not brilliance Avoid: - Building anything on the free SEED downloads without reading the licence first. It is published openly but it is not open-source in the ordinary sense: Naver's own licence requires you to display "Powered by HyperCLOVA X" in your interface, to prefix any model you derive from it with "HyperCLOVA X", and to pass its Prohibited Use Policy down to anyone you share it with. It also reserves the right to withhold a licence entirely — at Naver's sole discretion — if your service passes 10 million monthly active users or "directly competes with any product and service provided by NAVER." - Anything you built or bookmarked against CLOVA X or Cue:. There is no drop-in replacement for either. - Reading Korean legal, medical or contractual material through this and then through a browser translator. You get a Korean answer summarising Korean sources, and a non-Korean reader is then two lossy steps away from the original wording. Starter: Ask it something only Korean sources know Naver's June 2026 launch wired the answer box straight into its own Places, blog and café data, so the recommendations are assembled from reviews Koreans wrote inside Naver. Perplexity, searching the open web, cannot reach most of that material because it was never indexed outside Naver in the first place. In English, that asks: Recommend three cafés in Seongsu-dong, Seoul where I can sit for a long time with a laptop. One line on why for each. Swap the neighbourhood and the activity — 헬스장 (gym), 미용실 (hair salon), 병원 (clinic) all sit on the same data. Prompt: 서울 성수동에서 노트북 들고 오래 앉아 있기 좋은 카페 세 곳 추천해줘. 각각 이유도 한 줄씩 써줘. Starter: Make it do the Korean, not you Naver publishes this model's scores against Korean-specific benchmarks — KMMLU, HAERAE, CLIcK — rather than the English suites most vendors quote, and politeness register is precisely the axis those measure. ChatGPT will also produce Korean, but nobody publishes its 존댓말 accuracy as a headline number. In English, that asks: Rewrite the English message below in Korean — polite formal register, for a landlord, four sentences or fewer. Change the recipient — 상사에게 (to a boss), 친구에게 (to a friend) — and watch the register move. Prompt: 아래 영어 메시지를 한국어로 바꿔줘. 집주인에게 보내는 정중한 존댓말로, 네 문장 이내로: "[paste your message]" Starter: Narrow it in the box, because there is nothing to click Naver retired its "related search terms" strip on 30 April 2026, three weeks after closing CLOVA X — the sideways nudge that used to sit under every result is gone, so refining a Naver search is now something you type rather than something you click. Google still hands you related searches under the page; here the box is the only lever you have. In English, that asks: Narrow that last answer by three conditions — takes reservations, has parking, reviews from 2025 or later. Ask it to show what it dropped and why — that is where you find out how thin the underlying set was. Prompt: 방금 답변을 세 가지 조건으로 다시 좁혀줘: 예약 가능, 주차 가능, 2025년 이후 후기만. Starter: Take the model home instead The SEED versions are free open-weight downloads — the model file itself is published, so you can run it yourself. This is the only route to HyperCLOVA X that needs no Naver account, no Korean phone number and no Korean-language interface, and it exists because Naver ships the weights with prepared builds listed for llama.cpp, Ollama and LM Studio on the model's own page. Google's Korean-capable Gemini models cannot be downloaded at all, at any size. Start at the 1.5-billion size. If your machine handles it comfortably, step up; see this site's Local & Open-Source AI page for how to run model files at all. Pitfalls: - Downloading the wrong app. "네이버 클로바 / NAVER CLOVA" in the app stores is the old smart-speaker companion app, not this. The AI answers live in the regular Naver app and on naver.com. - Reading the generated summary and skipping what sits under it. Those sources are Naver blog and café posts — user-written, and a well-known share of the restaurant and product ones are sponsored. The summary does not separate a paid post from an unpaid one. - Assuming the download is the same model as the search box. Naver describes the search AI as running on a model built on HyperCLOVA X and then tuned on its own service data; the SEED downloads are the small open siblings. A local SEED model feeling weaker than the search answers is expected, not a bad install. Where to next: - Local & Open-Source AI: https://www.launchpadhq.io/discover?category=local-open-source-ai - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai - Research tools: https://www.launchpadhq.io/discover?category=research-academic-tools ### Getting Started with NotebookLM Canonical: https://www.launchpadhq.io/platform/notebooklm/guide Data: https://www.launchpadhq.io/data/v1/guides/notebooklm.json Last reviewed: 2026-07-13 Tool: Gemini Notebook (https://www.launchpadhq.io/platform/notebooklm): starting price Free; free tier yes; data as of 2026-10-20 Upload your own documents and NotebookLM answers questions using only those sources — with citations, and even a podcast-style audio summary. You create a notebook, add your sources — files, links, notes — and then chat with them: ask for a summary, a comparison, an explanation, and it answers from those documents only, with citations you click to see the passage. Add another source and ask again; the answers change with what's in the notebook. What it is: - NotebookLM is Google's source-grounded research assistant. You upload your own material — PDFs, Google Docs, web pages, YouTube videos, notes — and it answers questions using only those sources, citing exactly where each answer came from. Because it's anchored to what you gave it, it rarely makes things up. It also turns your sources into study aids: an Audio Overview (a podcast-style conversation about your material), video overviews, mind maps, flashcards, and summaries. It's for students, researchers, and anyone making sense of a stack of documents. Before you start: - Free with a Google account, no time limit: up to 100 notebooks, 50 sources each, and a daily chat allowance that covers a real study session, plus a few Audio Overviews a day. Most people never need to pay. - Paid tiers (bundled into Google's AI plans, which start at a low monthly price) mainly raise those limits and add heavier features. - One structural thing to know: it's cloud-only (no offline mode), and every source caps at a large but real size limit — a single enormous file may not import. Security: - NotebookLM's defining trait is that it answers only from the sources you upload, with citations — so unlike an open chatbot, it rarely fabricates, and you can trace every answer back to a specific document. That grounding is the whole point. Security: Where your documents stand: - Google states it does not train its models on the documents you upload, and your sources stay private unless you choose to share a notebook. - But your files are processed on Google's servers (it's cloud-only) — so for truly confidential or regulated material, the personal version isn't the right home; that's what the enterprise version exists for. - Each notebook is self-contained — it can't pull from your other notebooks or the open web, which is a privacy feature as much as a limitation. Security: - It's grounded, not infallible: it can still misread a source or miss context. Click the citations to confirm, exactly as you would with any research tool. Best at: - Answering questions from your own documents, with citations - Turning a pile of PDFs into a searchable, summarizable knowledge base - Making study aids — audio overviews, mind maps, flashcards — from your sources - Staying grounded, so it rarely invents facts Okay at: - Very large single files (each source has a size ceiling) - Working across many notebooks at once (each is siloed) - Confidential/regulated material (cloud-only; use the enterprise version) Avoid: - Uploading sensitive files that shouldn't leave your control - Treating the audio/summary as verified without checking citations - Expecting it to answer from outside your uploaded sources Starter: Make a notebook and add sources Create a notebook and upload a few related documents — a PDF, a Google Doc, a web article on one topic. Everything grounds on your sources; this is the step that makes NotebookLM different from a chatbot. Start with material you already need to understand — lecture notes, a report, a contract. Starter: Ask a grounded question Shows the core value — a cited synthesis of your material, not the open web. Follow up by asking it to explain any point "like I'm new to this." Prompt: Summarize the main argument across these sources, and list the points where they disagree. Cite where each point comes from. Starter: Generate an Audio Overview Create the podcast-style Audio Overview and listen to two AI hosts discuss your sources. It's the feature that made NotebookLM famous, and a genuinely different way to absorb dense material. Great for reviewing on a commute — but still check the citations for anything you'll rely on. Pitfalls: - It only knows what you give it. That's the point — it won't answer from the open web, so add the right sources. - Cloud-only, files on Google's servers. Fine for most material; not for truly confidential or regulated files (use the enterprise version). - Grounded isn't infallible. It can still misread a source — click citations before relying on an answer. Where to next: - Research & Academic Tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Document & PDF Processing: https://www.launchpadhq.io/discover?category=document-pdf-processing ### Getting Started with Notion AI Canonical: https://www.launchpadhq.io/platform/notion-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/notion-ai.json Last reviewed: 2026-07-13 Tool: Notion AI (https://www.launchpadhq.io/platform/notion-ai): starting price Free, from $10; free tier yes; data as of 2026-10-26 AI built into the Notion workspace — draft and summarize in your docs, and ask questions across everything you've written. You trigger the AI inside a page — a keyboard shortcut or a menu — and tell it to write, edit, or summarise what's there, then accept, redo, or discard what it produces. Or you ask a question of your whole workspace and get an answer with links back to the pages it drew on. What it is: - Notion AI is the AI woven into Notion, the all-in-one workspace for notes, docs, wikis, and databases. Rather than a separate chatbot, it lives inside your pages: it drafts and rewrites text where you're working, summarizes and translates, autofills databases, and — the standout — "Ask Notion" answers questions using your entire workspace (and connected apps) as the source, with newer agents that carry out multi-step tasks. It's for people who already keep their work in Notion and want AI grounded in their content. Under the hood it uses models from several AI labs, with an auto mode that picks one per task. Before you start: - The big change to know: Notion removed the standalone AI add-on. Full Notion AI (Ask Notion, agents, AI meeting notes) now comes only with the Business plan, priced per user. The Free and Plus plans include just a small trial of AI — not a usable free tier. So while Notion-the-workspace has a great free plan, Notion-AI is effectively a paid feature. - If you don't already live in Notion, the AI's main advantage (grounding in your own workspace) only shows up once you've got content in there. - Try the free trial responses first to see if it fits before committing to Business. Security: - On privacy, Notion is strong: it does not use your workspace content to train AI models, and its AI providers are contractually barred from training on your data — on all plans. Your content stays yours. Security: How it handles your workspace: - Permission-aware: the AI can only see content you already have access to — it respects your existing page and database permissions, which matters a lot on a shared team workspace. - Grounded in your docs: Ask Notion answers from your actual content with links back, so it's far less prone to inventing facts than an open chatbot — but it can still misread, so check the source page. - Data retention varies by plan: Enterprise gets zero retention with the AI providers; other plans have the providers hold data briefly (up to about 30 days) before deletion. Security: - Because it can read across your whole workspace, be mindful of what you put in Notion in the first place — the AI is only as private as the workspace it's grounded in. Best at: - Drafting, summarizing, and rewriting right inside your docs - Answering questions from your own workspace (Ask Notion), with links - Turning plain-English requests into database formulas and autofills - AI grounded in your content rather than the open web Okay at: - Value if you don't already use Notion (the grounding needs your content) - Cost for solo users (full AI now means the Business plan) - Absolute factual reliability (grounded, but still verify against the source page) Avoid: - Expecting real AI on the free/Plus tiers (it's a small trial now) - Assuming answers are infallible (check the linked source) - Putting content in Notion you wouldn't want the workspace AI able to read Starter: Draft inside a doc Writing where you already work — no copy-paste from a separate chatbot — is Notion AI's most immediate everyday win. Then highlight the draft and ask it to "make this more concise" or "change the tone to more formal." Prompt: Write a short, friendly project kickoff note for a team of five, covering the goal, the timeline, and who to contact with questions. Starter: Summarize a long page Turning a wall of meeting notes into a usable summary is the task people reach for most. Works on any long doc — paste in notes first, then summarize. Prompt: Summarize this page into five bullet points and pull out any action items with who owns them. Starter: Ask your workspace a question Open Ask Notion and ask something only your own notes would know, like "what did we decide about the launch date?" This is the feature that makes Notion AI different — a grounded answer from your content, with links to where it came from. It only knows what's in your workspace, so it works better the more you keep there. Pitfalls: - Real Notion AI now means the Business plan. The standalone add-on is gone; Free/Plus give only a ~20-response trial. - It's grounded, not infallible. Ask Notion links its sources — click through and confirm before relying on an answer. - Its value scales with your content. The workspace-grounding advantage only appears once your work actually lives in Notion. Where to next: - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai - AI Plugins for Business Software: https://www.launchpadhq.io/discover?category=ai-plugins-business-software ### Getting Started with Ollama Canonical: https://www.launchpadhq.io/platform/ollama/guide Data: https://www.launchpadhq.io/data/v1/guides/ollama.json Last reviewed: 2026-06-25 Tool: Ollama (https://www.launchpadhq.io/platform/ollama): starting price Free; free tier yes; data as of not dated Run AI models privately on your own computer — free, offline, and nothing you type leaves your machine. You install the Ollama app, pick an AI model to download onto your own computer, and chat with it in a window much like ChatGPT — except everything runs locally, with nothing sent online. What it is: - Ollama lets you run AI models directly on your own machine. Nothing you type leaves your computer, there's no monthly bill, and it works offline. It's the most popular on-ramp to 'local AI,' and the opposite trade-off from ChatGPT: you give up some quality and do a little setup, and in return get complete privacy and unlimited free use. It used to be a command-line-only tool, but there's now a proper app — so it's friendlier than its reputation. The real question isn't whether you're technical; it's whether your computer is powerful enough. Before you start: - The local app is free and open-source (the code is public and free to use) — no account, no card, no usage limits. Running models on your own computer is genuinely unlimited and private, forever. - One thing that'll confuse you the moment you Google it: Ollama now also sells a paid Cloud option for running big models you can't fit on your own machine. That's a separate, optional service — it requires an account and sends your data to their servers. This guide is about the free, private, local path. If you run into 'Ollama Cloud,' that's the paid thing, and you can ignore it to get started. - Your computer does the work, so hardware matters most. Rough floor: about 8 GB of RAM (your computer's short-term working memory — on Windows, check Task Manager; on Mac, 'About This Mac') to run a small-to-mid model smoothly. Underpowered means slow or won't run at all. This — not any setup step — is the main thing deciding whether Ollama feels pleasant or painful. - Friendlier than it used to be. There's now a proper desktop app: install it, open it, pick a model from a menu, and chat — no typing commands required. A command line (a text window where you type instructions instead of clicking) is still handy for housekeeping like listing or removing models, but you don't need it to get going. Getting set up safely (Download only from the official site — ollama.com. This is the one that matters most: don't grab it from a random search result, a bundle site, or a link someone sent you — lookalike sites can bundle malware, and you're installing software on your own machine.): - Check your hardware before you install, not after. Confirm roughly 8 GB of RAM free before committing. - Know where the models live. Leave the install location at its default. The thing to watch: downloaded models are several gigabytes each and pile up on your hard drive (a model's size is often shown in 'parameters' — a rough proxy for how big and capable it is; more is smarter but heavier). - Confirm it works. Running one small model and getting a reply back is your 'set up correctly' checkpoint. Security: - Here the usual warning flips. Because the local app runs entirely on your own computer, it's the most private option there is — the 'never paste anything sensitive' rule for cloud chatbots mostly doesn't apply, because nothing you type is sent anywhere. That privacy is real, and it's the whole point of using it. Security: Three things keep it that way:: - Stay local. Leave the defaults. The one way to undo the privacy is exposing it to your network or the internet (see the pitfalls below). - Get models from trusted sources. Stick to the official model library — same instinct as only installing software from sources you trust. - Know that 'Ollama Cloud' is the exception. The paid cloud option DOES send your data to Ollama's servers — that's the trade-off for borrowing their hardware. Ollama is private-by-default only when you run locally. If privacy is why you're here, stay on the local path. Best at: - Privacy — prompts and files never leave your machine, which is right for genuinely sensitive material - Free, unlimited use - Working offline - Learning how AI actually works by tinkering Okay at: - Raw quality — local models are smaller than the giant cloud ones, so they're useful, not magic - Speed, which is entirely down to your hardware Avoid: - Expecting frontier-level answers — if you want the best and don't care about privacy, cloud chatbots win - Running big models on weak hardware — the fastest route to a bad first impression - Reaching for it when you just want to chat and aren't curious about the local angle — start with ChatGPT, come back when privacy or cost is a real reason Starter: On a modest computer (~8 GB RAM): a small general model Pulls a small general-purpose model — around 3–4 billion parameters, like a small Gemma or Llama. It loads fast, runs smoothly, and proves your setup works end to end. On 8 GB this is the line between 'works well' and 'too slow to enjoy.' It's also the honest starting point — a tiny model isn't local AI at its best, so don't judge the whole approach by it. First run of any model is slow — it's loading into memory, not broken. After that it's quick. Starter: With more room (16 GB+ RAM): a 7–8B general model Steps up to a 7–8 billion-parameter general model — the point where local AI starts feeling genuinely useful rather than just a demo. This is the tier where most people decide local AI is worth it. It needs the extra memory headroom, which is why it's gated on 16 GB+ rather than being the default first pull. Starter: For coding: a code-specific model Pulls a model tuned for programming — a free, private coding helper running entirely on your machine with no per-token cost. A private coding assistant with no metered API bill is one of the most common reasons people install Ollama at all. Worth trying once your hardware can handle a mid-size model. Starter: Use it for things you'd never paste into the cloud A venting draft, a touchy work email, anything personal — that's the text most people hesitate to put into a cloud chatbot. Running locally, none of it leaves your computer, so the privacy you traded some quality for finally earns its keep. Swap in any sensitive editing job — a difficult message, a rough complaint, a journal entry you'd rather keep private. Prompt: Help me rewrite this so it sounds calm and clear instead of angry: [paste your message] Starter: Learn something with no internet It works on a plane, off-grid, anywhere — and you can run it as much as you want for free. Learning by tinkering is one of the real reasons to keep a small local model around, even when a cloud model would answer a little better. Tell it your level — 'explain like I'm completely new to this' — and it'll calibrate. Prompt: Explain [a topic I'm learning] in plain language, then ask me three questions to check I understood. Starter: Build a tiny thing — if your hardware can handle it Honestly gated: this needs a code-specific model and ideally 16 GB+. Within that limit, self-contained jobs — one file, a short script, 'find the bug in this snippet' — are where a small local code model genuinely delivers; what it can't do is build a whole app across many files. Expecting that is the fastest way to wrongly conclude local AI is useless. Prompt: Write a single self-contained HTML file for a simple kitchen timer — all the HTML, CSS, and JavaScript in one file, with no external libraries. Pitfalls: - Biting off too big a model — the #1 mistake. A model your RAM can't handle makes you wrongly conclude 'local AI is bad.' Start small. - Expecting ChatGPT quality — it's impressive for what it is; calibrate your expectations to the model's size. - Models quietly eating disk — they're gigabytes each. Check what you've got with ollama list and remove ones you don't need with ollama rm. - Accidentally opening it to the network — it defaults to your machine only, which is safe. If a tutorial tells you to expose it to '0.0.0.0' or your network, understand you may be letting other people reach your model — don't, unless you know exactly why. Where to next: - Local & Open-Source AI: https://www.launchpadhq.io/discover?category=local-open-source-ai - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Open WebUI Canonical: https://www.launchpadhq.io/platform/open-webui/guide Data: https://www.launchpadhq.io/data/v1/guides/open-webui.json Last reviewed: 2026-09-07 Tool: Open WebUI (https://www.launchpadhq.io/platform/open-webui): starting price Free; free tier yes; data as of not dated The chat interface you run yourself — you bring the model, it brings everything else. You install Open WebUI on your own computer and open it in a browser tab — except the site runs on your machine, not someone else's. That's what self-hosted means. What you type goes to whichever AI you've connected. What it is: - A browser-based chat window for AI you host yourself. It has no model of its own — the model is the AI "brain" that does the thinking — so you connect one separately. - Why this over Jan or LM Studio: those are desktop apps, one person, one computer. This one keeps running in the background, so your household reaches it from their own phones and logins. Before you start: - Free, and the code is public — no account, no trial, no card. The company sells licensing to large organisations; nothing you'd run is paywalled. - Open WebUI is not an AI. Install it and you get an empty model dropdown — sort out the AI half first. Ollama is the usual answer, and this site has a page for it. - The real question isn't whether you're technical, it's whether you want a small server in your house. You'll need Docker (a way of running an app in a self-contained box, so you don't install its parts by hand). If you only want AI on your laptop, Jan or LM Studio do it in an afternoon. Getting set up safely (https://openwebui.com/): - Install only from the official project — the site above, or open-webui/open-webui on GitHub. Lookalikes exist, and this one holds your keys and chat history. - Follow its official quick-start, not a forum post; commands change between versions. Docker is the recommended path, Python and a desktop app the alternatives. - Open WebUI currently finds a local Ollama by itself in the standard Docker setup; an online provider needs its address and key in Settings. - Make the admin account yourself, immediately — the first account on a fresh install becomes the administrator and sign-up closes behind it. Then open the address the docs give you (currently port 3000 for Docker, 8080 for Python and desktop) and send one message; an empty model dropdown means no AI is connected. Security: - Its privacy posture isn't fixed — your model source decides it. Security: What your setup actually determines: - A model on your own hardware: conversations never leave the machine. It works with the internet unplugged. - An online provider connected with an API key — a password that identifies your app and spends your money, whoever on your instance uses it — sends those chats to them, under their terms. - Check which model is selected before pasting anything sensitive: the dropdown is the privacy boundary. Security: - Securing it is your job: keep it on your home network unless you have a reason not to. An instance left open is a stranger's free chat window on your hardware. Best at: - One shared chat window for a household or small team - Local and online models in one place, switched from a dropdown Okay at: - One person on one computer — more machinery than you need - Document chat and web search — supported, and fiddly to get right Avoid: - Access from anywhere, without the job of securing a server - Any expectation of intelligence on its own — it ships with none Starter: Connect exactly one model source, then stop Points Open WebUI at a single model source — a local Ollama it finds by itself, or one online provider's address and key in Settings — and sends one message to confirm the dropdown fills. The settings surface is large, and the temptation is to configure everything before it works. Later problems are easier to diagnose from one known-good path. Starter: Lock the front door Keeps the instance reachable only on your home network, makes the admin account yourself, and checks that sign-up closed behind it. Unlike a desktop app, this is a server the moment it starts — the only tool in this category where "who else can reach this?" is a question you answer yourself. Starter: Ask the same question of two models One question against a model on your machine and one online, in the same window, shows what this is for, and where a small local model's limits are. Prompt: Explain in plain language why an AI on my own laptop might answer more slowly than one on a company's servers. Under 200 words. Pitfalls: - Assuming self-hosted means private. It means private if your model is local. Add an online provider's key and those chats go to them like any cloud tool. - Fighting your Python version. That install route currently supports a specific range and refuses newer releases outright. Docker sidesteps it. - Installing it and never updating it. The project currently ships releases every few weeks, and a self-hosted app doesn't update itself. Where to next: - Local & Open-Source AI: https://www.launchpadhq.io/discover?category=local-open-source-ai - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with OpenAI Codex CLI Canonical: https://www.launchpadhq.io/platform/openai-codex-cli/guide Data: https://www.launchpadhq.io/data/v1/guides/openai-codex-cli.json Last reviewed: 2026-09-08 Tool: OpenAI Codex CLI (https://www.launchpadhq.io/platform/openai-codex-cli): starting price Free; free tier yes; data as of not dated OpenAI's coding assistant for the command line: your machine, one project folder, paid for by your ChatGPT sign-in. You open a project folder on your computer and start Codex in a terminal (a text window where you type instructions instead of clicking). Describe the change you want in plain English. It reads the files, shows you the edits and commands it proposes, and waits for your approval. You read the result and say what to fix. What it is: - Codex CLI is OpenAI's coding agent — AI that takes actions on its own rather than only answering — running on your own machine: it opens your real files and runs commands rather than handing you code to paste. - It is one front door onto one product: Codex also lives in the ChatGPT app, a VS Code extension and a desktop app. The CLI is the one you drive where the code sits. - The name is recycled, so older search results mislead: the 2021 "Codex" was a model OpenAI sold through its API and retired in March 2023. Today's is a different thing, a coding agent from 2025. Before you start: - The entry requirement is code already on your disk and comfort opening a command line. Never used one? A browser-based coding assistant is the better first step; this will still be here. - You do not buy Codex; you sign in with a ChatGPT account and your plan pays for it. OpenAI's pricing page lists it on every plan including the free one, while its own CLI documentation still says Plus or above — so expect to need a paid plan (Go at $8/month, or Plus at $20/month). Billing your own OpenAI developer account per use is the other route. - Usage runs as rolling five-hour and weekly windows rather than a balance you can watch, and Plus and Pro can buy extra credits instead of upgrading. Getting set up safely (OpenAI's official Codex CLI documentation): - It installs in one command: use OpenAI's installer, the official Homebrew cask or npm package, or the GitHub releases page — lookalikes exist. On Windows the fence that keeps Codex inside your folder behaves most predictably under WSL2. - Open it in a project that is saved and tracked, with nothing uncommitted — that tracking is how you read back and undo every line it changed. - /permissions sets what it may do: read-only, a workspace default that edits inside the folder you opened and asks first about anything else, or full access, which removes both the fence and the questions — one keystroke, and it stays granted. Security: - Your code goes to OpenAI's servers to be worked on — the mechanism, not a leak, but the line to check before a client's private project. Security: The setting most people miss: - On personal ChatGPT plans OpenAI may train on your Codex content, and its own data-controls FAQ says the opt-out in ChatGPT's settings does not affect Codex's. You turn that off inside Codex. - Everything it reads is input, including text other people wrote: a README can be phrased to steer the agent rather than inform you. The approval prompts are the check on that. Best at: - A change spanning several files, watched and approved step by step - Getting oriented in a project whose build and test commands nobody wrote down - Chores you can describe precisely: tidying, dependency bumps, filling in missing tests Okay at: - Predicting your spend, since usage runs as rolling windows rather than a visible balance - Unattended runs: codex exec works inside a script, but nobody is approving anything Avoid: - Depending on a particular Codex model staying put — OpenAI shut off its gpt-5-codex and gpt-5.1-codex variants in July 2026 - Code you are not permitted to send off your machine Starter: Start read-only and make it explain the project back to you Codex picks its opening permissions by checking whether the folder is under version control, so your first session may already be allowed to write. Asking a question with /permissions on read-only shows which preset you landed in. Prompt: Before changing anything: list the commands this project uses to build, test and run, and name the files you would touch to add one new setting. Do not edit anything. Starter: Have it write the project's AGENTS.md, then fix the file by hand Codex reads AGENTS.md before doing any work, merging a global one in ~/.codex with the project's. It is a shared convention with a published spec, not a Codex-only file, so rules you write once travel to other agents. It writes a draft — add your exact test command and anything it must never touch. Prompt: /init Starter: Give it a build task while the internet is still switched off Codex's fence has network access turned off by default, so this matches what it can actually do: a package it cannot download becomes a question, not an install you find later. Prompt: Add [the feature] using only what is already in this project — do not add a dependency. If you need something that is not here, stop and tell me what and why. Pitfalls: - Following instructions written for a different Codex: the CLI, the ChatGPT app and the editor extension are three front doors onto one product, and anything from before 2025 describes the retired model. - Turning off model training in ChatGPT and believing Codex is covered. It keeps its own setting. Where to next: - Other coding agents and assistants: https://www.launchpadhq.io/discover?category=ai-coding-development - Developer APIs and services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - Run models on your own machine: https://www.launchpadhq.io/discover?category=local-open-source-ai ### Getting Started with OpenEvidence Canonical: https://www.launchpadhq.io/platform/openevidence/guide Data: https://www.launchpadhq.io/data/v1/guides/openevidence.json Last reviewed: 2026-10-24 Tool: OpenEvidence (https://www.launchpadhq.io/platform/openevidence): starting price Free; free tier yes; data as of 2026-10-25 A free AI reference for verified US healthcare professionals: ask a clinical question and get an answer that cites the medical literature. Not built for patients. You ask a clinical question in plain language — on the web or in the app. OpenEvidence searches peer-reviewed medical literature and guidelines and writes an answer with citations you can open and check. You pick how deep it goes: a quick answer in seconds, or a longer structured report that can take five minutes or more. What it is: - OpenEvidence is an AI tool for clinicians. Its terms say it is for physicians and other healthcare professionals and is not intended for use by patients or other consumers. Its user guide describes evidence-based answers to clinical questions, grounded in peer-reviewed literature, where every answer cites its sources. Typical uses it lists include quick consults, differential diagnosis, treatment options, drug information and guideline summaries. - It has content agreements with major publishers: OpenEvidence's about page names NEJM, JAMA, Nature, NCCN and Cochrane. Its user guide describes three answer models — Osler, the fast default; Sackett, more thorough and may ask clarifying questions; and Snow, a comprehensive structured report (this replaced the older DeepConsult feature). An EvidenceGrade on each answer rates the strength of the evidence behind it. - There's more for clinicians who sign up, including Visits, which records patient encounters and drafts clinical notes, and Patient Take-Homes, plain-language handouts a clinician sends to a patient. OpenEvidence's user guide says patients who receive a take-home can read it but cannot ask their own questions. If you use Visits, its terms say you must tell others in the room you're recording where the law or your organisation requires it, and let them opt out. - The model choice matters more than it first appears. OpenEvidence's guide describes Osler as giving direct answers in seconds, Sackett as taking seconds to a minute and sometimes asking you to clarify, and Snow as producing a full structured report that takes five minutes or more. Matching the model to the question saves you waiting for depth you don't need. Before you start: - It's free for clinicians — and ad-supported. OpenEvidence's advertising policy says it is committed to providing the service free to healthcare professionals, paid for by a limited number of clearly marked ads, and that advertisers cannot influence answers. Its user guide says all three answer models are available to verified clinicians with unlimited usage at no cost. - You'll be asked to prove you're a professional. Its terms say registration asks for your name, email, profession, specialty and NPI number (or another licensing number), which OpenEvidence may try to verify. Its privacy policy says the service isn't intended for anyone under 18. - Not available everywhere. OpenEvidence's terms say the service is not available in the EU, UK or EEA. - If you're not a clinician, this isn't for you. OpenEvidence says so in its terms. Patients looking for health information should talk to their own clinician; the general chat assistants on Launchpad HQ are not a substitute for medical advice either. Security: - Do not enter patient information. Leave out names, dates of birth, record numbers, addresses, photos and any detail that could identify a patient. OpenEvidence's security page says HIPAA covered entities may choose to send protected health information under its standard Business Associate Agreement, but its terms make you solely responsible for consents and for de-identifying anything you enter. Unless your organisation has approved OpenEvidence and confirmed that agreement covers you, ask about a de-identified case instead — "a 67-year-old with…" works just as well. Security: What OpenEvidence says about your data:: - Its privacy policy (updated September 17, 2026) says your raw queries are confidential, that it does not train AI models on protected health information, and that it secures protected health information in compliance with HIPAA. - Its terms say it may use de-identified data for research and development, and its privacy policy describes ads using cookies and pixels — but not ads based on your queries, uploads, chats or health information. - It states it is SOC 2 Type II certified and encrypts data at rest. Security: - OpenEvidence's terms say it is for educational and informational purposes, is not a diagnostic service, and does not replace the clinical judgment of a qualified professional. Its acceptable use policy requires a professional to review content before it's shared with a patient, and to follow your institution's AI policies. Best at: - Fast, cited answers to focused clinical questions, with sources you can open. - Summarising guidelines and comparing treatment options in a structured format. - Free access for verified US clinicians, with no usage cap on its answer models. Okay at: - Being right every time. Its terms provide the service as-is with no accuracy warranty — check the citations. - Long, complex questions. Snow is thorough but can take five minutes or more. Avoid: - Entering identifiable patient information outside an approved organisational agreement. - Using it as a patient or consumer — it isn't designed for that. Starter: Be specific OpenEvidence's prompting guide contrasts a vague question like "management of T1DM" with one that includes the key labs. Detail gets a more useful answer — and notice there's no patient identifier in it. Prompt: Management of new-onset type 1 diabetes in an adult with an HbA1c of 11% and no ketoacidosis: initial insulin regimen and first-week follow-up. Starter: Ask for a format Its prompting guide recommends asking for an output format, such as a table. A clinical curbside on macrocytic anemia is one of its own examples. Prompt: Differential diagnosis for macrocytic anemia in adults, as a table ordered from most to least common, with the first-line workup for each. Starter: Compare options with constraints This is close to an official example of stating the outcome you want. Constraints like allergies narrow the answer to what's actually usable. Prompt: Alternative treatments for acute otitis media in a child with penicillin, cephalosporin and macrolide allergies. Starter: Check the evidence grade OpenEvidence uses this example to show EvidenceGrade. Look at the grade and open two citations — it's the habit that makes the tool safe to rely on. Prompt: Risk of Lyme disease after a tick has been attached for 5 hours. Pitfalls: - Treating an answer as the final word. Open the cited sources, especially before acting on a dose or a rare diagnosis. - Choosing the slowest model for quick questions. Its guide says to start with Osler and move to Sackett or Snow only for harder questions. - Pasting notes straight from a chart. Strip identifiers first, every time. - Using it outside the US without checking. The terms exclude the EU, UK and EEA. - Skipping the ad labels. The free service shows marked advertisements; OpenEvidence says advertisers can't influence answers, but treat sponsored content as sponsored. Where to next: - More healthcare AI: https://www.launchpadhq.io/discover?category=healthcare-ai - Research & academic tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Meeting & note tools: https://www.launchpadhq.io/discover?category=meetings-notes ### Getting Started with OpenRouter Canonical: https://www.launchpadhq.io/platform/openrouter/guide Data: https://www.launchpadhq.io/data/v1/guides/openrouter.json Last reviewed: 2026-10-23 Tool: OpenRouter (https://www.launchpadhq.io/platform/openrouter): starting price No fee up to $25,000 of list-price inference per month, then 5%; free tier yes; data as of 2026-10-14 One account and one API key for hundreds of AI models from dozens of companies — with a browser chat room to compare them before you write any code. You make an account, add credits, and create an API key (a secret password your code sends with each request). Your app sends a request to OpenRouter naming the model you want; OpenRouter passes it to a company that runs that model and returns the answer. You can also chat with models in the browser, no code needed. What it is: - OpenRouter is a middleman between you and the companies that run AI models. Instead of opening separate accounts with OpenAI, Anthropic, Google and the rest, you open one account, keep one balance, and reach all of them through a single connection. Its pricing page lists more than 500 models from more than 80 providers on paid accounts. - It is built for people writing code. OpenRouter's API is OpenAI-compatible, meaning software already written for OpenAI's API can usually switch over by changing the web address it talks to and the model name. Its docs say it can also fall back automatically: if the model or provider you asked for errors, it tries the next one on a list you supply. - There is a no-code side too. The Chat Playground at openrouter.ai/chat lets you send one prompt to several models side by side and compare the answers, which is the quickest way to see whether a cheaper model is good enough for your job. OpenRouter notes that chat-room conversations are stored only in your browser, not synced to its servers. - OpenRouter does not make its own models. Every answer comes from a third-party provider, and the same model can be offered by several providers at different prices and speeds; by default OpenRouter balances between them, favouring cheaper ones while weighing uptime. Before you start: - You can start free, within limits. OpenRouter's free plan gives access to a set of free models (their IDs end in :free). Its rate-limit docs set a per-minute and a daily cap on free models, and the daily cap rises sharply once you have bought a small amount of credits in total. - Paid models run on prepaid credits. OpenRouter says it passes through each provider's price with no markup on the model itself, and instead charges a platform fee when you buy credits (on the standard pay-as-you-go plan: 5.5% fee on credit purchases). Its terms set a small minimum purchase. Prices are per million tokens (chunks of text, roughly three-quarters of a word), usually different for what you send and what comes back. - Unused credits are not a savings account. OpenRouter's terms let you refund unused credits only within 24 hours of buying them (the fee is not refunded), and say it may expire credits that go unused for a year. Buy small amounts while you learn. - You'll need a little code to use the API. If you only want to chat with AI, a regular chat app is simpler; OpenRouter earns its keep once you are building something or comparing models. Getting set up safely (Create keys only on openrouter.ai itself (Settings → Keys), and store each one in your project's environment variables — OpenRouter's docs say never to commit keys to a code repository — rather than in a chat, email or document.): - Put a credit limit on every key. When you create a key you can give it a spending cap that resets daily, weekly or monthly. OpenRouter's own guidance is to always set one, because a key with no limit can drain your whole balance — including any auto top-up you have switched on. - Decide on privacy before your first real prompt. In Settings → Privacy you can stop OpenRouter routing your requests to providers that may store or train on your data, and restrict routing to providers with zero data retention (they keep nothing). Strict settings can make some models unavailable; that error is the setting working. - Confirm it works with a free model. In the Chat Playground, add the Free Models Router and send a short message. A reply, with the name of the model that answered, means your account is set up. Security: - Your API key is the thing to protect. Anyone holding it can spend your credits, so keep it out of code you share, screenshots and public repositories. OpenRouter's docs say it scans GitHub for leaked keys and emails you if it finds one; if that happens, delete the key in Settings → Keys and make a new one. Security: What happens to your prompts:: - OpenRouter's data-collection docs say prompt and response logging is off by default and always opt-in. It does keep metadata such as token counts and timing. - There is an opt-in that lets OpenRouter use your inputs and outputs in exchange for a 1% discount. Leave it off for anything private. - The providers that actually run the models have their own policies, and some may keep or train on what you send. OpenRouter lets you exclude those providers in your privacy settings — separately for free and paid models. Its pricing table lists data-policy routing as not included on the Free plan, so check what your plan allows. Security: - Don't send personal, client or confidential information until you have set those privacy controls and checked which providers your chosen model routes to. Best at: - Trying many models from one account and one bill instead of signing up with each company. - Comparing models side by side in the Chat Playground before you commit code to one. - Keeping an app running when one provider has an outage, using fallback lists. Okay at: - Free experimentation. The free models are real, but OpenRouter itself calls them usually unsuitable for production use, and the free daily cap goes quickly. - Predicting costs. Prices are pass-through and can change; OpenRouter says requests keep working at a model's new price, so check the model page before a big job. Avoid: - Creating a key with no credit limit. - Buying a large block of credits up front — refunds are only possible within 24 hours. - Sending sensitive data before you have set your privacy routing. Starter: Chat for free in the browser Open openrouter.ai/chat, click Add Model, type "free" and choose the Free Models Router, then send a message. Each reply shows which free model answered. It proves your account works and shows you what OpenRouter does — many models behind one door — without a key, code or payment. Ask the same question twice and notice that a different model may answer. Starter: Race two models on one prompt In the Chat Playground, add a well-known paid model and a cheaper or free one, then send them the same prompt from your own work and compare the answers. Most tasks don't need the most expensive model. Seeing the gap yourself is the fastest way to pick a model you can afford to run all day. Check each model's price per million tokens on openrouter.ai/models before you choose. Starter: Make a key with a ceiling Go to Settings → Keys, create a key with a name you'll recognise, and set a small credit limit that resets monthly. The limit is what turns a leaked key or a runaway script from a disaster into a nuisance. Doing it before you write code means you never run without one. Starter: Point existing code at OpenRouter If you have code written for OpenAI's API, change its base URL to https://openrouter.ai/api/v1, use your OpenRouter key, and set the model to an ID copied from the models page. Because the API is OpenAI-compatible, this is usually the whole migration. It is the clearest demonstration of why developers use OpenRouter. OpenRouter's Request Builder generates a working example request in several languages if you are starting from scratch. Starter: Stay free on purpose When testing in code, use only model IDs ending in :free, or the openrouter/free router, and keep paid models out of any fallback list. OpenRouter's docs warn that if a free model's fallback list includes a paid model, the paid one is billed. This keeps experiments at zero cost. Pitfalls: - Web search is never free. Adding web search to a request (the :online suffix or the web plugin) costs extra even on free models. - A negative balance blocks free models too. OpenRouter's docs say you need to top up before free models work again. - Models come and go. A retired model returns a "no endpoints found" error; pin exact model IDs and recheck them occasionally. - The same model can cost different amounts. Prices vary by provider, and default routing balances on price and uptime rather than always choosing the cheapest. - Deleting your account forfeits your credits. Spend down or refund (within 24 hours) first. Where to next: - AI APIs & Developer Services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - AI coding tools to build with: https://www.launchpadhq.io/discover?category=ai-coding-development - Run models on your own computer instead: https://www.launchpadhq.io/discover?category=local-open-source-ai ### Getting Started with OpusClip Canonical: https://www.launchpadhq.io/platform/opus-clip/guide Data: https://www.launchpadhq.io/data/v1/guides/opus-clip.json Last reviewed: 2026-09-12 Tool: Opus Clip (https://www.launchpadhq.io/platform/opus-clip): starting price Free, from $15/mo; free tier yes; data as of not dated Turn one long video into a batch of short clips for TikTok, Reels and Shorts — without opening an editor. You paste a link to a long video or upload the file, then wait while it watches the whole thing. It hands back a set of short vertical clips, each one already cut, captioned and ranked. You play them, keep the ones you like, adjust the wording or the timing, and download them one at a time. What it is: - OpusClip takes one long video — a podcast episode, a livestream, a conference talk, a webinar — and cuts it into short vertical clips for TikTok, Reels and Shorts. It decides which moments are worth clipping, so your job becomes reviewing its picks instead of scrubbing through an hour of footage. - It is built for people who already have long footage and no time to edit it. Making a video from scratch is a different job and a different product — the same company sells one called Agent Opus, which this page does not cover. - What comes back is a first pass, not a finished post. Before you start: - The free plan currently gives you a monthly allowance of processing minutes, refreshed monthly. What counts is the video you feed in, not the clips that come out: OpusClip's help centre charges by the minute of original video imported, and the pricing page calls the same thing credits (the platform's unit of spend — each thing you make costs some). A single long podcast can eat most of a free month in one upload, so the allowance, not the price, is the constraint. - You need a long video you already have and are allowed to use. On the free plan the help centre lists your sources as YouTube and files from your own computer; Google Drive, Vimeo, Zoom, Rumble and StreamYard are listed as paid-plan additions, as are posting straight to social and scheduling. - Check your language before anything else. OpusClip says it supports "over 25 languages," but all except English, German, Spanish, French and Portuguese are marked as still being tested, with "inaccurate emojis or highlighted keywords" as the stated symptom. Two gaps are absolute rather than rough: the help centre states plainly that Hindi is not supported, and that "we do not support any languages that read right to left, including Hebrew, Kurdish, Arabic or Persian." - A free export carries an OpusClip logo watermark (a mark stamped on the output showing where it was made) and comes out at up to 1080p — normal HD, fine for phone screens. It is not removable on the free plan, though the help centre notes that upgrading strips it retroactively from clips you already made. The 7-day Pro trial also watermarks: the vendor lists "Videos exported during the trial will include the OpusClip watermark" first among what the trial does not include. - Whether you will need to pay turns on more than the watermark. The terms of service (effective 17 November 2025) say you will use the Services "for your own internal, personal, non-commercial use," and license commercial use only "if you are a user of Paid Services" — so clips for a business, a client or a channel you earn from sit on the paid side. There are several paid tiers, each cheaper paid annually; current prices are on its Launchpad page. Security: - What matters here is the size of what you hand over. To find a 40-second moment, OpusClip takes the whole file — so everyone who appears or speaks anywhere in that recording is uploaded, not just the part you wanted. Worth a thought before you feed it a client call or a private workshop. Security: What the vendor's own documents do and don't say: - The privacy policy (effective 17 November 2025) says data is kept "for as long as necessary to provide you with our Services," and gives no retention period for the video files themselves. - It does not say, either way, whether uploaded video is used to train OpusClip's AI. It lists "testing, research, internal analytics and product development" as a purpose, which is not the same as an answer. - The one explicit opt-out is regional: EU/EEA users are told "you are free to opt out of any use of your personal data for our AI research and model-development activities," by email with the subject line "AI R&D – Opt-Out." No equivalent is described for anyone else. Security: - One mismatch is worth naming rather than smoothing over. OpusClip's marketing and help pages describe the free plan as an ordinary creator product and never mention the non-commercial limit that sits in the terms of service. Where a binding document and a friendly one disagree, the binding one governs. Best at: - Turning one long talking-head recording — podcast, webinar, livestream, interview — into a batch of short clips in one pass - Adding animated captions automatically, the part that takes longest by hand - Recropping a wide recording into a tall, phone-shaped one, keeping the speaker in frame Okay at: - Deciding what is actually interesting. The picks are plausible rather than curated; plan on keeping two or three out of ten. - Precise trims. This is a clipping tool with editing attached, not an editor. - Short source videos. It hunts highlights inside long footage, so a three-minute upload gives it little to choose between. Avoid: - Clipping and posting footage you did not make. OpusClip's terms make you warrant you have "all rights necessary" to what you upload and that it infringes nobody's copyright. The tool will process it; that is not permission. - Anything where a mis-heard word would matter. Captions are generated from speech and baked into the picture, so a mangled name or figure means re-exporting and re-uploading, not a quiet edit. Starter: Run one stretch first, with the box empty Upload a video you already have, type nothing into the prompt box, and let OpusClip pick on its own. OpusClip's help centre describes a Processing timeframe slider as the way to cut what an upload costs you — you point it at one section of a long file rather than the whole thing, which makes a first test cheap. It names the slider but does not say which plans have it, so check yours once you are signed in. Pick something with one person talking clearly. Starter: Steer it with words that are actually spoken The free plan clips by speech alone, and OpusClip's help centre is specific about what that means: keywords work when "the keywords are present in the speech or transcript." On its own account, typing funny returns nothing useful unless somebody says the word "funny." Listen for the phrases you already repeat. You are searching a transcript, not describing a mood. Prompt: biggest mistake, what I'd do differently, nobody tells you, the one thing Starter: Describe the moment instead of the words OpusClip's prompts manual says a strong prompt names a subject, a visible action, an emotional tone and a sound — this one names all four, the shape of its own worked example ("Clip my exciting reactions to tasting shawarma"). The same manual lists non-English text among the things that make a prompt (the message you type) fail, so write it in English even when the video is in another supported language. This free-text style of clipping is the model OpusClip calls ClipAnything, and its pricing comparison puts it above the free plan — the free tier reads speech only. Two reads of that comparison table disagreed about whether it starts at the entry paid tier or the one above it, so check before you buy on the strength of it. The 7-day trial is the way to try it without paying. Prompt: Find the moments where I answer a question from the audience, and include my reaction right afterwards. Keep the laugh if there is one. Starter: Ask again, and ask for a compilation OpusClip's help centre says the prompt box reappears at the top of the results page and that "you can reprompt as many times as you like to get more clips." The prompts manual also lists compilations as their own category of request — "Compile all Man City's scoring" is its example — returning one clip assembled from scattered points in the recording. The docs describe that rather than demonstrate it, so treat it as their claim. The help centre does not say whether re-prompting a video you already uploaded spends more of your monthly allowance. Watch your balance the first time, and assume it might until you have seen otherwise. Prompt: Compile every time the guest mentions [their company] into one clip. Pitfalls: - A bad batch still costs you the minutes. There is a refund path, but the help centre limits it to Pro accounts, within 14 days, and only if you "haven't downloaded any clips yet" — free and trial users are told outright they cannot reprocess a video for a credit refund. - Free clips stop being exportable after three days. The pricing page puts free-plan storage at "Expires after 3 days" and the export window at the same limit. Download what you want the day you make it. - The Virality Score is a prediction, not a measurement. OpusClip scores each clip 0 to 99 on four things it names — Hook, Flow, Value and Trend — all judged before a single human has watched it. A 92 is not evidence it will outperform a 71. - The marketing page is not the plan page. The homepage advertises imports from Google Drive, Vimeo, Zoom, Twitch and more; the help centre's own plan breakdown gives the free plan YouTube and local files. Both are OpusClip pages and they do not match — trust the plan comparison table. - Everything you make will look like everything else you make. The free plan includes one brand template and no custom fonts, so your clips carry the same default caption look as every other OpusClip clip unless you restyle each by hand. Where to next: - Video creation and editing: https://www.launchpadhq.io/discover?category=video-creation-editing - Voice and speech: https://www.launchpadhq.io/discover?category=voice-speech - Sales, marketing and SEO AI: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai ### Getting Started with Otter AI Canonical: https://www.launchpadhq.io/platform/otter-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/otter-ai.json Last reviewed: 2026-07-13 Tool: Otter AI (https://www.launchpadhq.io/platform/otter-ai): starting price Free, from $16.99/user/mo billed monthly; free tier yes; data as of 2026-10-25 An AI notetaker that joins your meetings, transcribes them live, and hands you a searchable summary afterward. Otter works by connecting once and then running on its own. You link your calendar and grant access to your meeting tool; from then on its bot shows up as a participant in your scheduled calls, captures the audio, and turns it into a live transcript with speakers labelled. Afterwards you get a summary and a searchable record. What it is: - Otter is a meeting assistant. You connect it to your calendar, and its bot joins your Zoom, Microsoft Teams, or Google Meet calls, transcribes everyone in real time, and produces a summary with action items you can search later. It's aimed at people who sit in a lot of meetings — students in lectures, salespeople on calls, anyone who'd rather pay attention than scramble to take notes. Before you start: - The free Basic plan gives you a limited monthly allowance of transcription minutes, a cap on how long any one conversation can be, and only a handful of file uploads ever (lifetime, not monthly). It's enough to test the quality and the meeting integration, but a couple of long meetings will exhaust it. Pro, a paid monthly plan that is cheaper billed annually, gives you a much larger allowance. - The single most important thing: an AI bot joining a call to record it affects everyone on the call, not just you. Handle consent before you turn it on (more below). - Set up the calendar and meeting-tool connection first — that's the "recipe" that makes Otter run on its own. Security: - Otter's whole model is that a bot joins your meeting and records everyone in it. In many places, recording a conversation requires the other participants' knowledge or consent — and it's simply good practice regardless of the law. Tell people the notetaker is on, or let the visible bot do that for you, before the real conversation starts. Security: Where your meetings go: - Every recording and transcript is uploaded to and stored in Otter's cloud — nothing is processed locally, and it needs an internet connection. - That means don't run it on conversations covered by confidentiality, legal privilege, HR sensitivity, or health information unless you've confirmed it's allowed. (HIPAA-grade handling is an enterprise-only add-on.) - Anyone you share a transcript with can read everything that was said — check before forwarding. Security: - As of early 2026, Google Meet began flagging third-party notetaker bots (Otter among them) as a potential security concern, and organizations on strict security settings can auto-eject the bot. If your workplace runs a locked-down Meet, test that Otter can actually join before you rely on it. Best at: - Capturing lectures, interviews, and recurring meetings hands-free - Turning a long call into a searchable summary with action items - Letting you stay present in a conversation instead of typing notes Okay at: - Accuracy on crosstalk, heavy accents, or noisy audio (good, not perfect) - Non-English meetings (a limited set of languages is supported) - Bulk transcription of uploaded files (the free tier's import limit is tiny) Avoid: - Confidential, privileged, or HR-sensitive conversations without clearance - Recording people who haven't been told - Relying on the summary verbatim for anything high-stakes without checking the transcript Starter: Connect your calendar and meeting tool Link Otter to your calendar and grant access to Zoom, Teams, or Google Meet so its bot can auto-join scheduled calls. This is the setup that makes everything else automatic; without it you're uploading files by hand. Only want it on some meetings? You can control which calls the bot joins. Starter: Turn on a consent habit Decide how you'll let participants know they're being recorded — a note in the invite, a verbal heads-up, or relying on the visible bot — and make it your default. It's the difference between a useful tool and an awkward (or non-compliant) surprise. For external calls, put the disclosure in the meeting invite so it's on record. Starter: Run one real meeting, then read the summary Let Otter capture a low-stakes internal meeting, then open the transcript, try the search, and ask its chat "what were the action items?" One real run shows you the accuracy and whether the summary is trustworthy for your use. Compare the summary against your own memory of the call to calibrate how much to trust it. Pitfalls: - The free tier's caps bite fast. The monthly minutes, the per-conversation limit and the lifetime upload limit all run out quickly — a single all-hands can blow the per-meeting cap. - Minutes don't roll over and there's no pay-as-you-go. Hit the cap and Otter simply stops until the cycle resets or you upgrade. - The bot is visible to everyone. That's a feature for consent, but plan for the reaction on external calls. Where to next: - Meetings & Notes: https://www.launchpadhq.io/discover?category=meetings-notes - Workflow & Automation: https://www.launchpadhq.io/discover?category=workflow-automation ### Getting Started with Outreach Canonical: https://www.launchpadhq.io/platform/outreach/guide Data: https://www.launchpadhq.io/data/v1/guides/outreach.json Last reviewed: 2026-09-15 Tool: Outreach (https://www.launchpadhq.io/platform/outreach): starting price Custom; free tier no; data as of not dated The sales platform your employer buys for you — and how to get something out of its AI once you have a seat. You sign in to a web dashboard that already holds your company's accounts, deals and email history. You type a question in plain English — about a customer, a deal, a call — and it answers from those records, then offers to draft the next email. You edit what it wrote and send it. What it is: - Outreach is the system a sales team runs its day out of — customer records, call recordings, scheduled emails and pipeline in one place — with a layer of agents (AI that takes actions on its own rather than only answering) sitting on top of all of it. - It is not a tool you go and get: Outreach is sold by its sales team to whole revenue teams, so the realistic way you meet it is that your employer already bought it and gave you a seat. This page is written for that moment. - Its AI headline is Outreach Omni, a chat box in the web app, the mobile app and Slack that answers questions about your accounts and deals from your company's own records and drafts follow-ups from them. Alongside it sit Kaia, which records and transcribes your sales calls, and a set of narrower agents for meeting prep, deal risk and email personalisation. - The company rebranded to outreach.ai in April 2026. The old outreach.io address now redirects there, though the support site and its release notes deliberately stayed on the old domain, so a link that still says outreach.io is not necessarily out of date. Before you start: - You cannot buy this by yourself, and there is nothing here you can sign up for tonight. Outreach publishes no prices — every plan on the pricing page ends in "Request pricing" — and its Terms of Service (last updated 12 April 2024) say service begins on the effective date in a signed Order and then auto-renews in successive twelve-month periods. No free trial is advertised anywhere on the site; it is possible one exists inside a sales conversation, but nothing public says so. Outreach also publishes no seat minimum and no floor contract value. If you want to learn how AI inside a sales system actually behaves before an employer hands you one, HubSpot's free tier gives you a real account with customer records and AI drafting in it, and most of what you learn there transfers directly to this. - Billing is two things at once, which matters even when it is not your money: Outreach charges per seat and per use, through credits (the platform's unit of spend — each AI task costs some). Each plan includes a fixed pool, larger on the higher Amplify plans, and a credit is spent on "a specific AI-powered task performed by the platform—such as generating a personalized message or completing an individual search." Outreach does not publish what any given action costs, so how far a pool goes for your team is something you only learn from watching it. Extra credit packs can be bought at any time, so nothing stops mid-task; the cost just moves. - The four plans are not a ladder, and which one you are on decides which AI you actually have. On Outreach's own plan comparison, the Omni chat is on all four, but sequences and the email Personalization Agent start at Amplify Core, call recording and conversation intelligence start at Amplify Essentials, and AI forecasting is Amplify Pro only. - You will also need your admin to have switched the AI on. Omni is an explicit toggle an Outreach Admin sets under Administration > Agents & AI > Features, so "I don't see it" is usually a permissions answer rather than a you answer. Outreach's August 2026 release notes also mention an AI Control Hub, but what an ordinary user can see or change there is not documented publicly, so it may well turn out to be another admin-only screen. - Everything on this page about how Omni behaves comes from Outreach's own documentation and marketing rather than from hands-on use, because the product cannot be reached without a paid seat. Treat the interface detail as Outreach's account of its own software. Security: - The people with the most at stake in this system are not you and not your employer. Outreach holds your customers' contact details and, where your company uses Kaia, recordings and transcripts of calls those customers were on. Outreach's security page lists SOC 2 Type II, ISO 27001, ISO 27701, ISO 42001 (the AI-management standard) and HIPAA, which is a genuinely strong shelf of audits; the full trust documentation behind it is not public, and the page tells you to ask your account executive for Trust Center access. Security: Three things worth knowing about the AI specifically: - The training commitment is narrower than the certificate list suggests. Outreach's privacy statement says "Google Workspace Data is not used to develop, improve, or train AI and/or ML models." That is one named source, and the statement makes no promise about anything else in your instance — a gap in the public documentation rather than a confirmed practice in either direction. - Read the document, not the index. Outreach's legal-documents page lists the privacy statement as last updated 17 February 2026, while the statement itself says last modified 21 August 2026. Where a vendor's index and its binding document disagree, the document governs. - International transfers are covered by paperwork, not by keeping data local. Outreach relies on the EU-US, UK and Swiss Data Privacy Frameworks and on standard contractual clauses rather than promising your data stays in a region. Security: - None of this is a reason to be nervous about using the tool your employer handed you. It is a reason to ask, once, who at your company decided what gets recorded. Best at: - Answering "what is going on with this account" without you building a report first — it reads the emails, calls and fields already logged against it. - Turning a call you just had into a transcript, a summary and the follow-up email, in the same place the deal record lives. - Making a multi-step outreach plan run the same way across a whole team, which is the original job the product was built for and still the thing it is least replaceable at. Okay at: - Calls that are not in English. Kaia transcribes in around 28 languages, but Outreach's own documentation is explicit that "real time content cards, action items, recording search by topic are not supported in additional languages" — so on a French or Japanese call you get words, not the AI layer on top of them. - Forecasting. The AI projection is a model built over your own pipeline data, so it inherits your team's habits rather than correcting them. Avoid: - Firing a sequence at a list you did not earn. Outreach's Acceptable Use Policy — last updated 10 January 2021 and still in force — requires that "you must have permission to email the recipient when using our service to send bulk campaigns," requires an opt-out method and your valid physical mailing address in commercial messages, bans sending from a pseudonym or assumed identity, bans group addresses like hello@ and sales@, and prohibits sending that "generates an unacceptable number of bounces" or spam complaints. Those are contract terms you are personally agreeing to, not etiquette — and that policy predates every AI feature on this page and contains no AI-specific clauses, so how it applies to agent-generated sending is not spelled out. - Trusting a confident summary of an account nobody has been logging into. Omni answers from your company's records; when the records are thin it produces a fluent, specific, wrong answer rather than an error, and there is nothing on screen to tell the two apart. Starter: Find out what you actually have Before you look for any feature, ask your Outreach admin two questions: which Amplify plan the company is on, and whether the Omni agent is switched on for your team. Outreach publishes its admin documentation openly while keeping the product behind a contract, so you can read exactly which switch controls Omni and still have no way to see it from your own seat. Compared with HubSpot, where the tier you are on is printed in your own account settings, the first move here is a conversation rather than a click. Ask the same person how many AI credits the plan includes and who watches the balance. It is a shared pool, and you are about to start spending from it. Starter: Ask where a deal stands Outreach's own examples of Omni questions run to "Summarize where this deal stands and what's at risk." The last sentence is the part Outreach does not suggest, and the one that makes the answer checkable — it separates what is in your CRM (the system your sales team keeps its customer records in) from what the AI filled in around the gaps. Run it on a deal you already know cold. That is how you calibrate how much of the answer to believe on the deals you don't. Prompt: Summarise where this deal stands, what has happened on it in the last two weeks, and what is most at risk. At the end, list anything you inferred rather than read from an actual record. Starter: Make it show its evidence Omni answers out of rows your colleagues logged rather than out of a judgement about the account, so it can name the email, call or field behind each recommendation and date it — which turns a ranked list into something you can spot-check in thirty seconds instead of taking on trust. If it cites nothing older than today, it is reading activity rather than judging accounts. Ask it what it would need to do better. Prompt: Which of my accounts should I prioritise today? For each one, name the specific email, call or field you based that on, and the date. Starter: Draft the follow-up, then take it back Outreach markets "Draft a follow-up to address the pricing objection" as a one-line request. The constraints matter because this draft goes out from your own connected mailbox under your name, and an AI reading a deal record will cheerfully promise a timeline it found in a note. Paste the reply you actually sent back in and ask what it got wrong. Two or three rounds of that is worth more than a better first prompt. Prompt: Draft a follow-up email to [name] addressing the pricing objection from our last call. Under 120 words, no superlatives, plain sentences. Do not invent any commitment, discount or date we have not already agreed. Starter: Send one sequence to yourself first Put your own address into a one-person version of any sequence before it touches a customer, and read every step as it arrives — timing, signature, the unsubscribe line, how the AI-personalised opener actually reads. Outreach sends through the mailbox you connect rather than its own servers, so what lands in your inbox is exactly what a prospect gets, rendered by the same provider. It is the only test that catches a personalisation token resolving to a blank. Do it again from a phone. Most sales email is opened there, and a paragraph that reads fine on a laptop often does not. Pitfalls: - Your sending ceiling is your mailbox's, not Outreach's. Outreach's own documentation puts Google at 2,000 emails per account per day — dropping to 500 if the account is new or flagged — Microsoft 365 business at 10,000 per day and 30 per minute, and Outlook 365 personal at 300 per day. A sequence built for a big list does not warn you before it runs into that wall. - Deliverability damage lands on your company's domain, not on the software. A burst of bounces or spam complaints costs your colleagues their inbox placement too, and it is slow to undo. - Recorded calls have a consent question attached, and it is not answered by the software. Outreach's privacy statement does not set out recording-consent requirements; those come from where you and the other person are sitting, and some places require both sides to agree. Worth asking your admin what your company's policy is rather than assuming the default setting is the legal one. Where to next: - Sales, marketing and SEO tools: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai - Meeting and note-taking AI: https://www.launchpadhq.io/discover?category=meetings-notes - AI inside business software: https://www.launchpadhq.io/discover?category=ai-plugins-business-software ### Getting Started with Pabbly Connect Canonical: https://www.launchpadhq.io/platform/pabbly-connect/guide Data: https://www.launchpadhq.io/data/v1/guides/pabbly-connect.json Last reviewed: 2026-09-17 Tool: Pabbly Connect (https://www.launchpadhq.io/platform/pabbly-connect): starting price $25/mo; free tier yes; data as of not dated Automation you can buy once instead of renting — and what "lifetime" actually covers. You build workflows in your browser. Each one starts with a trigger (the "when" — the event that starts an automation), then you add one or more actions (the "then" — what happens once the trigger fires), picking apps and fields from menus rather than writing code. You run it manually to check the data, then switch it on. What it is: - Pabbly Connect links apps that don't naturally talk to each other — your email, your spreadsheets, your payment tool, your CRM (the system a sales team keeps its customer records in) — so something happening in one makes something happen in another. By Pabbly's own count it connects to around 2,000 apps. - What makes it unusual isn't the automation. It's the price tag. Almost everything in this category is rented monthly; Pabbly also sells Connect as a one-time purchase — pay once, keep the software, with no recurring subscription. - It is built and run from India by a company called Magnet Brains, and it positions itself openly as the cheap alternative to Zapier — its own pricing page runs Zapier's prices alongside its own. Before you start: - Start on the free plan, which is genuinely free and not a trial. You get a small monthly task allowance, no card required, and unlimited separate workflows — but each one is limited to two steps (a trigger plus a single action). Sign up, connect one app you already use, and build something from the list below before you think about money. - A "task" is Pabbly's unit of spend, and it is narrower than you'd guess: only successful actions that touch an outside app count. The trigger is free, and so are Pabbly's own internal steps — filters, routers, formatting — so a workflow that looks like four steps may bill as one. Every plan is priced by that number. - The one-time offer is the real decision. Pabbly currently sells three one-time Connect plans, priced by the monthly task allowance; the cheapest also caps how many workflows you can build, while the higher two allow unlimited multi-step workflows. Prices display in your local currency, so you may see a rupee figure instead. The sales page promises all future integrations and updates at no extra cost. But "lifetime" covers the software, not the help: Pabbly's terms say lifetime plans include one year of support from purchase, after which you renew support by paying one third of the original purchase price. Treat it as a large payment now plus a possible recurring one later. - There is a 30-day money-back guarantee, worded on the sales page as a full refund, no questions asked — which is the real safety net on a four-figure decision. Security: - A one-time purchase changes the security question. A subscription gets a second look every time the renewal invoice lands; something you bought outright has no invoice, so a workflow you set up today can still be running in five years, still holding a live connection into your email and your files. Buying once means you have to schedule the review yourself. Security: Three habits that make that survivable:: - Diary a connection review. Once a year, open the list of apps you've linked and disconnect everything you no longer recognise. Nothing else will prompt you to. - Protect the Pabbly account harder than the apps it connects to. It is the one login that can reach every app you have joined to it, so a strong unique password and two-factor authentication matter more here than on most logins. - Revoke from both ends. Any connection can also be cut from the connected app's own security settings, which is the faster route if you ever lose access to Pabbly itself. Security: - On data: Pabbly's privacy policy, last updated 18 February 2026, says data is stored with Amazon Web Services in the United States and Hetzner in Finland, and that Connect automatically removes workflow history older than 15 days. That short window helps privacy and hurts debugging. The policy states compliance with European privacy law, but Pabbly is an Indian company, so if where your data is handled matters to you contractually, check that before you commit money. Best at: - Replacing a monthly automation subscription when your volume is steady and predictable — that swap is the whole pitch, and the one-time plans are priced to win it. - Straightforward "when this happens, do that" chains between common business apps: forms, spreadsheets, email, payments, CRM. Okay at: - Elaborate branching logic. Routers and filters exist, but the drag-and-wire visual canvas that Make and n8n are built around is not the model here. - Newer or niche apps. The catalogue is large, and smaller than Zapier's. Avoid: - Buying a one-time plan expecting to stack it, resell it or move it. Pabbly's terms state you can't buy multiple one-time plans or transfer one between accounts — it is tied to the account that bought it. - Assuming the largest one-time plan is safe ground. Pabbly's terms cap one-time plans at a lower task count than the largest one-time tier its sales page advertises. Those two vendor documents disagree, and the terms are the binding one. Ask support in writing which governs before you buy the top plan. - Assuming a dispute would be simple. The terms are governed by Indian law with any action filed in the courts of Bhopal, and the terms page carries no effective date or version. Starter: Find out how many small automations you actually need When a new row is added to a Google Sheet → then send yourself an email. Free workflows are unlimited but two steps each, so the useful first experiment is counting how many separate little automations your week really contains — the exact number the cheapest one-time plan caps. Zapier and Make both bill by volume rather than by how many automations you keep, so neither makes you ask that question. Build a second one for a different sheet rather than making the first one longer — on this plan that costs you nothing. Starter: Run something on a schedule, then go and read its history When a set time arrives each morning → then email yourself a line pulled from a spreadsheet or a form. With run history cleared after 15 days, a daily job is the cheapest way to build the habit of reading a run log while the log still exists — by the time a monthly automation misbehaves, the evidence may already have been cleared. Deliberately break it — point it at an empty sheet — and look at what the failed run records. Starter: Put one real lead somewhere it won't get lost When someone submits your form → then create the record in your spreadsheet or CRM. Because only the write at the end bills, this shape runs at one task per lead, which makes it the cleanest way to measure your true monthly volume against the smallest one-time plan's allowance before you commit the money. Neither Zapier nor Make asks you to do that sum, because neither sells you a plan upfront. Leave it running a full month before you decide — a single busy week is exactly the sample that makes people buy a bigger plan than they need. Pitfalls: - The countdown doesn't mean what it looks like. The one-time offer is sold behind a "limited time only" timer that has been a fixture of that page. If the price is right for your volume it will still be right tomorrow; don't let a clock make a four-figure decision. - Your allowance appears to reset each month rather than accumulate. Pabbly's materials describe tasks as an amount you get every month, which would mean a plan sized for your busiest month is money you also spend in every quiet one. Size it from a measured week, and if carry-over matters to you, confirm it with support first. - A three-step idea can't be tested on the free plan at all. The free version you evaluate is an approximation of the thing you would be buying — a real limitation when the decision is a single large payment. - Check your specific apps, not the headline number. The integration count is a marketing figure; whether the two apps *you* need are both properly supported is a five-minute check in the integrations directory, and it matters more here than on a monthly tool, because the purchase can't be moved if it turns out not to fit. Where to next: - Workflow & Automation: https://www.launchpadhq.io/discover?category=workflow-automation - AI Plugins & Business Software: https://www.launchpadhq.io/discover?category=ai-plugins-business-software - Sales, Marketing & SEO: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai ### Getting Started with Paperpal Canonical: https://www.launchpadhq.io/platform/paperpal/guide Data: https://www.launchpadhq.io/data/v1/guides/paperpal.json Last reviewed: 2026-09-15 Tool: Paperpal (https://www.launchpadhq.io/platform/paperpal): starting price Free, from See pricing; free tier yes; data as of not dated A beginner's guide to editing an academic paper with AI — and to telling your journal you did. You write or paste your draft into Paperpal's editor, and suggested changes appear alongside your text sentence by sentence; you click each one to accept or ignore it. Select a passage and a small menu offers rewrites of it. There's also a box where you type a question in plain English and get an answer back. What it is: - Paperpal is an academic writing and language-editing assistant from Cactus Communications, the academic-services company behind Editage. It is built for research papers, theses and journal manuscripts rather than general writing, and its suggestions are tuned to how academic English is expected to sound. - It runs in three places — its own web editor, an add-in for Microsoft Word, and a Chrome or Edge extension that puts it inside Google Docs and Overleaf (the browser-based editor many maths and physics papers are written in). One account covers all of them, and the vendor says your plan benefits are the same whichever one you use, so the allowances appear to be shared across them rather than counted separately. - Alongside editing it bundles pre-submission checks — plagiarism, reference accuracy, AI detection (a tool that guesses whether text was written by AI) and a report on how your manuscript measures against journal requirements — plus a research box that answers a typed question with citations. Before you start: - Signing up is free and takes an email, with no card. The free plan gives you a monthly allowance of language-editing suggestions, and a separate small daily allowance for the generative features — paraphrase, trim, translate, academic tone and AI Review. The monthly bucket resets monthly and the daily one daily, so a bad afternoon costs you a day, not a month. - Whether you will need to pay is a question about manuscript length, not ambition. The free allowance is roughly an abstract-and-introduction pass, not a thesis. Prime raises the monthly edit allowance a long way and Pro makes editing unlimited, and both cost less billed annually. Paperpal's pricing page loads its numbers from a logged-in account and may show something different in your region, so check the price you are shown at checkout. - The main editing engine reads English and only English. Paperpal's own documentation lists non-English text among the things Language Suggestions skip entirely, so a paper written in Spanish or Mandarin gets no grammar suggestions at all. Its Translate and Rewrite tools do work in other languages, but they are a different part of the product from the editor you will spend most of your time in. - Look up your journal's or university's AI policy before you edit a word, not after. Elsevier, to take one publisher, requires a declaration when generative AI was used to improve the readability and language of a manuscript, while exempting basic checks of grammar, spelling and punctuation (policy updated June 2026). Paperpal sits on both sides of that line — a comma fix and a rewritten paragraph are the same two clicks — so which of its features you use decides what you have to declare. Security: - Paperpal's commitment on training is written down and it reaches the free plan: your documents, uploads and personal information "are never used for model training and remain fully confidential, on every plan" — and it says the same specifically of work edited through its Overleaf integration, down to not recording which suggestions you accept. It is certified against the international security and AI-management standards (ISO 27001 and ISO 42001) and says it complies with European privacy law and with US student-records law. Security: - That commitment is the reason an unpublished manuscript is a more defensible thing to put here than in a chat window. It is still a company's own statement about a cloud service, not something you can verify yourself. Security: Still worth thinking about: - Your draft is uploaded and stored so it can be re-checked later; you can delete documents and uploaded files yourself when you are done. - A co-author's unpublished data, or anything covered by a confidentiality agreement, is not yours alone to upload. Ask first. - Patient-identifiable material needs its own clearance. Paperpal announced infrastructure built for the US rules covering patient data in March 2026, but that is an arrangement for life-sciences organisations, not something a free account sits inside. Best at: - Tightening English in a draft you have already written — it is an editor, not a ghostwriter, and it is at its best when there is something to edit - Catching the errors that academic English written by non-native speakers tends to produce: article use, tense in methods sections, over- and under-hedging Okay at: - Producing text from nothing. Titles, outlines and abstracts come from filled-in forms rather than a free-form instruction, so you get a competent template shaped like your field, not your argument - The journal-fit report. Free and Prime accounts both get the cut-down version of it; the full report is Pro-only, which is easy to miss because the feature appears on every tier Avoid: - Letting Paperpal's plagiarism or AI-detection scores settle anything. Your institution runs its own tools against its own databases, including sources Paperpal cannot see, so the numbers will not match — and Paperpal has published its own explanation of why text it has edited can be flagged by an AI detector, which means a clean score from its checker is not a prediction of what your university's will say - Relying on it for the parts of a paper it does not read. Language Suggestions skip tables, text boxes, figure and table captions, equations and numbers, abbreviations, text inside quotation marks, the code used to typeset maths, and anything inside an image — so errors in exactly the places reviewers look survive a clean pass - Paraphrasing your way out of a similarity flag. Rewriting a matched passage until the checker stops matching it does not make it cited, and Rewrite makes doing that accidentally very easy Starter: Run one paragraph before you run the paper Paste a single difficult paragraph — your methods section is a good candidate — into the web editor and work through the suggestions one at a time, accepting and rejecting individually rather than applying them in bulk. Paperpal meters the free tier in individual suggestions rather than in documents, and one dense methods paragraph can easily eat dozens of that monthly allowance. Doing a paragraph first tells you how far the free allowance will actually stretch on *your* writing before you spend it discovering that on a full draft. Pick the paragraph you have rewritten most times yourself — that is where you find out whether its suggestions are better than your own. Starter: Ask the research box a real question Paperpal's Research box answers out of a library of published papers and hands back citations you can save straight into your own, so its failure mode is a thin answer rather than a confidently invented reference. Ask a question the literature actually has an answer to. Vague questions come back with vague citations. Prompt: What evidence is there that spaced repetition improves long-term retention in adult learners, and what are the main criticisms of that evidence? Starter: Turn on AI Footprint before you start editing Switch on AI Footprint from the editor's menu before you accept your first suggestion. It marks which parts of the document Paperpal added or changed, and keeps those marks across writing sessions. Most editing tools leave you unable to reconstruct afterwards which sentences were yours — which is exactly what a disclosure statement, or a supervisor's question, requires you to do. Paperpal added AI Footprint alongside one-click disclosure templates drawn from Springer, Wiley and IEEE wording in September 2025 for precisely that reason. Turn it on first — switching it on halfway through cannot tell you about the edits you already accepted. It is not listed on the current plan-comparison table, so whether a free account gets it is not publicly documented; if you are relying on the trail for a disclosure statement, check it is switched on and working before you edit. Starter: Ask a PDF instead of skim-reading it The PDFs you question this way sit in the same library Paperpal's citation tools draw from, so a paper you have just interrogated is one step from being a formatted reference in your draft rather than something you have to go and find again. Asking for page numbers is the point — it gives you something to check the answer against. Prompt: Summarise this paper's method in five bullet points, then list every limitation the authors admit to and the page each one appears on. Starter: Translate a sentence, not a paper Select one passage in another language and use Translate on it, then edit the English yourself rather than accepting it whole. Translate, Paraphrase, Trim and Academic Tone all cap at about 300 words per selection and need at least ten words to run at all, so these tools are built for sentence-level work — feeding them a whole section gets you nothing rather than a warning. Translate into English, then run the result past the editor as if you had written it. The help centre documents translation between 50-plus languages while an earlier vendor post described it as translating into English, so check the direction you need works before you rely on it. Pitfalls: - Real-time suggestions are documented up to around 35,000 words, and Paperpal advises splitting anything larger. A book-length thesis goes through in chunks, which also means your chapter numbering and cross-references are on you. - Each submission check is separately metered and separately gated, so "Paperpal has a plagiarism checker" and "my plan will run one on this manuscript" are different claims. Read the plan comparison table before you rely on any of them the night before a deadline — consistency checks and reference checks are not on the free plan at all, and the free plagiarism check returns what Paperpal calls a Lite Report, which appears to be a similarity range with top sources rather than a full itemised one. Exactly what is withheld is not spelled out publicly, so expect less detail and confirm it once you are signed in. - Paperpal Preflight is a separate product. It is the submission-readiness service offered through publisher partners, it is not included in any subscription, and it carries a per-manuscript charge (rechecks on revisions are included in that charge). Seeing Preflight in a journal's workflow does not mean your Paperpal plan covers it. - Accepting every suggestion flattens your voice. Extensive mode deliberately produces more and heavier rewrites than Essential mode, and on a paper you will have to defend in a viva or a review response, more is not automatically better. - The plan table has recently grown a row for a category of feature Paperpal has not yet published anything describing, marked as limited on the free plan — so you may find something in the product that this page does not cover. - It edits what you wrote, not what you meant. A sentence that is grammatically immaculate and scientifically wrong comes back immaculate. Where to next: - Document & PDF tools: https://www.launchpadhq.io/discover?category=document-pdf-processing - Education & learning AI: https://www.launchpadhq.io/discover?category=education-learning-ai - More research tools: https://www.launchpadhq.io/discover?category=research-academic-tools ### Getting Started with PathAI Canonical: https://www.launchpadhq.io/platform/pathai/guide Data: https://www.launchpadhq.io/data/v1/guides/pathai.json Last reviewed: 2026-09-12 Tool: PathAI (https://www.launchpadhq.io/platform/pathai): starting price Custom; free tier no; data as of not dated AI that reads tissue slides inside hospital and drug-company labs — sold to institutions, not to people. A lab scans a glass slide of tissue into a very large digital image, which lands in a list in the browser of the doctor reading it. They open it, pan and zoom like a map, switch on AI tools that mark up what they see, then type their findings into a report. What it is: - PathAI is a Boston-based company building AI software for pathologists — the doctors who examine tissue samples under a microscope and say what a disease is. Its own description of who uses the platform: "biopharma, reference laboratories, academic medical centers, and health systems." - It sells two things: a platform that holds, displays and routes a lab's scanned slides, and a set of AI models (the AI "brain" that does the actual thinking) — PathExplore, IHCExplore, LiverExplore, IBDExplore, TumorDetect, ArtifactDetect, PathAssist Derm, AIM-MASH — which run on it and mark up or measure what is in an image. - PathAI used to own a pathology laboratory business too; it was sold to Quest Diagnostics in 2024, and PathAI has described itself since as "an independent company, dedicated to advancing AI and digital pathology solutions for biopharmaceutical companies and pathology laboratories." Before you start: - You cannot buy this as an individual, and there is no price to look up. Every route on pathai.com ends at a "Request a Demo" form or an email address — [email address on the web page] for the diagnostic platform, [email address on the web page] for research partnerships, both ending in a sales conversation rather than a download. As of 11 September 2026 we found no sign-up page, no published price list and no free trial anywhere on the site. - Two products share the AISight name, and the small print tells them apart. One footnote reads: "AISight® Dx is FDA-cleared for primary diagnosis in the US with the Hamamatsu NanoZoomer® S360MD, Leica Aperio® GT 450 DX, Roche VENTANA® DP200 and DP600 slide scanners and is CE-IVD–marked for primary diagnosis in the EEA, UK, and Switzerland." The other reads: "AISight® is for Research Use Only. Not for use in diagnostic procedures." Research Use Only is a legal status, not modesty: PathAI is not permitted to sell that product for working out what is wrong with a patient. "Primary diagnosis" is its phrase for the opposite case — the diagnosis a patient's treatment is based on. - Those footnotes name four places: the US, the EEA, the UK and Switzerland. PathAI announces deployments elsewhere — Brazil's A.C.Camargo Cancer Center in May 2026 — but we could not establish from its public pages what regulatory status applies outside those four. - One route ends with somebody outside a customer organisation receiving something. The PathExplore Data Access Program offers approved academic researchers the measurements PathAI's software produced from slides in a public cancer-research archive called TCGA; the announcement states: "If approved, there is no cost to academic researchers to access and use these data sets." What arrives is the software's output, not the software. - Everything else PathAI makes public is reading: its news index and resource library are open, and the US regulator's clearance record is a free public database — the starter actions below are built around both. If you came wanting to ask an AI about medicine in plain language, this site's Getting Started with ChatGPT page will get you typing in a minute. Security: - If PathAI's software ever touches your data, it is because a laboratory you sent a tissue sample to is running it. Its privacy notice, effective 13 February 2025, says: "In certain cases, your access and use of our Platforms may result in your providing access to Protected Health Information ('PHI') as defined by the Health Insurance Portability and Accountability Act of 1996 ('HIPAA')", and "our processing of PHI is governed by Customer Agreements, Terms of Use, and Business Associate Agreements, as applicable." The contract protecting a patient is the laboratory's, not the patient's. Security: Two documents that are easy to mistake for each other: - The Terms of Use on pathai.com are website terms — effective 3 May 2018, last updated 10 May 2022 — and say so, noting that PathAI "software products and related services may also be governed by separate Terms of Service." Those product terms are not published, so the binding document for anyone using the platform is one the public cannot read. - The privacy notice reserves the right to share "aggregated user statistics and de-identified information derived from Personal Information, with third parties for any purpose." - PathAI announced ISO 27001 certification — an independent audit of how a company manages information security — in March 2019. We found no more recent statement on its site, so treat that as a 2019 fact, not a current one. Best at: - One browser-based place for a lab's scanned slides — PathAI pitches AISight as "a central hub for case management, image management, and best-in-class artificial intelligence tools." - Running other companies' algorithms next to its own: the site lists Deep Bio, Paige, Visiopharm and DoMore Diagnostics as partner algorithms on the platform. - Measuring the same thing the same way at scale — PathExplore is advertised as over 600 standardised measurements per sample, at single-cell detail, across 14 cancer types. - Drug trials: PathAI announced on 9 December 2025 that AIM-MASH is the first AI-powered pathology tool to receive FDA qualification under the Drug Development Tool programme, for clinical trials of MASH, a serious liver disease. Okay at: - Being understood from outside. Nearly everything published about how the platform behaves is PathAI's own description of it, and none of it can be checked without an account. Avoid: - Anything diagnostic on a product carrying the Research Use Only footnote — PathExplore, PathAssist Derm and the AISight image management system all carry "Not for use in diagnostic procedures" in PathAI's own materials. - Reading "FDA" as though it meant one thing. On PathAI's own pages it does three jobs: AISight Dx is cleared (510(k) K243391); AIM-MASH is qualified as a drug development tool for clinical trials and is, in PathAI's words, "not for use in diagnostic procedures"; PathAssist Derm got Breakthrough Device Designation in March 2026 and is still Research Use Only. A designation is a place in a review queue, not permission to diagnose anyone. - Assuming the platform's clearance covers the algorithms running on it. K243391 is filed under the regulation "Digital Pathology Image Viewing And Management Software" — the viewing and workflow layer; PathAI's own and its partners' algorithms carry separate statuses. Starter: Look up the clearance yourself Open the US FDA's public 510(k) database and search the number PathAI prints in its own footnotes: K243391. The record comes back: device AISight Dx; applicant PathAI, Inc., Boston; decision date 26 June 2025; decision "Substantially Equivalent"; regulation name "Digital Pathology Image Viewing And Management Software"; product code QKQ. PathAI prints the K-number in its marketing footnotes, so the trip from a vendor claim to the government record behind it is one search — and the "regulation name" line names what was actually cleared. Starter: Read the footnote before the headline Open PathAI's PathExplore page and its AISight Dx page in two tabs and read the small grey line at the bottom of each before anything else. PathAI's own navigation lists "AISight® Image Management System (RUO)" and "AISight® Dx Digital Pathology Platform" — the three-letter difference in the name, and the footnote under it, carry the entire regulatory distinction. Starter: Apply for the free academic data set, if that is you Email [email address on the web page] to apply to the PathExplore Data Access Program. It is the only path on the site ending with a person outside a customer organisation actually receiving something — the software's output on public images, not the software. Starter: Follow the release notes rather than the press releases On PathAI's news index, look for the numbered entries: AISight Dx v2.17 in October 2025, v2.18 in December 2025, v2.19 in March 2026, v2.20 in May 2026, v2.21 in July 2026, each headline naming what changed — dashboards, viewer usability, reporting, configurability. A numbered release roughly every two months is a harder signal than a partnership announcement: partnerships tell you who signed, the version stream tells you the platform is genuinely shipping. Pitfalls: - Searching "PathAI" and landing on a laboratory. Quest Diagnostics bought select assets of PathAI Diagnostics; the digitised Memphis lab was rebranded AmeriPath and became Quest's national AI and digital research centre, and Quest separately licensed the AISight platform for its own labs. That laboratory is not the company that makes the software. - "Self-service" in PathAI's announcements does not mean you. Its 18 November 2025 release, "self-service integration of PathExplore on AISight", describes an institution already running AISight switching PathExplore on without PathAI's implementation help. - Expecting the demo link to show you the product. "Request a Demo" books a conversation. PathAI does run demo.aisight.pathai.com, but loading it gave only a browser-requirements notice — Chrome or Edge, JavaScript on — with no visible way to create an account, so we could not establish what sits behind it. Where to next: - Healthcare AI: https://www.launchpadhq.io/discover?category=healthcare-ai - Research & academic tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Chat assistants anyone can open today: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Perplexity Canonical: https://www.launchpadhq.io/platform/perplexity-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/perplexity-ai.json Last reviewed: 2026-07-13 Tool: Perplexity AI (https://www.launchpadhq.io/platform/perplexity-ai): starting price $20/month; free tier yes; data as of 2026-10-26 An answer engine: ask a question, and it searches the live web and replies with a written answer plus clickable citations. You type a question and Perplexity searches the web, reads the pages it finds, and writes an answer with numbered source links inline. You click a link to check the claim it supports, then ask a follow-up in the same thread and it searches again with that context. What it is: - Perplexity is an answer engine — a cross between a search engine and a chatbot. You ask a question in plain language; it searches the web in real time, reads multiple sources, and gives you one written answer with numbered citations you can click to check. Unlike a general chatbot working from training data, it's grounded in current web results; unlike Google, it hands you a synthesized answer instead of a list of links. It's for anyone who does research or fact-finding — students, journalists, the curious. Before you start: - The free tier is genuinely capable: unlimited basic cited searches, all the focus modes, saved research spaces, on any device. The deeper Pro-search mode is capped on free (a handful every few hours). - Pro is a paid monthly plan (cheaper annually; students can get a discounted rate) for unlimited deep searches, file uploads, and switching between frontier models. - The one habit to build from day one: click the citations. More on why below. Security: - Perplexity is citation-transparent, but not always citation-accurate. It shows you sources, which is a real strength — but independent testing has found meaningful error rates: it can attach a statistic to a source that doesn't actually contain it, or cite the wrong page. Treat it as a research starting point, not the final word. Security: How to use it without getting burned: - Click through the citations on anything that matters — verify the source actually says what Perplexity claims before you rely on it. - When you write it up, cite the original source, not "Perplexity." It's the tool that found the source, not the source itself. - Don't use it for high-stakes decisions (medical, legal, financial) as a primary source — use it to find the real sources, then check those. Security: - On data: Perplexity states it doesn't use free searches to train models, and its enterprise tier adds explicit no-training guarantees. As with any web tool, avoid pasting confidential material into your queries. Best at: - Answering current, factual questions with sources attached - Fast first-pass research and finding the right sources to read - Replacing a pile of Google tabs with one cited synthesis - Letting you verify every claim in one click Okay at: - Citation accuracy (transparent, but verify — error rates are real) - Deep judgment calls (it synthesizes whatever ranks highest, not the most considered view) - Niche/technical topics (it may cite a forum post as if authoritative) Avoid: - Creative writing, coding, or image generation (wrong tool) - Trusting a cited number without clicking the source - Citing 'Perplexity' instead of the original source in real work Starter: Ask a current factual question A question with a moving, factual answer is exactly where Perplexity beats both Google and a plain chatbot — and every claim comes sourced. Ask about something you're actually deciding on; then click every citation. Prompt: What are the current pros and cons of induction stoves versus gas, and roughly what does an induction range cost right now? Starter: Verify a citation Open the linked source and confirm it actually says what the answer claimed. This is the single habit that makes Perplexity trustworthy instead of just confident. If a source doesn't support the claim, re-ask the question or discount that point. Starter: Research a topic you'll write up Shows the intended workflow — use it to find and understand sources, then write it yourself citing those sources. Ask follow-ups in the same thread to drill into whichever source looks strongest. Prompt: Give me an overview of how tariffs affect consumer prices, with sources I can read to go deeper. Pitfalls: - Transparent isn't the same as accurate. It shows sources but can misattribute — always click through on anything that matters. - It's a research tool, not a writing or coding tool. For creative work or code, use a general assistant instead. - Cite the source, not the engine. In real work, trace every claim to the original and cite that. Where to next: - Research & Academic Tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Perplexity Comet Canonical: https://www.launchpadhq.io/platform/perplexity-comet/guide Data: https://www.launchpadhq.io/data/v1/guides/perplexity-comet.json Last reviewed: 2026-10-15 Tool: Perplexity Comet (https://www.launchpadhq.io/platform/perplexity-comet): starting price $17/mo; free tier yes; data as of 2026-10-24 A full web browser with an AI assistant built in — it can read the page you're on and, if you let it, click and type on it for you. You install Comet and browse the web as you would in any other browser. When you want help, you open the assistant panel beside the page and type a request in plain words. It reads the page you're on and answers, or clicks and types on the page itself while you watch, stopping to ask before anything important. What it is: - Comet is a web browser made by Perplexity, the company behind the Perplexity AI search app (which has its own page on this site). It looks and behaves like Chrome — Perplexity's help centre says most Chrome extensions (add-ons) work in it — but every page you open has an assistant beside it that can read that page, summarise it, compare it with your other open tabs, and draft replies from it. - The bigger difference is that the assistant can also act. Comet works as an agent (AI that takes actions on its own rather than only answering): Perplexity's own guide describes it comparing products and going through to checkout, searching your inbox once you connect Gmail, and moving meetings on your calendar. That ability is the reason to choose Comet over a chatbot in a separate tab, and it is also the reason this page spends more time than usual on safety. - Google now builds a similar assistant into Chrome itself, so 'a browser with AI in it' is no longer unusual. What Comet gives you is a browser built around the assistant from the start, with Perplexity's search underneath and its own set of controls for limiting what the assistant can touch. Before you start: - You can start for nothing. Perplexity says it will always provide a free version, and that the free version includes Comet. It also sells paid plans — Pro at $17/mo (billed annually) and Max at $167/mo (billed annually) — but Perplexity's public pages, as checked for this guide, don't spell out what either adds inside the browser, so download the free version and see whether you hit a wall before paying for anything. - Comet is something you install, not a website. Perplexity lists it for Mac, Windows, iPhone and Android; Linux isn't on that list. Perplexity's setup guide has you sign in to a Perplexity account the first time you open it, so expect to create one if you don't have one. - Moving over is quick. Comet can bring in your bookmarks, history and saved passwords from the browser you use now (Settings → import bookmarks and settings), and you don't have to make it your default browser to try it — running it alongside your current one is a sensible way to start. - The assistant becomes most powerful, and most risky, on sites where you're already logged in — your email, calendar, shopping accounts. Read the security section below before you connect any of those. Getting set up safely (Download Comet only from Perplexity's own site — perplexity.ai/comet, or the download page its help centre links to at perplexity.ai/download-comet. The download button sends you through a Perplexity link-redirect address (perplexity.sng.link) on the way to the installer; that's expected. Don't install a 'Comet' from a search ad, a download-bundle site or a link someone sent you: a browser sees every site you log in to, so a fake one is about the worst thing you can install.): - On a phone, follow the link from perplexity.ai/comet rather than searching your app store for 'Comet', where similarly named apps can turn up. - When you import from your old browser, you choose what comes across — bookmarks, history, passwords. Bringing passwords is optional; you can leave them behind and add them later. - Comet updates itself when you relaunch it. You can check for updates by hand under Settings → About — worth doing now and then, because several of the security fixes described below arrived as updates. - Your 'it works' checkpoint: open any article, open the assistant panel, and ask it to summarise the page. A summary that matches what's on screen means you're set up. Security: - An assistant that reads web pages can be tricked by the pages it reads. Text hidden in a page — white-on-white, tucked inside a comment, faint lettering in an image — can carry instructions the assistant mistakes for yours, and in Comet those instructions can reach whatever you're logged in to. This is not hypothetical. Between August 2025 and February 2026, outside researchers publicly demonstrated it in Comet several times: Brave showed a hidden Reddit comment steering Comet, when asked to summarise the page, into reading a one-time login code from the user's open Gmail and posting it back to the attacker; LayerX showed a single crafted link doing similar work; and Zenity Labs showed a booby-trapped calendar invite leading Comet into a logged-in 1Password vault. Perplexity fixed the issues as they were reported — not always on the first attempt, and Brave said in its own update that the August 2025 attack had not been fully mitigated — and Perplexity itself commissioned a security firm to attack Comet before launch. Security: What Perplexity has built, and what's still on you:: - Comet asks before the high-stakes steps. Perplexity says it pauses for your approval before sending an email or message, changing your calendar, placing a final order, or filling in personal details it doesn't already have — even when nothing looks suspicious. Read what it's about to do before you approve; that confirmation is the last line of defence. - Block the sites that matter most. Settings → Privacy and security → Comet Assistant lets you switch the assistant off entirely or list sites it must never act on or read. Perplexity's own example is your bank. - Be selective about what you connect. Every account you link, and every site you stay logged in to, widens what a hidden instruction could reach. Don't connect your work email to a personal experiment. - Perplexity also says it scans the pages Comet retrieves for hidden instructions and tells you when it blocks something. That's a filter, not a guarantee — researchers have repeatedly found ways past such filters. Security: - On what leaves your computer, Perplexity's help centre says the assistant doesn't upload your browsing history, cookies, saved passwords or local files by default, and sends page content to Perplexity only when a request needs it — asking it to summarise the tab you're on, for example. When it does, Perplexity says that context is kept for up to 30 days and then deleted automatically; those conversations appear as 'Temporary threads' in your Perplexity library, where you can delete them sooner. It also says it doesn't sell this data. Best at: - Questions about the page you're already on — summarising it, pulling out dates or steps, translating a section — without copying anything into another app - Comparing several open tabs at once, by mentioning each tab in your request - Repetitive web chores you'd happily hand off and then check — building a basket for review, gathering options into one list Okay at: - Multi-step tasks on unfamiliar or busy sites. The agent can get there, but it's slower than doing it yourself while you're still learning to trust it - Being your everyday browser. It's a capable Chrome-style browser, but you're switching browsers to get the assistant, not for the browsing itself Avoid: - Letting the assistant loose on a page you don't trust while you're logged in to email, banking or a password manager — the documented attacks above all worked through exactly that combination - Assuming every site welcomes it. Amazon sued Perplexity over Comet shopping on Amazon; a court blocked it in March 2026, an appeals court lifted that block in August 2026, and the underlying case is still undecided Starter: Summarise a page you trust Each time you send a request, Comet pulls in the visible text, headlines and details of the page you're on — so a summary is the natural first test. Start on a site you trust: the Brave attack described above came through exactly this kind of request, which is why the page you ask about matters in Comet in a way it doesn't in a chatbot that never sees your logged-in tabs. Add another open tab with @ — for example, '@ the other review: where do these two disagree?' — to compare pages side by side. Prompt: Summarise this page in five bullet points, then list any dates, prices or deadlines it mentions. Starter: Block your bank before you do anything else Go to Settings → Privacy and security → Comet Assistant and add the sites the assistant should never read or act on — your bank, your password manager, anything financial. Comet's block list is the one control that works before any page gets a chance to say anything to the assistant — a blocked site isn't used for tasks or as context at all. It's a two-minute job. Starter: Hand over a task where nothing is lost if you say no This is an example from Perplexity's own use-case guide, and it lets you watch the agent click through a site, then meet the pre-order confirmation described in the security section. Declining at that point costs you nothing, so it's the safe way to learn what an approval request looks like before one matters. Use a store you already have an account with, and say no at the confirmation the first time. Some stores object to AI shopping (see the Amazon case above), so expect some sites to behave differently. Prompt: Build a grocery cart with the ingredients for lasagna for four people, ready for me to review before checkout. Starter: Save a request you'll repeat as a shortcut Type / in the search bar or assistant panel, choose 'Create a shortcut', write the request once, name it, and from then on call it up by typing / and its name. A Comet shortcut stores more than your wording — it also saves which Perplexity search mode, AI model and sources to use — and choosing it fills in the request without sending it, so you can add details first. It's a way to make Perplexity's search behave the same way every time without retyping your instructions. Good first shortcuts: 'summarise this page for a busy reader', or 'find the cancellation policy on this site'. Pitfalls: - The first time an advanced task runs, Comet asks whether to allow it this time only, always, or not at all. Choose 'this time' until you've seen it work a few times. 'Always' removes a check you may later want back. - Confusing Comet with Perplexity. Perplexity's answer engine is a website and app you ask questions in; Comet is the browser built around it. Help articles and settings for one don't always apply to the other. - Walking away mid-task. The assistant works while you carry on browsing, and you can run several tasks at once — but on a logged-in site, keep an eye on what it's doing rather than coming back later to see what happened. Where to next: - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai - Perplexity (the answer engine): https://www.launchpadhq.io/platform/perplexity-ai - Research AI: https://www.launchpadhq.io/discover?category=research-academic-tools ### Getting Started with Photoroom Canonical: https://www.launchpadhq.io/platform/photoroom/guide Data: https://www.launchpadhq.io/data/v1/guides/photoroom.json Last reviewed: 2026-10-14 Tool: Photoroom (https://www.launchpadhq.io/platform/photoroom): starting price $7.50/mo billed yearly (8,000 AI credits, 1,000 exports). Max and Ultra are listed as yearly billed plans; free tier yes; data as of 2026-10-14 A beginner's first half-hour with the AI photo editor built for product listings. You drop a photo of a product into the editor and Photoroom cuts the object away from its background in a second or two. From there you pick a backdrop from a list or type a description of the scene you want, look at what comes back, and keep re-describing until it fits. What it is: - Photoroom is a photo editor built around one trick — separating a product from what it was photographed against — and around what comes next: a plain backdrop, a generated scene, better lighting, a straighter crop. - It is built for people who sell things, not photographers. Its customer stories are Depop, Decathlon and Palm Angels, and its tools are named product staging, virtual try-on, ghost mannequin, listing images. - It runs as an iPhone, iPad and Android app and as a browser app at app.photoroom.com; one account works across all of them, though not identically. - It will also invent a whole image from a written description, dress a virtual model, and generate short product videos — the expensive end of the product, and the part to try last. Before you start: - The free plan is licensed for personal use only. Photoroom's terms, effective 29 July 2026 and in force today, license free accounts "for your own personal, non commercial purposes," and its help centre puts it plainly: "You cannot use Photoroom for commercial use." If the photo is going on an Etsy, eBay or Depop listing, the free plan is for deciding whether you like the tool. - Free accounts currently have weekly and monthly export caps, counted across anything that produces a file — downloads, camera-roll saves, shares. Editing is not what runs out; saving is. - Free export resolution depends on which Photoroom you use: the vendor documents a 750 × 550 pixel ceiling on the iPhone, iPad and Android apps against 2000 × 2000 in the browser. Those figures were last revised in August 2025, so treat them as the published position rather than a guarantee about today. - Generative tools run on credits (the platform's unit of spend — each thing you make costs some); background removal, retouch, preset shadows, expand, resize and the image enhancer do not, listed by the vendor as unmetered under fair use. Free accounts get some credits, but Photoroom does not publish how many — its help centre says only "a low limit," and the pricing comparison table has no free column at all. - Paying starts at Pro, billed yearly, with a much larger allowance of AI credits and exports. Max and Ultra are higher yearly billed tiers. You move up for volume and for the commercial licence, not for better background removal — that is the same tool the free plan already gives you. Security: - Photoroom trains its models on what you upload, by default, and its privacy policy is blunt about the consequence: switching the setting off later "shall not have retroactive effect on processing activities already carried out." Find the switch in your first five minutes. Security: Where the switch lives: - iPhone, iPad and Android: Create → Content → your workspace → Manage account → Data control → "Improve model for everyone" - Browser: log in at app.photoroom.com → Create → Preferences → Your profile → scroll to "Terms & conditions" → "Improve model for everyone" - It is per-person. Photoroom states there is currently no way to set it once for everybody sharing a workspace, though it says it plans to add one. Security: - The terms also make you warrant you hold every right and consent needed for anything in an image you upload, naming likenesses specifically — a friend's hand holding your product is a consent question here, not just a courtesy. Security: - Photoroom is a French company, its privacy policy is written to European rules naming the CNIL as its supervising authority, and it publishes retention periods per category of data, so access and deletion rights are exercisable. Best at: - Separating a product from a cluttered background in one tap — still the thing it does best - The same cutout, backdrop and crop across a whole folder, on the paid plans - Reshaping one image into the several aspect ratios different storefronts each insist on - Describing a backdrop in words instead of hunting for a stock photo that nearly matches Okay at: - Images of people. The virtual-model and try-on tools work but are the most credit-hungry part of the product, and Photoroom's own prompt guide is written around reducing retries — which tells you retries are normal. - General photo editing. No layers, curves or masks in the way a photo editor usually means them. - Being the same product everywhere. The phone and browser apps differ in prompt options and in what they will export, so a task you gave up on in one may work in the other. Avoid: - Planning a high-volume month around "unlimited exports." That phrase still appears on Photoroom's own plans-overview article, while its pricing table and its newer, more specific export-limit article both give the Pro cap above. Assume the number, not the adjective. - Buying Pro for the video generator: Photoroom's own plan comparison marks video "Not included" on Pro despite video sitting in the top-level product menu. - Budgeting a fixed monthly output from the advertised credit figure outside the US. The vendor's pricing FAQ states that "the monthly credit amount included with your plan may vary depending on your country or region," without publishing how much it varies. Starter: The cutout, on a photo you already have Open the background remover in a browser, drag in a photo of something on your kitchen table, and download a transparent PNG — a cutout with no background baked in. Tested against Canva, which also removes backgrounds: Canva's lives inside a design you already opened and gives back a design. Photoroom's is a page you drop a file onto, and gives back the transparent cutout itself — the asset you reuse against every later background. Try an awkward edge — a wicker basket, headphones — not a clean rectangular box. Starter: Describe the scene, not the product Tested against Midjourney: a Midjourney prompt (the message you type) must describe the product too, because Midjourney invents the whole picture and can invent your mug wrong. Here the mug is already cut out and pinned in place, so every word goes on the scene — which is why Photoroom's prompt guide says to name the setting, materials, lighting, camera angle and intended use, and end with what you do not want. Change only the last sentence between attempts — "do not include reflections" — negative phrasing the vendor documents as supported. Prompt: A ceramic mug on a pale oak table beside a folded linen napkin, morning window light coming from the left, a soft shadow under the mug, shot from slightly above at a shallow angle, clean and minimal, for an e-commerce product photo. Do not include other objects, hands or text. Starter: Run the same description on two different engines In the AI Backgrounds settings, run your description once on an earlier engine (v2 or v3) and once on a Studio one, then compare. Tested against Canva's Magic Media, which picks the model (the AI "brain" doing the work) for you without mentioning there was a choice: Photoroom exposes the dial and documents the trade — v2 and v3 give lower-resolution output and cost less to run, the Studio engines higher-resolution output and cost more. Starter: Invent the photo you never took Paste this into "Create any image" under AI Tools, which Photoroom documents as a browser-app feature. Tested against ChatGPT, which also makes pictures from a sentence: what comes out here lands in the same editor as your real photographs, cropped, resized and exported by the same tools. Keep the last line and change the wall: "a terracotta tiled wall," "weathered timber." Prompt: A navy canvas tote bag hanging on a hook against a whitewashed plaster wall, late afternoon light raking across from the right, shot at eye level and centred with plenty of empty wall above it, for an e-commerce product photo. Do not include people, hands or other bags. Starter: Export one throwaway file before you do any real work Export one junk image from whichever Photoroom you plan to live in, before editing anything you care about, and check the file's pixel dimensions. Tested against remove.bg, which shows the resolution before you commit: Photoroom does not surface its ceiling while you edit, and the phone and browser ceilings differ by a factor of four (see Before you start). One throwaway export buys that answer. Pitfalls: - Editing a folder spends the allowance per image: a batch of ten counts as ten exports, so one bulk run can eat most of a free week. - Re-running an AI tool counts as a new generation, and so does editing an image the AI just made: the tenth background attempt costs what the first did. - Photoroom automatically returns credits when a generation fails on a technical error, but an image that arrives intact and simply isn't what you pictured is a normal, paid-for result. - Photoroom compresses imported images to 2000 × 2000 by default, and a low-resolution import stays low-resolution inside a larger design. Keep your originals somewhere else. - Subscribe inside the phone app and Apple or Google bills you, not Photoroom — cancellations and refunds go through that store. Photoroom's own refund window is one week from purchase. Where to next: - Image generators and editors: https://www.launchpadhq.io/discover?category=image-generation-editing - Video creation tools: https://www.launchpadhq.io/discover?category=video-creation-editing - Sales and marketing AI: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai ### Getting Started with Pictory Canonical: https://www.launchpadhq.io/platform/pictory-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/pictory-ai.json Last reviewed: 2026-09-17 Tool: Pictory AI (https://www.launchpadhq.io/platform/pictory-ai): starting price $29/mo; free tier yes; data as of 2026-10-14 Turn writing you already have into a narrated video, without opening a video editor. You paste a script or point it at something you have already written, and it splits the text into scenes, finds matching clips for each one, reads the words aloud in a chosen voice and puts captions on screen. You then work down the scene list swapping any clip or line you do not like, and download the finished video. What it is: - Pictory is a browser tool that turns writing you already have — a script, a blog post, a PDF, a slide deck — into a narrated video with captions and matching footage. - It does not film or draw anything original by default — the clips are ready-made and the voice is synthetic. Pictory's help centre names Storyblocks and Getty Images as the sources of the visuals and Melodie as the source of the background music. - It also works in the other direction: upload a webinar, a Zoom recording or a podcast episode and Pictory writes out everything said, so you can shorten the video by deleting sentences from the text. - A newer layer generates visuals from scratch — images, short AI video, and computer-generated presenters — rather than pulling them from the library, and it is priced separately from everything else here. Before you start: - There is no free tier. Pictory currently gives you a 14-day trial that its pricing page keeps tight, with small limits on how many videos you make and how long they run; its signup pages say no card is needed to start. After that it is a paid product, starting with the Starter plan, a little cheaper if you pay for a year up front. - Assume trial exports carry Pictory's branding. Its plan-comparison table lists "no watermark" (a mark on generated output identifying it as AI-made) against the paid plans only, and its free script-to-video landing page also files watermark-free export under the paid tiers. Neither page states outright that the trial is watermarked, so this is an inference from where the vendor puts the line rather than something Pictory says — worth confirming in the app before you build anything around trial output. - Two separate allowances run at once. Video minutes cover finished videos — each plan has a monthly allowance, larger on Professional than on Starter. AI credits (the platform's unit of spend — each thing you make costs some) are a different pot entirely, spent only when you generate original AI images, AI video or a presenter, and Pictory's credits article says unused ones roll into the next cycle. - The language you narrate in decides which voice pool you draw from. Pictory's help centre lists 7 languages for its standard voices — English, Dutch, French, German, Italian, Portuguese and Spanish — and 29 for the premium voices it licenses from ElevenLabs, which every plan meters by the minute. Whether transcription and auto-captions cover the same 29 languages is not stated on any Pictory page we could reach. Security: - Everything you put into Pictory goes to its cloud and stays there. Pictory advertises independent security auditing and European privacy compliance, and its privacy policy (last updated 17 February 2026) extends deletion, access and portability rights to all users rather than only EU ones. What that policy does not say, either way, is whether your uploaded content is used to improve Pictory's own AI. Treat that silence as a genuine unknown, not as a no. Security: Think twice before uploading: - Internal recordings — all-hands, strategy calls, anything nobody cleared for a third-party vendor - Client footage covered by a contract or an NDA (a signed agreement not to share someone's confidential information) - Unreleased product material - Recordings of other people's faces and voices, where they have not agreed to a vendor processing them Security: - Your text does not stay inside one company: the premium voices are ElevenLabs', and the generated visuals come from an assortment of outside image and video models that Pictory names in its credits pricing. Pictory's privacy policy names its cloud hosts but does not list these AI suppliers individually, so you cannot tell from the public documents exactly whose servers a given generated shot passed through. Best at: - Getting captions onto footage without hand-timing a single one - Volume: the tenth version of an explainer costs about what the first did Okay at: - Clip choice. A concrete sentence gets a usable shot; an abstract one gets a generic office corridor you will want to swap - Voice performance. The standard voices read cleanly but flatly; the ones that sound genuinely human come out of the premium pool - Slide decks. It converts PowerPoint, metered against a monthly slide allowance, but the result is a narrated slideshow rather than a redesign Avoid: - Publishing anything with Pictory's background music to YouTube without following Pictory's own new rule. Its help centre states that from 1 September 2026 you must add the appropriate clearance code — a short string in your video description that tells YouTube's system you are licensed — before publishing, or the track's owner can claim your video. Pictory points to a separate article for where that code comes from, which we could not retrieve, so budget time to find it. - Lifting a clip or a music bed out of a finished Pictory video to use on its own. Pictory's terms of service (last modified 14 July 2026) licence that third-party material only as part of the video you made in Pictory, and separately bar using any of it to train an AI system. - Expecting the footage to be about you. A video about your product shows somebody else's product unless you upload your own material, which is a slower workflow than the one Pictory is fast at. Starter: Paste a mixed script and watch what it picks Pictory chooses footage sentence by sentence on keywords, so a script mixing concrete nouns with abstractions shows you both outcomes in one pass — the bank-account lines get usable shots, "order of operations" gets a stock boardroom. invideo AI starts instead from footage you upload and an edit you describe, so it never makes this guess. Run the same script twice with two different themes to see how much of the result is the theme rather than the writing. Prompt: Most people think saving money is about willpower. It is not. It is about order of operations. First, work out what leaves your account automatically every month. Second, move the savings transfer to the day you get paid, before anything else goes out. Third, leave the rest alone. Nothing here requires you to spend less on coffee. It requires you to spend after you have saved, not before. Try it for one pay cycle and look at the balance. Starter: Cut a long recording by deleting sentences Upload a webinar, Zoom call or podcast episode, wait for Pictory to write out everything said, then delete the filler sentences from that text and watch the matching video go with them. Descript does transcript-based cutting too, and does it on a permanent free plan with a monthly cap on how much media you can work with. What Pictory adds on the same file is offering to pick the summary sentences for you and rebuild the recording as a short — Descript leaves every choice to you. Starter: Export once, change something, export again Download your first finished video, then go back into the same project, swap a clip or fix a caption, and download it a second time. Pictory's help centre says video minutes are deducted only on the first download of a project and that later downloads of the same project do not touch the quota, so finishing inside one project is cheaper than starting a fresh one each time you change your mind. invideo AI bills a monthly credit pot that its own pricing page says does not roll over, so the same habit buys you nothing there. Starter: Spend a few credits on one deliberate generated shot Pictory publishes a per-model, per-second price for this — its credits article lists a per-second credit cost for generated video that varies widely depending which model you choose, and a separate rate for a presenter. Descript meters AI credits per plan too, but against editing features rather than a published per-second rate, so there is no equivalent meter to calibrate against. Run the same description on the cheapest and the dearest model in the list and decide whether you can see the difference at the size you publish. Prompt: A slow overhead shot of a wooden desk at dawn, one open notebook and a cooling cup of coffee, warm low light, no people, no text on screen. Starter: Find out which voice pool your language falls in Before writing anything long, open the voice picker and check whether the language you need is in the seven standard-voice languages or only in the premium set. Pictory meters the two pools differently: standard voices are unlimited on a paid plan, while the premium voices come out of a capped monthly minute allowance, currently around 60 minutes on Starter — so the language you write in decides whether narration is effectively free or rationed. Descript publishes one flat transcription-language list per plan, currently 25 languages including on its free tier, with no second pool to run out of. Pitfalls: - The first pass's clip choices are the weakest thing about the output — "our team has grown" reliably returns a stranger's open-plan office. Budget time to swap shots; publishing pass one is how Pictory videos come to look like Pictory videos. - The trial clock runs on days, not on use — it expires whether or not you make anything, so start it on a day you already have a script ready. - Annual and monthly plans describe the same allowance with very different-looking numbers — Starter's minutes appear as a yearly total on one and a monthly figure on the other. Those are the same thing said two ways, not a bigger pot for paying annually. - Any single video is capped at 30 minutes on the paid plans, so a two-hour webinar cannot come out whole. It has to be cut down or split first, which is what the transcript editor is for. Where to next: - Write the script first: https://www.launchpadhq.io/discover?category=text-conversational-ai - Voice and speech tools: https://www.launchpadhq.io/discover?category=voice-speech - Other video tools: https://www.launchpadhq.io/discover?category=video-creation-editing ### Getting Started with Pika Canonical: https://www.launchpadhq.io/platform/pika/guide Data: https://www.launchpadhq.io/data/v1/guides/pika.json Last reviewed: 2026-10-14 Tool: Pika (https://www.launchpadhq.io/platform/pika): starting price Free, from $10/mo; free tier yes; data as of 2026-10-14 Turn one photo into a five-second video that squishes, melts or explodes — usually in a couple of taps, usually free. You upload a photo, type a description, or hand it a video you already have, and Pika gives back a clip a few seconds long. Adjust and run again if it isn't right — each run spends some of your credits, the platform's unit of spend. What it is: - A short-form AI video generator from Mellis, Inc., on the web and in an iOS app. Pika calls what it makes "social-first content" and its homepage offers "a reality-bending video with just one tap": the tap is real, the reality-bending is five seconds long and often takes three goes. - It is built around Pikaffects, a menu of pre-made effects: instead of describing what should happen you upload a picture and pick "melt", "squish" or "cake-ify". Runway is the stronger craft tool for building shots; Pika is the faster toy with a real free tier. Veo leads on realism and sound but needs a Google subscription; Pika is cheaper and more fun, not more realistic. - Its video model is currently at version 2.5; 2026's releases have been audio tools, agents and a storefront reselling other labs' models rather than a new flagship. Alive and shipping, just not racing on raw quality. Before you start: - Pika's pricing compare table (pika.art/pricing, as of 2026-10-14) lists the Free plan with no monthly credits included (you buy credit packs instead) and no watermark on downloads. - Paid monthly credits reset; unused ones vanish and only purchased top-ups carry over. Starter, Creator and Fancy are monthly plans with growing credit allowances, cheaper billed annually; current prices are on its Launchpad page. - Free is a look, not a workflow — you buy a pack or subscribe before you make anything regularly. Bring an email or Google sign-in and a photo; image-to-video is where Pika is strongest. First step: sign up at pika.art, upload a clear photo of one object on a plain background, open Pikaffects, pick an effect. Security: - Your generations are not fully private, and Pika says so: its FAQ answers whether videos in your account are private with "No, but let us explain." Others see only what you share, but Pika may feature yours on its Templates page under your display name; deleting the video removes it. Security: Faces, likenesses and licensing: - Pika's privacy policy says it collects "Biometric information, such as scans of facial features or voice data" for features driven by a face or voice. A selfie is not a picture of a teacup. - AI Trendmaker and its like are built around your likeness — a selfie plus a sound becomes a video of you. Uploading images or likenesses of other people without consent is explicitly prohibited by Pika's terms: friends, colleagues, celebrities, strangers. - Where law permits or you consent, Pika may use your inputs and outputs to improve its models. It does not claim ownership — you keep them — but you grant a broad licence to use, reproduce and display what you upload while stored. Security: Commercial use — check before you rely on it: - The pricing page currently lists "Commercial use" on every plan, including free Basic. The FAQ contradicts it, limiting that to Pro and Fancy — and is stale elsewhere on the same page, so the pricing page is the better guide. But the terms defer to it, which makes the disagreement matter. - If money depends on the answer, email [email address on the web page] and get it in writing first. Best at: - One-photo party tricks: Pikaffects on a single clear subject — melting, squishing, cake-ifying - Short social clips where the effect is the idea - Swapping one object inside footage you already have, via Pikadditions and Pikaswaps — free at 480p, for a modest credit cost Okay at: - Stitching a first frame to a last frame with Pikaframes, currently around 20–25 seconds Avoid: - Anything needing a long, continuous shot — the default generation is five seconds - Faces, hands and readable text held steady across a clip — these still drift - Real people who have not agreed to it — prohibited by Pika's terms Starter: Melt one object with Pikaffects Uploads a photo of one object on a plain background, opens Pikaffects, picks Melt, and generates. Pikaffects is on the free plan for image-to-video at a low credit cost, your cheapest finished result and no writing at all. Use one clear subject on a plain background — clutter confuses the effect. Starter: Add a camera move to a still photo Pika's own FAQ names the camera words it responds to — bullet time, vertigo, timelapse, dolly down. Unusually direct guidance from a vendor, and a move named on that list is a safer bet than a cinematic adjective the FAQ never mentions. Swap dolly push in for vertigo and run it again. Same photo, completely different feeling. Prompt: Slow dolly push in toward the mug on the table, steam rising, warm morning light, everything else still Starter: Swap one thing inside real footage Pikaswaps modifies one region while preserving the original footage and its sound — a genuinely different skill from generating from scratch. Prompt: Replace the coffee cup on the desk with a small potted cactus, keep everything else exactly the same Pitfalls: - Burning your month on re-rolls. AI video is a slot machine. Running the same idea five times because the third nearly worked is how people hit zero by Tuesday. - Writing like a novelist. Paragraphs of mood and adjectives produce mush. One subject, one action, one camera instruction. Where to next: - More AI video tools: https://www.launchpadhq.io/discover?category=video-creation-editing - Make the images you feed it: https://www.launchpadhq.io/discover?category=image-generation-editing ### Getting Started with Pinecone Canonical: https://www.launchpadhq.io/platform/pinecone/guide Data: https://www.launchpadhq.io/data/v1/guides/pinecone.json Last reviewed: 2026-10-14 Tool: Pinecone (https://www.launchpadhq.io/platform/pinecone): starting price $20/mo flat; free tier yes; data as of 2026-10-14 The searchable memory behind "chat with your own documents" apps — free to start, with one part the docs say you can use in a browser without code. Pinecone's docs describe a console in a browser: you create a store for your material and upload documents into it, then ask a question in a chat box and get an answer with citations pointing at the files it came from. Programs do the same thing without the browser, sending questions in and getting matches back. What it is: - Pinecone is a vector database — a store that finds things by meaning rather than by matching exact words. Ask it about "refund policy" and it surfaces a paragraph on money-back guarantees though not one word lines up. It is the part you never see: when an app answers questions about files you uploaded, something has to hold that material in a searchable form. - It is built for people making things — with one documented exception. Pinecone's docs present its assistant as usable from the console: upload documents, ask questions about them in a chat box, citations included, no code. They describe that route rather than walk through it, so treat the browser-only path as the vendor's claim and expect to confirm it once you are signed in. Everything past it assumes you write software. - Ignore the homepage headline: the company currently leads with Nexus, an enterprise product that installs into a large company's own cloud, and it is not what the free plan gives you. Before you start: - The free Starter plan is a standing allowance, not a trial — Pinecone's docs say it "has no monthly minimum," and it does not expire. It covers a modest amount of stored data, one project and two people. Paid plans step up from Builder, a flat monthly fee, to Standard, which has a monthly usage minimum, to Enterprise, with a much higher minimum — current prices are on its Launchpad page. - It blocks rather than bills. On the flat-fee plans, using up a monthly allowance returns an error and an upgrade prompt instead of an overage charge. - Nothing installs. Everything runs on Pinecone's servers, in a browser at app.pinecone.io — which also means your uploaded material lives on their infrastructure. - If you only want to ask questions about your own files, a consumer AI assistant with file upload gets you there faster. Pinecone earns its place when the material is bigger than that, or when something you are building must search it for other people. Security: - Your documents leave your computer; that is the arrangement with a hosted service. Pinecone's security page says your data "is only used for servicing API calls" — not for training models — and lists SOC 2 Type II and ISO 27001 certification, GDPR readiness, and HIPAA compliance on request. Data is encrypted in transit and at rest. Security: Two things worth knowing: - The exposure is a different shape from a chatbot's: you are not typing one message you might regret, you are uploading a whole body of material, and all of it becomes retrievable by anything holding the key. - An API key is a password that identifies your app — and this one can read, change and delete everything you have stored, not merely spend money. One key per project; delete any you think has been seen. Best at: - Finding relevant passages across a large pile of material by meaning, which keyword search cannot do - Being the memory behind an app other people use — fully managed, so there is no server to keep alive - Turning uploaded documents into something searchable for you, rather than handing you an empty store and the job of filling it Okay at: - Being explored without code — Pinecone documents a console and an assistant playground, but everything past them assumes a developer - Small projects: machinery built for billions of items, pointed at a few hundred documents Avoid: - Running it on your own machine. There is no self-hosted version; the Docker build called Pinecone Local is an emulator the docs call "not suitable for production" that loses everything when it stops. If data must never leave your computer, look elsewhere. - Using it as an AI in its own right — it stores and retrieves, and something else does the thinking - Parking real work on the Starter plan: one project, two users, community-Discord support Starter: Start in the console: make an assistant and upload a few documents In the console at app.pinecone.io, creates an assistant and uploads a few files to it — the no-code route Pinecone's docs describe. Most vector databases hand you an empty store and leave you to turn documents into something searchable. Pinecone's assistant does that step and returns answers with citations naming the file behind each claim — which is why the docs point a non-developer here first. If some part of the console turns out to want code, that is where this page's no-code route ends. Starter caps files at 10 MB each and 1 GB across your whole account, so start with a handful, not an archive. Starter: Ask it something using words your documents never use A plain vector store returns matching fragments and leaves the answering to you; Pinecone's assistant returns an answer with its source attached. Following that citation back to the file is how you check whether the match was really about meaning or just lucky word overlap. Prompt: Using only the documents I uploaded, answer this: [ask about a topic using different words than the documents themselves use]. Say which file each part of the answer came from. Starter: Open the usage page before you build anything on it Finds the usage view in the console and reads how many read and write units your test queries consumed. Pinecone bills read and write units rather than queries, and the docs put a query at roughly one read unit per gigabyte of data it searches. The same query gets more expensive as your collection grows while your query count stays flat — that, not the number of questions you ask, is what decides affordability. Pitfalls: - The ceiling is account-wide, not per index. On Starter, 2 GB is your whole organisation's stored data and 1 GB is the assistant's file storage. - "Serverless" does not mean free. It means you are charged for reads, writes and storage instead of for a machine that runs all the time. - Plan mechanics change in the changelog, not on the pricing page — egress became a metered monthly allowance on 1 September 2026 and the Marketplace was retired on 31 August. Check the release notes before designing around a limit. Where to next: - More AI APIs & developer services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Poe Canonical: https://www.launchpadhq.io/platform/poe/guide Data: https://www.launchpadhq.io/data/v1/guides/poe.json Last reviewed: 2026-10-15 Tool: Poe (https://www.launchpadhq.io/platform/poe): starting price Free, from $4.99/mo; free tier yes; data as of 2026-10-13 One app, many AI chatbots from different companies — paid for with a shared pool of points instead of a separate subscription to each. You pick a bot — Poe's name for each AI you can talk to — type a message, and it answers in the same chat window. Each reply spends points (Poe's unit of spend — each message costs some). Follow up, or ask a different bot the same thing and compare what comes back. What it is: - Poe is a chat app made by Quora that puts AI chatbots from many different companies in one place. Instead of opening one company's app for one AI, you open Poe and choose which one to talk to. Poe's own help pages list bots from OpenAI, Anthropic, Google and Meta, alongside image, video and audio generators. - Each of those AIs runs on a model (the AI "brain" that does the actual thinking), and the model belongs to the company that built it, not to Poe. Poe's terms say plainly that Quora does not create what the bots write. Poe is the front door and the bill; the thinking happens elsewhere. - The reason to use it rather than one company's own app is comparison. Poe's app listing describes comparing answers from different AIs and combining bots "all within a single chat thread," so you can see for yourself which one suits a task instead of taking anyone's word for it. - Anyone can also build a bot of their own on top of an existing one, and other users publish theirs. That is why Poe's bot list is much longer than the handful of big-company AIs — many of the bots you will scroll past were made by individuals, not by the companies whose models they use. Before you start: - You can start without paying. You need a Poe account. - Every bot costs a different amount, and the cost is shown before you send. Each bot's page displays how many points a message takes, and a "Rates" section explains how that price is worked out. A quick text bot and a video generator spend points at very different speeds, so which bots you use decides how far a given balance goes. - Poe sells subscriptions; we have not verified its plans — check poe.com. If you lean on image or video generation, expect to run out of points quickly compared with lighter chat. Check Poe's own site rather than any figure quoted elsewhere. - If you do subscribe, decide where first. Poe's purchases FAQ says you can only upgrade on the same platform you subscribed on, and switching means cancelling and subscribing again. Subscriptions bought in the app stores renew automatically unless cancelled at least 24 hours before the period ends. Security: - Poe sits between you and several other companies, and that is the privacy fact that matters. Its privacy policy (last updated January 2024) says the companies whose models power a bot, and the people who built that bot, may receive the contents of your chats. It says your account details, such as your username and email, are not passed to them. Conversations are not visible to other Poe users unless you share them. Security: Worth knowing before you type anything personal: - Bots built by outside developers are a different case from the big-company ones. Poe's privacy policy says developers of bots built with code 'may view and store your anonymized chats on their servers to train their models.' Keep sensitive questions to the official bots. - Sharing a chat publishes it. Choosing messages to share copies them into a post other Poe users can see; delete it from your profile if you change your mind. - Memory is off by default. If you turn it on, Poe builds up context from your chats and shares that context with the other model companies you talk to on Poe. You can view or delete it in Settings. - Temporary chats are not saved to your history and their messages are deleted within 24 hours, but files you upload to one are kept for up to 15 days for safety review. Security: Never paste in: - Passwords, card numbers or Social Security numbers — Poe's own privacy policy asks you not to - Other people's private information - Confidential work material Best at: - Putting the same question to AIs from different companies and seeing the difference side by side, without an account at each one - Finding out which AI you actually prefer before paying any single company for it - Turning a set of instructions you keep retyping into your own reusable bot, with no code Okay at: - Image, video and audio generation — available, but these bots cost far more points per use than text bots - Long, ongoing conversations — they work, but Poe sends chat history along with each new message, so a long thread costs more per reply than a fresh one Avoid: - Assuming a community-made bot costs what an official one does — creators can set their own price per message, and Poe's per-message budget setting only warns you about those bots after the fact - Relying on any bot for medical, legal or investment decisions — Poe's own FAQ says to consult a professional Starter: Ask two companies' AIs the same thing Send it to one bot, then send it again to a bot from a different company, in the same chat or a new one. Every bot's page shows its points cost, so you see both how the answers differ and what each one cost. Swap in any topic you are trying to learn. The differences are usually clearest on explanations, not on simple facts. Prompt: I want to understand how a credit score is calculated. Explain it in under 150 words for someone who has never had a credit card, and end with the single most common misunderstanding. Starter: Check the price tag before and after Open a bot's page and look at its Rates section, send one short message, then open that message's info menu to see what it actually cost. Settings lets you view your points history. Poe charges per message from one shared balance, and the cost differs bot to bot — which is not how a single-company chatbot like ChatGPT works, where you never see a per-message price. Two minutes of checking tells you which bots a given balance can comfortably cover. Starter: Build your own bot in a form, no code Click 'Create a bot' in the sidebar, choose a base bot (the existing AI it runs on) and paste this into the prompt field. Unlike ChatGPT's custom versions, the base can come from any company Poe carries, so you can rebuild the same coach on a different AI and compare. A setting called 'Prompt visible from bot profile' decides whether other people can read your instructions. Add a greeting message so you remember what the bot is for when you come back to it. Prompt: You are a patient writing coach. When I paste a paragraph, reply with: one sentence on what works, the two most important fixes, and a rewritten version. Keep it friendly and never longer than my original. Starter: Start a fresh chat for each new task Poe has a setting called auto-manage context that limits how much earlier chat history is sent with each message, because sending history costs points. Starting a new chat for a new job keeps each reply cheap and keeps the AI from mixing in an unrelated earlier conversation. Prompt: Here is a long email I need to answer: '[paste]'. Summarise what the sender is actually asking for in three bullet points, then draft a reply under 120 words. Pitfalls: - A long chat gets more expensive as it goes. Turning off auto-manage context lets the bot see more of the conversation but spends more points per reply. - In a group chat, the person who calls a bot pays for it. Bots only reply when mentioned, unless automatic replies are switched on, and the points come from whoever triggered the reply. - It sounds equally sure of itself whichever AI you pick. Comparing two answers helps you spot disagreement, but two bots agreeing is not the same as being right — check anything that matters. Where to next: - Chat assistants: https://www.launchpadhq.io/discover?category=text-conversational-ai - Image generators: https://www.launchpadhq.io/discover?category=image-generation-editing - Research tools: https://www.launchpadhq.io/discover?category=research-academic-tools ### Getting Started with QuillBot Canonical: https://www.launchpadhq.io/platform/quillbot/guide Data: https://www.launchpadhq.io/data/v1/guides/quillbot.json Last reviewed: 2026-10-15 Tool: QuillBot (https://www.launchpadhq.io/platform/quillbot): starting price $8.33/mo; free tier yes; data as of 2026-10-14 Reword, tidy and check writing you already have — with honest limits on what its AI detector can tell you. You paste your own writing into the box on the left, pick a tool, such as one that rewords it or checks its grammar, and click its button. A reworked version appears alongside it. You keep the changes you like, adjust a setting, and run it again until it says what you meant and still sounds like you. What it is: - QuillBot is a collection of writing tools on one website. Its best known is the Paraphraser, which rewords text you give it; around it sit a grammar checker, a summarizer, a translator, a citation generator, a plagiarism checker, an AI detector (a tool that estimates how likely it is that a piece of text was written by AI) and a "Humanizer" that smooths stiff, machine-sounding drafts. Unlike a chatbot, almost everything here starts from text you already have rather than a blank page. - It is aimed squarely at students and people who write for work. QuillBot is a business of Learneo, Inc., and it follows you out of its own website: there are add-ons for the Chrome, Edge and Safari browsers and for Microsoft Word, plus apps for iOS, Android, Windows and macOS. - The menu has grown well past rewording — it now lists an AI Chat (marked "Beta"), image tools and a shelf of PDF utilities. None of those is why people come here, and you can ignore them on day one. The rewording and checking tools are the product. Before you start: - There is a free plan, and QuillBot's own grammar-checker and paraphraser pages say you can start without signing up. The plans page lists word limits on free paraphrasing and summarizing. Several modes and tools inside the site carry a Premium label, so you will meet those as you click around rather than all at once. - If you do pay, Premium is currently listed at $8.33/mo (billed annually), meaning you are charged for the year up front. QuillBot's plans page does not publish a month-to-month price. Its FAQ says you can pause a subscription from your Account page instead of cancelling it, and you are not billed while it is paused. - Whether you are allowed to use it is your school's or employer's call, not QuillBot's. QuillBot's own Humanizer page tells you to disclose AI use according to your institution's or employer's policies, and its help centre says that rewording a source without citing it can still count as plagiarism. Check the rule that applies to you before you hand anything in. - The browser add-on works on nearly all websites without signing in, according to QuillBot's help centre. Google Docs is the exception: there you have to be signed in to a QuillBot account before it switches on. Security: - Everything you paste into QuillBot is sent to its servers to be reworked — it is a website, not a program running privately on your computer. The browser add-on goes further than the website, because it works inside the text boxes of the sites you write on, including your documents once you sign in. QuillBot's Trust Center says it uses the text you paste for training to improve its services, that storing your text is switched on by default for everyone, and that it does not sell your data. The switches to stop it storing your text or training on it are documented for the browser add-on (version 4.37.0 or later), under Extension Settings → View All Settings → Data and Privacy. Security: Before you paste anything unpublished:: - Read QuillBot's Trust Center, specifically its answer to "How does QuillBot use my data?" (quillbot.com/trust-center). It explains what happens to the text you paste and which settings you have — read it yourself rather than assuming. - Don't paste other people's unpublished work — a friend's essay, a colleague's draft, a client's document — for the same reason you'd ask before emailing it to a stranger. - If you install the add-on, it is easy to forget it is there. Turn it off on sites where you write anything private; the switch is in the QuillBot pop-up in your browser toolbar. Best at: - Rewording a sentence of your own that says the right thing in a clumsy way — and showing you several versions to choose between. - Catching grammar, spelling and punctuation slips in about six languages, not only English — QuillBot's grammar page names English, Spanish, German, Dutch, French and Portuguese. - Turning a long article into a short paragraph or a bullet list, with a slider for how short. - Producing a formatted citation in APA, MLA or Chicago style for a source you've used. Okay at: - Its AI detector. QuillBot itself calls the result "probability estimates, not verdicts", says "no AI Detector can provide 100% accuracy", and notes that heavily reworded or lightly AI-edited text is harder to catch. - Changing a lot of words at once. The more of your text you let it swap, the more likely the meaning drifts — see the first starter below. Avoid: - Treating a detector percentage as proof about someone else's writing. QuillBot's detector page leads with a "99% detection rate, according to independent evaluations" — and the same page says "Never make an academic or employment decision on a detector score alone." Believe the second sentence. - Running a draft back and forth between the Humanizer and the AI Detector until the number drops. QuillBot sells both, and its own Humanizer page says the score is "not something to optimize against" and that the Humanizer "isn't a tool for evading AI detection." - Rewording someone else's work to make it count as yours. QuillBot's help centre says QuillBot "is not designed to bypass plagiarism detection systems such as Turnitin." Starter: Reword a paragraph, then move the Synonym Slider Paste a short paragraph you wrote yourself — under 100 words keeps you inside the free limit — click Paraphrase, then drag the Synonym Slider above the text boxes and watch the rewrite change. QuillBot's help centre describes the slider as four levels: 1 makes the fewest word swaps and stays most accurate, 4 makes the most and is the least accurate, and it starts at 2. It re-runs the moment you move it, so you can see exactly where your meaning starts to slip — a dial Wordtune's rewrite suggestions don't give you. The slider isn't available in every mode. If it disappears, switch back to the Standard mode. Starter: Paraphrase a source, then cite it Take one sentence from an article you're using, reword it in the Paraphraser, then open the Citation Generator and make an APA, MLA or Chicago citation for that article to put next to it. QuillBot keeps the Paraphraser and a Citation Generator under the same menu, so the citation the rule above calls for is one click away. Doing both together, every time, builds the habit that keeps paraphrasing on the right side of your school's rules. Starter: Check your grammar, one suggestion at a time Paste an email or paragraph into the Grammar Checker, click Check grammar, and go through each highlighted suggestion — accepting or rejecting it — instead of fixing everything at once. QuillBot highlights grammar, spelling and punctuation problems for you to review, and checks the six languages listed above — useful if you write in more than English. Reading each change is how you learn the mistake rather than just losing it. Try it on something you wrote in a second language. Starter: Test the AI detector on writing you know is yours Paste a few paragraphs of something you definitely wrote yourself, without AI, into the AI Detector and read both the overall percentage and the notes on each section. The detector sorts text into three kinds — written by AI, written by a person, or written by a person and refined with AI — rather than just yes or no. QuillBot also says that when a result is unclear its detector leans toward calling text human. Seeing what it says about writing you know is human tells you how much any score it gives is worth. Try a very formal piece of your own, like a cover letter. QuillBot's own page warns that formulaic human writing can score higher than expected. Starter: Shrink a long article into bullet points Paste a news article or report into the Summarizer, choose Bullet Points instead of Paragraph, and move the length slider to see the summary grow and shrink. QuillBot's Summarizer offers exactly two built-in formats — Paragraph and Bullet Points — and a four-step length slider. Being able to set the shape and length with a click, rather than asking a chatbot like ChatGPT and hoping it obeys, makes it easy to compare a 3-line and a 10-line version of the same text. Pitfalls: - Assuming the reworded version means the same thing. Read every output against your original before you use it, with an eye on any name, number or term that has to stay exact. - Panicking when your own writing scores as 'AI'. As the fourth starter shows, a high score on something you wrote yourself is a known limit of the tool, not an accusation. - A grayed-out QuillBot icon in Google Docs. QuillBot's help centre lists the usual causes: you're not signed in through the add-on, the document is in an older format (try a new document), it's set to Pageless instead of page mode (File → Page setup), or the add-on is switched off in its pop-up. Where to next: - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai - Research tools: https://www.launchpadhq.io/discover?category=research-academic-tools ### Getting Started with Quizlet's AI Tutor Canonical: https://www.launchpadhq.io/platform/quizlet-q-chat/guide Data: https://www.launchpadhq.io/data/v1/guides/quizlet-q-chat.json Last reviewed: 2026-09-12 Tool: Quizlet Q-Chat (https://www.launchpadhq.io/platform/quizlet-q-chat): starting price Free, from $35.99/year; free tier yes; data as of not dated Quizlet retired Q-Chat. Here's the AI study chat that's inside Quizlet now — what it does, and who can actually use it. You open the chat from your Quizlet home page or from a set you're already studying, type a question about whatever you're stuck on — or hand it a photo of your notes — and it answers in the window. You keep asking follow-ups, and it can turn what it just explained into flashcards. What it is: - Q-Chat was Quizlet's AI tutor, launched in March 2023 and built on the technology behind ChatGPT. It is no longer available: Quizlet's help centre has no page for it in any language, and independent write-ups put its removal around June 2025. Quizlet does not appear to have announced the retirement anywhere public. - What Quizlet has instead is Ask Quizlet — a chat window inside Quizlet that explains things from your coursework, points you at study sets, and turns its own explanations into flashcards you can study. We have not seen Quizlet describe it as a renamed Q-Chat, so treat it as the current tool rather than the old one under a new label. - The reason to use it rather than a general chatbot is where it sits: it opens on top of the set you're revising, in the account you already study in. Before you start: - Start here, because it decides everything else: Quizlet's help centre says Ask Quizlet "is currently only available for users in the United States who are 14 years old or older." If you're outside the US or under 14, nothing on this page will appear for you, and the honest move is a general chatbot instead — start with our Text & Conversational AI category. - You need a Quizlet account and to be signed in, but you don't need to pay. Quizlet's help page for Ask Quizlet names no subscription requirement and no usage cap. That's worth a pinch of salt rather than a guarantee: Quizlet does cap its other AI features by plan — its study-guide generator is "limited access" on a free account and unlimited on a paid one — so a cap on the chat may exist without being written down. Check for a limit once you're signed in. Paid Quizlet Plus is a yearly subscription (also sold monthly as an in-app purchase on Apple devices), and what it buys is more of the other AI tools, not obviously more chat. - You can type a question, or attach your own material and add a prompt (the message you type) saying what to do with it — Quizlet says up to three files at a time, in .pptx, .docx, .pdf, .png, .jpg or .jpeg, and under 25 MB each if you're on a phone. - The conversation is not kept. Quizlet says plainly: "Chat history isn't saved across pages. If you leave Ask Quizlet, the conversation will reset, but minimizing it will keep your chat active." So if it explains something well, copy it out or turn it into flashcards before you navigate away. Security: - Two things we could not establish from Quizlet's own public pages — its privacy policy would not load for us — are which company's model (the AI "brain" that does the actual thinking) is answering you, and whether your conversations are used to improve it. Quizlet has not published the equivalent detail for this one that we could find. Security: Worth a think before you upload: - A classmate's graded work, a photo of someone else's paper, anything with other people's names on it — that's their information going onto a company's servers, not yours. - If you've joined a teacher's class, Quizlet gives teachers reporting on how students study their sets. Quizlet's help pages describe that for study activities and don't say one way or the other whether chats are included, so don't assume the chat is invisible to a teacher just because it looks like a private window. Best at: - Getting unstuck on one specific thing while you're already inside the set you're revising, without opening another tab - Answering from your own material rather than from the open web — you hand it the notes and ask about those Okay at: - Sustained tutoring. It's shaped around explaining a study item and making cards from it, so it suits one sitting rather than a plan across a week. - Work that isn't flashcard-shaped — essay feedback, code, open-ended discussion. Quizlet's own description of it doesn't reach for any of that. Avoid: - Generating the answer you hand in. Quizlet's community rules are explicit — "Quizlet is meant for learning, not cheating" — and Quizlet says it can remove public access to sets and send warnings or take other administrative action against the people who made them. - Treating its explanation as checked. Quizlet's own help page for Ask Quizlet says nothing at all about accuracy, verifying answers, or double-checking what it tells you — no caveat, in either direction. An absent warning is not a promise that it's right. - Counting on it for anything current or off-Quizlet. Web searching is not among the things Quizlet says it does, so don't lean on it for this week's news or facts you can't check. Starter: The card you keep getting wrong Ask Quizlet opens on top of the set page itself, so the term and the definition you're arguing with are right there to copy — and the explanation it gives back can become new cards without you leaving the page. Paste the same thing into ChatGPT and you get a good answer in a window that knows nothing about the set you're revising. Add "and give me a way to remember it" — a mnemonic is the part you'll actually still have on the day. Prompt: I keep getting this one wrong: [paste the term and its definition from the set]. Explain it in plain language, give me one everyday example, and tell me the thing it's easiest to confuse it with. Starter: Hand it your notes instead of typing the question Quizlet says that when you upload, it will "respond based on your study materials" — so the answer is pinned to what your course actually covered rather than to the general version of the topic. Attach the photo of a whiteboard or a page of handwriting rather than retyping it — images are on the accepted list. Prompt: These are my lecture notes. Pull out the ten things most likely to come up in a test, then explain the two I'd be most likely to get wrong. [attach the file or a photo first] Starter: Check a set before you trust it Judging whether a set fits what you're studying, and pointing you at others, is one of the jobs Quizlet lists for this tool — and it can look at the set you're on. A general chatbot can only answer this if you paste all eighty terms into it first, which nobody does. Follow up with "make cards for what's missing" and you've closed the gap instead of just learning about it. Prompt: I'm revising [topic] for [the exam or test]. Does this set actually cover what I need, and what's missing from it? Starter: Make it test you, in one sitting Quizlet's Learn and Test modes check what you typed against the wording on the card; a chat lets you answer in your own words and see what it makes of that. Say "ask me the ones I get wrong again at the end" to get a second pass at your weak spots. Prompt: Quiz me on this set one question at a time. Wait for my answer before you say whether it's right, keep a running score, and stop when I say stop. Pitfalls: - Everything you find about Q-Chat is out of date, including Quizlet's own page. Quizlet's March 2023 announcement, "Introducing Q-Chat, the world's first AI tutor built with OpenAI's ChatGPT," is still live on its blog with no retirement notice on it. If a guide, a video or a teacher's handout tells you to click Q-Chat, it was written before the tool went away. - Blaming yourself when the chat isn't there. If no chat appears on your home page or set page, the likeliest reason is the eligibility limit above, not your account, your browser or your plan. - Assuming Quizlet's AI features share one rulebook. They don't: its study-guide generator is available to 14-year-olds in the US and UK, 15 in Australia, 16 in Canada, Ireland and New Zealand and 18 in a long list of other countries, and works in English, French or German — a far wider gate than the chat's US-only one. - Letting it write flashcards you never read properly. A wrong or half-right explanation doesn't just get read and forgotten — it gets made into a card and then memorised on purpose. Read what it generated before you study it. Where to next: - Education & Learning AI: https://www.launchpadhq.io/discover?category=education-learning-ai - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai - Research & Academic Tools: https://www.launchpadhq.io/discover?category=research-academic-tools ### Getting Started with Qwen Canonical: https://www.launchpadhq.io/platform/qwen/guide Data: https://www.launchpadhq.io/data/v1/guides/qwen.json Last reviewed: 2026-10-15 Tool: Qwen (https://www.launchpadhq.io/platform/qwen): starting price $0.05/$0.40 per 1M tokens; free tier yes; data as of 2026-10-14 Alibaba's free AI chat — and a first look at a model family you can later download and run yourself. You go to chat.qwen.ai, sign in, and type a message — your prompt (the message you type) — into the box. Qwen types an answer back, sometimes after searching the web and listing the pages it used underneath. You reply to steer it: shorter, simpler, more detail, try again. What it is: - Qwen is Alibaba's family of AI models (the AI "brain" that does the actual thinking), and Qwen Chat is the free website and app where you talk to them. Qwen's site credits the service as "Powered by Alibaba Cloud." - Qwen comes through three doors, and this page covers the first. Qwen Chat is the website at chat.qwen.ai plus apps for iOS, Android, macOS and Windows. Downloadable models are the second: Alibaba publishes hundreds of Qwen models as open-weight files (the model file itself is published, so you can run it yourself) on Hugging Face, a public site for sharing AI models. Run one with Ollama or LM Studio and your messages never leave your computer — see Local & Open-Source AI below. Each model carries its own licence, and they are not all the same, so read the one on the model's page before relying on it. The third door is for programmers building Qwen into their own software: Alibaba's Qwen Cloud, launched for global markets in May 2026, and Alibaba Cloud Model Studio. - Why start here rather than with ChatGPT or Gemini: Qwen describes its chat as one place to reach both its open-source and its proprietary models, alongside web search, reading documents and pictures you upload, making and editing images, and voice and video chat. What you learn here carries straight over if you later want to run a Qwen model privately on your own machine — something you cannot do with the models behind ChatGPT, Claude or Gemini's chat apps. Before you start: - Go to chat.qwen.ai (or start from qwen.ai, which links to it) and create an account — Qwen's rules say you must be signed in to use it. Its privacy policy describes signing up with an email address or with a Google, Apple or GitHub account. Qwen describes the chat as free to use and open to all. - It is licensed for personal use, not work. Qwen's service agreement (effective 10 September 2026) says the service is "for private and non-commercial use only." Learning, home projects and personal writing are fine; drafting for your employer or for paying clients falls outside that licence. Businesses are pointed at Alibaba's developer platforms instead. - The agreement sets the minimum age at 18 (or older, if your country requires it), and says features "may also vary among jurisdictions" — so a button a guide mentions may simply not exist where you live. Security: - Qwen Chat, the Qwen apps and the desktop client are run by NTH Power Global Tech Singapore Pte. Ltd., a company registered in Singapore. Its privacy policy (effective 7 September 2026) says your account details and conversations are collected, processed and stored in Singapore, on Alibaba Cloud's Singapore servers, and warns that Singapore's data-protection laws 'may differ from those in your jurisdiction.' The service agreement is governed by Singapore law, with disputes settled by arbitration there unless your country's consumer law says otherwise. Security: What Qwen's own documents say about training on your chats: - The privacy policy says Qwen may use your conversations to train its models after stripping out details that identify you, and that it does not train directly on your raw input. - The service agreement is broader — it may use your content 'to train models' — and gives one way out: email [email address on the web page] and ask to opt out. Neither document mentions a switch in the app's settings. - Even after you opt out, rating an answer with thumbs up or down saves that whole conversation as feedback, and the agreement says feedback is still used for training. Skip the rating buttons on anything private. Security: Qwen's own list of things not to upload: - Passwords, keys or login details of any kind — the agreement has you promise your uploads contain none - Health records, ID cards and financial account information - Your precise location, or biometric data such as face or fingerprint scans - Other people's personal details, unless they've agreed — the policy makes that your responsibility Security: - You can delete your account yourself from Qwen's settings, per the privacy policy. The same policy says some data may be kept after deletion where the law or a pending dispute requires it, and it publishes no fixed retention periods. Best at: - Trying Qwen's models for free before deciding whether to download one and run it privately on your own computer. - Questions about recent events, with the pages it searched listed under the answer so you can check them. - Summarising and questioning documents and pictures you upload. - Everyday drafting, rewriting and explaining — the core chat job. Okay at: - Search-backed answers: useful, but Qwen's own agreement says replies 'may still contain errors even when reference sources are identified' — open the sources. - Image making and voice or video chat, which Qwen lists among its features; Qwen does not publish how much of them you can use. Avoid: - Anything you are paid to produce — the personal, non-commercial licence covered in Before you start. - Scripting it or bulk-copying its answers with another tool. The agreement bans 'extracting data or outputting content through automated or programmatic means'; the developer platforms exist for that. - Letting its answer decide something about another person — a loan, a job, a tenancy, a diagnosis — without a human checking it. Qwen's agreement forbids exactly that use. Starter: Ask about something recent — and make it show its sources Qwen's service agreement says it searches the web either when it guesses you want it to or when you switch its search function on yourself, and then lists the pages it drew on beneath the answer. Saying 'search the web' takes the guessing out, and asking for dates matters because the same agreement admits answers can be wrong even with sources attached. If no sources appear under the reply, it answered from what it already knew — ask again with search switched on. Prompt: Search the web: what has changed about [a rule, product or event you care about] in the last month? List each source you used with its date, and tell me which points only one source supports. Starter: Hand it a document you're allowed to share Qwen's agreement has you promise that anything you upload is yours or that you're authorised to share it. Practising on a public document — a product manual, a council notice, a blank lease template — lets you learn how it handles files without breaking that promise on day one. Follow up with 'Which of your bullet points is least certain, and why?' Prompt: Summarise the attached document in five bullet points a beginner could follow, then list three questions it leaves unanswered. Starter: Find out what it remembers about you Qwen's privacy policy names a Memory feature that keeps 'specific details from your conversations' to shape later ones, but Qwen publishes no help page on how to see or clear what it has kept — unlike ChatGPT, which documents its memory controls. Asking it directly in a fresh chat is the dependable way to learn what it carries forward. Open a new chat and ask: 'What do you remember about me?' Tell it to forget anything you'd rather it didn't keep, then ask again to check. Prompt: Remember this for future chats: I cook for two, I'm vegetarian, and I have 30 minutes on weeknights. Suggest three dinners for this week. Starter: Test-drive a model you could later download If Qwen Chat offers a choice of model, pick one Qwen also publishes as a free download, and ask it the kinds of questions you'd want answered offline — then compare it with the default. Qwen describes its chat as one place to reach both its open-source and proprietary models. That turns the chat into a free test drive: you find out whether a model is good enough before spending the disk space and setup time to run it in Ollama. Whether a model chooser appears for you, and which models it lists, is something you'll see once you're signed in. Pitfalls: - Lookalike sites and apps are everywhere. Search 'Qwen pricing' and the top results include unofficial sites with Qwen-style names quoting their own prices. Start from qwen.ai or chat.qwen.ai and follow their links to the apps rather than a search result. - 'Qwen' means two things. It's the chat app and it's the family of downloadable models. A guide about 'running Qwen' or a per-use price is almost always about the models or the developer platform, not the free chat this page covers. Where to next: - Local & Open-Source AI: https://www.launchpadhq.io/discover?category=local-open-source-ai - AI APIs & Developer Services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Ramp Intelligence Canonical: https://www.launchpadhq.io/platform/ramp-intelligence/guide Data: https://www.launchpadhq.io/data/v1/guides/ramp-intelligence.json Last reviewed: 2026-09-12 Tool: Ramp Intelligence (https://www.launchpadhq.io/platform/ramp-intelligence): starting price Free, from $12/user/mo; free tier yes; data as of not dated AI that works on your company's card spending — receipts, policy questions, and whether you're overpaying for software. Your company's card spending arrives in Ramp on its own. You text a photo of a receipt, or type a question into a chat box, and an answer comes back in plain English. Most of it happens while you aren't looking; your part is checking what it got wrong. What it is: - Ramp is a corporate card and spend-management platform — cards, bills, reimbursements and travel bookings in one place. Ramp Intelligence is the name it has given the AI layer running across all of that since May 2023. - Most of it isn't something you open. It reads receipts and invoices, checks transactions for fraud, suggests how each expense should be categorised, and answers "can I expense this?" over text message. Ramp increasingly calls these pieces agents (AI that takes actions on its own rather than only answering) rather than features. - Ramp Intelligence is a reason to notice Ramp, not a reason to move your company's money. Before you start: - This is not something an individual can sign up for. Ramp is "not accepting individuals, sole proprietors, and other types of unregistered businesses." If that isn't you, a general assistant like ChatGPT is the substitute — it just won't sit on your company's cards. - Outside the US we could not establish whether there is a path at all. The help article we read is titled for US-based applicants, implying a non-US counterpart, and Ramp publishes an international number for texting receipts — yet the qualification text says the business must be registered in the United States. We found no vendor page describing a Canadian, UK or EU entity route, so treat the US-only line as what Ramp's own document says rather than a settled answer. - You are applying for a corporate charge card, not a subscription you cancel: 30-day statement cycles, the full balance due at the end of each one and debited from the bank account you linked, no carried balance and, Ramp says, no interest and no personal guarantee. - The base plan is genuinely free per user, and everything below on this page sits inside it. The accounting half is where the money is — AI coding of fields, approval recommendations, AI-driven expense reviews and building reports by describing them are all marked excluded from the free plan on Ramp's own comparison table and included on Plus, priced per user per month plus a platform fee that scales with team size, with a 30-day trial and a discount for annual billing. - Then apply at ramp.com: formation details, a corporate officer's address and the last four digits of their Social Security number, and either a live connection to the business bank account or its three most recent statements as PDFs. That is an approval decision, not a signup, so expect a yes or a no at the end of it. Security: - Unlike a chatbot, Ramp holds your company's actual money. The login to protect hardest is whichever one carries the admin role — that is the account that can issue cards and move money. Security: What Ramp says about your data — and what we could not check: - Ramp's own engineering post of 13 June 2024 says it does not train models end-to-end on private customer data, that its general models (the AI "brain" that does the actual thinking) learn from "aggregated and masked" customer data by understanding its structure rather than its contents, and that contract data improves Ramp's models "only if they explicitly opt in." - That is the company describing itself, not the contract. Ramp's binding terms and privacy policy live at ramp.com/legal, which does not render without JavaScript, so we could not read the version in force today or confirm it says the same thing. If your business has rules about AI and vendor data, read those pages yourself once signed in. - Ramp's security page says it undergoes annual SOC 2 Type II audits (an audit of how a company handles customer data), abides by PCI standards for card data, and encrypts data at rest and in transit. - Receipts and memos you text travel over ordinary SMS to a published shortcode. Whatever is visible in that photo has gone by text message. - Uploading a vendor contract for a price check hands Ramp the contract — its help centre says it extracts amounts paid, products purchased and payment terms. Check whether yours carry confidentiality clauses first. - Ramp's documentation says assistant conversations are visible only to the person who started them. Best at: - Expense admin by text message — send the receipt photo, Ramp matches the transaction - Answering "can I expense this?" from your own written policy, in Slack, by text, or in Microsoft Teams - Spotting software you overpay for — Ramp claims a benchmark base of over 70,000 businesses Okay at: - Open-ended analysis — Ramp scopes the assistant to its products, your account and acting on your behalf, not "why did travel spend jump 40% in August?", and we could not test it without an account - Newness — the 29 April 2026 procurement agents include "Zero-Touch Sourcing", marked Early Access, so what you can switch on may be narrower than the marketing - Starting from nothing — the expense-policy agent works from a policy document you upload, and offers to draft one if you have none Avoid: - Reading ramp.com/intelligence as a list of what you get — it names a dozen AI features, says "Get started for free" and attaches no plan labels, while Ramp's own pricing comparison marks the accounting-side AI as excluded from free. The two vendor pages disagree; believe the one with columns. - Treating a policy answer as an approval — it answers from whatever policy document was loaded, however stale Starter: Ask Ramp what you owe it Ramp's plan table leaves one AI row ticked in the free column: "Instant, 24/7 AI help for any Ramp question or task". Two hedges. The pricing page never uses the name "Ask Ramp", so equating that row with the assistant its help centre documents is our inference; and that help centre puts the assistant at app.ramp.com/chat behind a left-sidebar icon we could not see without an account. It can act on your behalf as well as answer — see the pitfalls — so phrase your first few messages as questions. Prompt: I'm new to Ramp. Walk me through what I need to do this week to keep my expenses in good standing: what's outstanding on my account, which transactions are missing a receipt, and what I should do first. Starter: Text the policy question everyone is embarrassed to ask Text it to HIRAMP (447-267), the shortcode Ramp publishes in its help centre, with 844-331-1023 for international users. Navan and Brex route this through their apps; a published SMS shortcode on the free plan is Ramp's. The reply comes from your own expense policy, so it is as current as that document is. The same conversation works in Slack (type /ramp, press enter) and Microsoft Teams — though Ramp's docs say its full assistant is *not* available inside Slack's AI chat, so Slack gets the expense bot, not the whole thing. Prompt: What does our policy say, and what do you need from me? Starter: Send a receipt with the memo already written Text that line in the same message as the receipt photo and Ramp attaches it as the memo. Its help centre lists the memo alongside merchant, amount, department, receipt content and your past coding as the inputs it uses to predict the category — so eight words change where the expense lands. By email the same text needs the prefix memo: in the body; by text it does not. And a texted memo attaches to your *latest* transaction — see the pitfalls. Prompt: Client dinner with Acme — 2 attendees, Q4 renewal discussion Starter: Find out if you are overpaying for a piece of software Go to Insights → Price Intelligence and upload a contract for a vendor you already pay. Ramp compares it against anonymised spending data from other Ramp customers and shows an average price and how many businesses contributed to it. That is where the documentation puts the screen; we could not confirm it without an account. This works only because of what a card company can see: Ramp benchmarks against transactions flowing through its own cards, which a tool that does not issue the card, like Navan, has no way to build. Start with the subscription you are least sure about rather than the biggest one — the benchmark only helps where enough other businesses buy the same product. Starter: Connect your work inbox and stop chasing receipts Connect your Gmail or Outlook work account and Ramp collects emailed receipts on its own and attaches them to matching transactions. Ramp's documentation flags the limitation that makes this worth understanding rather than just switching on: the email connection does not support submitting memos or accounting fields, only the receipt itself. Do this one first if most of what you buy is software and subscriptions, where the receipt arrives by email. Pitfalls: - A texted memo lands on your latest transaction. Buy lunch, buy petrol, then text the lunch memo, and it goes on the petrol. - The assistant can act, not just answer. Ramp's documentation says it can approve requests and update transactions on your behalf, so a question phrased as an instruction changes your account while you thought you were asking about it. - Price Intelligence may simply not be there. It is on by default for Admins, Owners and financial roles; enabling it for custom roles requires an admin on Plus. If you cannot find it in Insights, that is a permission, not a bug. - Features can exist and still be switched off. Ramp's July 2025 agents launch told customers to find an "Early Access tab and toggle it on", so "Ramp has this" and "your account has this" are different claims. That launch also put the expense-policy agent behind Plus, while the current pricing table lists policy questions by SMS or Slack as free — we read the table as current, but if you are on the free plan and cannot find it, that history is the likeliest reason. - Slack is two steps, not one. Ramp's instructions are to type /ramp, press enter, and attach the receipt picture to the *next* message. Where to next: - Finance & Real Estate AI: https://www.launchpadhq.io/discover?category=finance-real-estate-ai - AI Plugins & Business Software: https://www.launchpadhq.io/discover?category=ai-plugins-business-software - Workflow & Automation: https://www.launchpadhq.io/discover?category=workflow-automation ### Getting Started with Read AI Canonical: https://www.launchpadhq.io/platform/read-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/read-ai.json Last reviewed: 2026-10-24 Tool: Read AI (https://www.launchpadhq.io/platform/read-ai): starting price Free, from $15; free tier yes; data as of 2026-10-26 An AI note-taker that joins your Zoom, Teams and Google Meet calls, writes up summaries and action items, and can search across your meetings, email and chat. You sign in and connect your calendar. Read's note-taker then joins your video meetings as a participant (on Google Meet it can also record without joining) and produces a report: a summary, topics, action items, key questions and a transcript. You can also ask Ask Read questions across your meetings and, if you connect them, your email and chat. What it is: - Read AI is a meeting assistant. Its help centre says it supports Zoom, Microsoft Teams and Google Meet: it joins Zoom and Teams as a participant named "read.ai meeting notes" by default, and on Meet it can join or use botless recording. For other platforms, or in-person meetings, there are desktop and iPhone apps that record locally. - Each meeting gets a report with a summary, topics, action items, key questions and a transcript. Read also measures how a meeting went: its Read Score combines sentiment and engagement, which its help centre says is measured from facial and verbal cues outside the EU and UK. You also get private speaker coaching on things like pace and filler words. - Ask Read is its search assistant. Its help centre says it can search your meetings, the web, and — if you connect them — Gmail or Outlook, your calendar, and Slack or Teams. Connecting email and chat also powers Readouts, summaries of your inbox and messages. Before you start: - Sign-up sets up auto-join. You can sign in with Google, Microsoft, Zoom, Slack or email. Read's step-by-step setup guide says the default is to join all meetings on your calendar and to send recaps to everyone with access — its own help centre warns people skim this step and are surprised when Read starts joining every meeting. - Free is small. Read's free plan caps how many meeting transcripts you get each month and how long a meeting can run. Its help centre adds that reports other people share with you count toward that allowance, and that Readouts on Free last only 14 days. New users start on a free trial of a paid plan, then drop to Free. - Paid plans are per user. Pro is $15/user/month billed annually, more if you pay month to month, and the two Enterprise plans cost more per user. The pricing page also says Read doesn't offer refunds. Getting set up safely (Sign up at read.ai. If you use Zoom, connect the Zoom integration from inside Read — its help centre says Read can now only join Zoom meetings when you're there, signed in with that connected account.): - Change auto-join during setup. Choose "only meetings you host" or "only meetings you invite it to" instead of all meetings. You can change it later under Account Settings → Meeting Recording → Auto-Join Preferences. - Send recaps only to yourself at first. Read's help centre says report sharing goes to everyone on the calendar invite, whether or not they attended. Change it under Account Settings → Report Sharing. - Using a Gmail address? Read decides who's internal by email domain, so with a personal address everyone counts as external and the "internal only" options won't work. - Confirm it works. From the Calendar page, turn on "Add Read?" for one upcoming meeting and check that a report appears afterwards. Security: - Read records other people, and recording-consent laws vary by place, so tell participants. In a meeting, anyone can type "Opt Out" in the chat: Read's help centre says it then leaves and deletes everything, with no report created. When you record locally with the desktop or phone app there's no bot, and getting consent is entirely your responsibility. Security: Worth knowing before you connect everything:: - Sentiment and engagement scores analyse faces and voices outside the EU and UK. They can feel intrusive in HR or client meetings; Read's help centre says any user can turn these metrics off in Account Settings. - AI training is opt-in, per Read's privacy page, and its privacy policy says you're not opted into its Customer Experience Program by default. - Email and chat access is optional. Read says it doesn't read your email unless you connect an email integration, and you can revoke access from Integrations. - Read states it is SOC 2 Type II audited and GDPR compliant; a HIPAA agreement is only available on annual Enterprise+. Best at: - Hands-off meeting notes with clear action items across Zoom, Teams and Meet. - Searching across meetings, email and chat in one place once you connect them. - Private feedback on how you come across in meetings. Okay at: - Heavy use on the free plan — its small monthly allowance, including reports shared with you, goes quickly. - Platforms beyond the big three. Webex, GoTo and RingCentral need the desktop app or a file upload, and uploads are paid. Avoid: - Leaving the default auto-join and sharing settings on. - Showing engagement scores about other people without their knowledge. Starter: Narrow it down Set auto-join to meetings you host and recaps to "Only you", then enable Read for one meeting from the Calendar page. Read's defaults join everything and share widely. Starting with one meeting lets you see exactly what a report contains before anyone else does. Starter: Ask about yesterday This is the example Read's own docs give for Ask Read. It's the quickest way to see whether searching your meetings saves you time. Prompt: What issues came up in my meeting yesterday? Starter: Read your coaching After a meeting where you spoke a lot, open your speaker coaching and look at pace and filler words. Read's help centre says coaching is always private to you, so it's a low-stakes way to get value from the meeting metrics. Starter: Practise opting out In a test meeting, type "Opt Out" in the chat and confirm Read leaves and no report is created. Knowing the exact words — and that anyone on the call can use them — makes it easier to tell others how to decline. Pitfalls: - Other people's Read still joins. Turning Read off for yourself doesn't stop an attendee's own Read account from joining; Read's help centre says a private meeting setting stops others adding it. - "Read Stop" isn't "Opt Out". "Read Stop" ends recording but keeps a report; "Opt Out" deletes everything. - Shared reports eat your free allowance. Reports others share with you count toward your monthly allowance on Free. - Custom vocabulary is paid. If names keep coming out wrong, the fix — a custom word list — needs a paid plan. Where to next: - More meeting & note tools: https://www.launchpadhq.io/discover?category=meetings-notes - Voice & speech tools: https://www.launchpadhq.io/discover?category=voice-speech - Workflow & automation tools: https://www.launchpadhq.io/discover?category=workflow-automation ### Getting Started with Recraft Canonical: https://www.launchpadhq.io/platform/recraft/guide Data: https://www.launchpadhq.io/data/v1/guides/recraft.json Last reviewed: 2026-09-07 Tool: Recraft (https://www.launchpadhq.io/platform/recraft): starting price Free, from See pricing; free tier yes; data as of 2026-10-26 The AI image tool that makes real logos and icons, not just pictures of them. You type a description of the image you want, pick a style, and Recraft draws it on a canvas in a few seconds. You edit what comes back in place, generate variations beside it, and keep what works. Then the same for the next piece. What it is: - An AI image generator with a design canvas, for people making things they intend to use — logos, icons, illustrations, patterns, mockups — rather than one-off art. - Its distinctive output is vector artwork: an image made of shapes, so it stays sharp at any size, unlike a photo, which blurs when you enlarge it. Most generators hand you a picture of a logo someone still has to redraw. Recraft draws the shapes and exports them as SVG, the file Figma, Illustrator and Canva all open. - The second piece is consistency: build a custom style from up to around ten reference images, then generate everything else in it — the difference between one nice icon and a matching set of twelve. - Canva is a layout tool with AI added; Recraft is a generator with editing added. Canva assembles a poster from existing pieces; Recraft makes the pieces. Before you start: - There is a free plan, it is genuinely usable for learning, and you do not own what you make on it. Recraft's documentation is blunt: free-plan images belong to Recraft, appear publicly in the community gallery, and aren't licensed for commercial use. - A browser and an email address is the whole requirement. - Everything costs credits, the platform's unit of spend: a small free allowance that refills daily, or a monthly paid pot that resets without rolling over. Vector generations currently cost more than plain images — a few credits each, more if you pick one of the higher-quality vector models (the AI "brains" doing the work) — so a vector-heavy afternoon empties an allowance fast. - If you want to use what you make, you'll be paying. Paid plans are monthly, with a discount for paying annually. Paying flips your images to private and hands you ownership and commercial rights for anything generated while subscribed — and those stay yours if you later cancel. Security: - The most important thing to know: on the free plan your generated images are public. Recraft's docs state they may appear in the community gallery and that Recraft, not you, owns them. Keep anything confidential or client-identifying out of a free account — what you type and what you upload alike. Security: Commercial use, by tier: - Free: owned by Recraft, publicly visible, not licensed for commercial use, and not listable on stock platforms, which require proof of ownership. - Any paid plan: full ownership and commercial rights for images generated while subscribed, client work included. - After cancelling: images made while subscribed stay private and stay yours, not retroactively made public. - On every plan: Recraft's terms forbid using its output to train other AI models or systems. Security: - You can upload your own images as references, or to trace and edit, and free accounts are capped at a few uploads a day. Only upload artwork you have the right to use — the public-gallery rule covers free-plan work built from it too. Best at: - Logos and app icons you need as clean, editable files rather than flat pictures - Matched sets — a dozen icons, an illustration series — that share one look - Turning an existing image or rough sketch into artwork that scales - Short, readable text inside a design: labels, badges, a poster headline - Repeating patterns and backgrounds, using its dedicated seamless-tile mode Okay at: - Photorealistic scenes and product photography — decent, not its headline strength - Long paragraphs of text in an image; short phrases are far more reliable - Multi-page layout — it makes the pieces, it isn't a layout tool Avoid: - Commercially usable assets on the free plan — the licence simply isn't there - Anything requiring a real person's likeness, or images passed off as photographs of real events - Asking for a logo "like" a famous brand's — that's legal exposure, not a shortcut Starter: Make a logo mark you can actually use This is Recraft's core trick: the result arrives as shapes you export into Figma or Illustrator, so you can recolour it, shrink it to a browser-tab icon, or hand it to a printer at any size. Swap "two colours" for "single colour, black" — one-colour marks are the real test, and what you need for stamps and dark backgrounds. Prompt: A minimalist logo mark for a coffee roastery, one coffee bean formed from two curved leaf shapes, flat vector, two colours, deep brown and warm cream, no text, plenty of space around it Starter: Build a matching icon set Consistency is what Recraft sells and what beginners underestimate. Generate the set, pick your favourite, save its look as a custom style, then generate the rest against it — that's how icon eleven still matches icon one. Prompt: A set of interface icons in one style, line art with rounded corners and the same line thickness throughout, single accent colour on a transparent background: a calendar, an envelope, a bar chart, a padlock Starter: Rescue a low-resolution image Turns a pixel image you already have into shapes you can rescale and recolour. Recraft has a dedicated tracing function and an editor on the same canvas, so you clean up the result without exporting elsewhere — useful for an old logo that only survives as a small blurry file. It works best on flat, high-contrast artwork; on a photograph you get a poster effect rather than a usable trace. Pitfalls: - The free plan is a demo, not a supply of assets. People generate a logo they love, put it on a real product, and only later read that free-plan images are owned by Recraft and publicly visible. Rights attach to what you make while subscribed, so subscribe before you generate the thing that matters. - Scalable output is a mode, not something you apply afterwards. Generate a pixel image and then ask for shapes and you get a traced approximation, not clean geometry. Choose the vector model or style up front when the output needs to scale. - Credits vanish faster than you plan for. Vector generations cost several times a basic image, the pro-quality vector models several times that again. Explore with cheap drafts, then spend vector credits once you know the composition you want. Where to next: - More AI image and editing tools: https://www.launchpadhq.io/discover?category=image-generation-editing - Marketing and brand tools that use these assets: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai ### Getting Started with Reducto Canonical: https://www.launchpadhq.io/platform/reducto/guide Data: https://www.launchpadhq.io/data/v1/guides/reducto.json Last reviewed: 2026-09-09 Tool: Reducto (https://www.launchpadhq.io/platform/reducto): starting price Free, from See pricing; free tier yes; data as of not dated Turns a messy PDF or scan into tidy, labelled text and tables for other software to use — a building block, not an app you chat to. You sign in to Reducto's browser workspace, drag a document in, and wait a moment. Back comes the same document rewritten as tidy, labelled data — headings, paragraphs, table cells — shown beside the original page. Change a setting, run it again, and put the two results side by side to see where they differ. What it is: - Reducto is a document parser: it reads a PDF, a scan, a spreadsheet or a Word file and hands back what was on the page as text, tables and figures with their positions marked, in a shape another program can read. - It is a building block rather than a finished product. It will not chat with you about your document, summarise it, or keep a library of your files — it produces the clean input that a chatbot, a search tool or an automated workflow needs in order to be accurate. - The company sells this to teams putting large document piles into AI systems, and says it has processed over a billion pages. Its browser workspace, called Studio, exists so you can watch what it produces before anyone writes a line of code. Before you start: - Reducto is built for people writing software, and its site assumes it. That is a fact about the product rather than about you: it emits data for other programs to consume, not answers for a person to read. If what you want is to ask questions of a PDF and get a readable reply back, ChatPDF is the tool for that job and this site has a page on it. - There is no sales gate on the entry plan. Reducto's pricing page currently lists a self-serve Standard plan that starts you with a pot of free usage — at its published parsing rate, enough for a sizeable batch of documents before you owe anything. Its Studio page says signing in takes "no setup, no credit card"; that is Reducto's own wording, worth confirming when you get there. - The two plans above Standard, Growth and Enterprise, are quoted by sales rather than published — and several of the data protections described under Security below start on those plans rather than the free one. - Have a genuinely awkward document ready before you sign in. The reason anyone goes looking for a parser is a scanned invoice, a financial statement whose table runs over a page break, or a form somebody filled in by hand. Security: - Every tool in this category uploads your document; Reducto's particular wrinkle is which route keeps it. Its data policy page says data submitted through the API (a way for programs to talk to each other without a person clicking) is set to expire within 24 hours — and then names Studio-run jobs as the exception to that window, without saying what window replaces it. So the no-code path a beginner would naturally take is the one route whose retention the policy does not state. Security: What Reducto's own documents say: - The same policy page states that Reducto never uses customer data for training "for users on our 'Growth' tier and above". That sentence is written so that it does not cover the free Standard plan you would be experimenting on. - Its terms of service run the other way: they grant Reducto a licence to "use, analyze, aggregate, reproduce, modify" what you submit, and to use it "to develop and improve the Services". The data policy page is not the document you agreed to; the terms are. - It says it has completed SOC 2 Type I and Type II — an audit of how a company handles customer data. A HIPAA-compliant processing pipeline, and a signed agreement covering health records, are offered to Growth and Enterprise customers, not on the free plan. - Reducto's own docs give two different retention windows: the FAQ says results are "generally available for 12 hours", the policy page says data expires within 24. Both are stated for API jobs. Neither covers the browser workspace. Security: - Practical version: until you know which plan you are on, test it with a document you would not mind a stranger reading. Short retention windows are reassuring but they do not undo the upload itself. Best at: - Complicated tables — merged cells, nested headers, rows that continue after a page break — which is the specific failure that sends people looking for a parser in the first place - Scans and photographs of pages, where the text has to be recognised off the image before anything else can happen - Showing its working: results link back to the exact region of the page a value came from, so you can check a number instead of trusting it - Feeding other software — 30+ file types in, one consistent shape of data out, so a spreadsheet, a slide deck and a scanned form all arrive looking the same Okay at: - Being explored without code. Studio is real and included on the free plan, but it is built for people who will eventually write some - Handwriting and less common languages. Reducto claims broad support; that is a claim to test on your own worst page - One-off jobs. Nothing stops you parsing a single PDF, but you are wiring up an industrial machine to open one envelope Avoid: - Asking questions about a document — it hands the contents to something that can - Being the finished workflow. There is no inbox to forward invoices to and no approval queue for a person to sign off in; Reducto stops at producing the data and assumes your system takes it from there - Trusting published head-to-head benchmarks, Reducto's own included. It publishes comparisons against named rivals on its site; treat those as a hypothesis to test on your documents, not a result Starter: Feed it your worst table You upload a financial statement or invoice whose largest table spans a page break, then check whether the rows after the break stayed in the right columns. Tables are the reason this category exists, and when a number lands in the wrong column the marked cell positions let you see where it came from rather than guessing. A parser that returns only text gives you no way to tell a misread character from a broken layout. Run the same file twice with one setting changed. Where the two runs disagree is where the document is genuinely ambiguous. Starter: Photograph a page and parse the photo You photograph one page of a printed form with your phone, upload it, and compare the result against the same page exported as a PDF. Reducto takes JPEG, PNG, HEIC and TIFF alongside PDFs, so one pipeline handles a phone snap and a born-digital file. That is the case that separates a parser from a text extractor: a clean PDF already contains its text and almost anything can read it. Then photograph the same page badly on purpose, at an angle and half in shadow, and find where it stops coping. Knowing the boundary is worth more than knowing the best case. Starter: Ask for four fields instead of the whole document You list the four values you would otherwise type into a spreadsheet from the document, then ask for exactly those four and nothing else. Parse and extract are separate products here with separate prices, because they are separate jobs: parse hands you the whole document, extract hands you the four values you were going to copy out of it anyway. Choosing the wrong one is the commonest way to overpay in this category. Ask for a field that genuinely is not in the document. How a tool behaves when the answer is absent tells you more than how it behaves when it is there. Starter: Price your real workload before you decide you like it You multiply your monthly page count by the published rate for the endpoint you need, then check that total against the free balance you start with. Reducto publishes a per-endpoint rate card — parse, extract, deep extract, split, deep split, classify and edit each priced separately — so you can cost a project off the pricing page without a sales call. Most vendors in this category quote by conversation, which makes the arithmetic somebody else's. Do it twice, once for the pages you have today and once for ten times that. The second number is what decides whether this becomes infrastructure or stays an experiment. Pitfalls: - The free usage is a balance, not a monthly allowance. The pricing page says standard rates apply after it, which reads as one pot you spend down rather than something that resets. Bulk-parsing a back catalogue is exactly how you would burn through it without noticing. - The endpoint names are the pricing, and the "deep" variants currently cost roughly double their plain counterparts on the published rate card. Reaching for the strongest option by reflex is the expensive habit here. - Upload limits depend on how you send the file, not on the file. Direct upload is currently capped at 100MB, a presigned upload at 5GB, and passing a link has no stated cap — so "too big" usually means you used the wrong door rather than that the document is unsupported. - This company is moving fast. Product names, endpoints and prices are worth re-checking against its own docs before you build anything on them. Where to next: - Other ways to get answers out of documents: https://www.launchpadhq.io/discover?category=document-pdf-processing - More developer services like this one: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - Reading long documents rather than parsing them: https://www.launchpadhq.io/discover?category=research-academic-tools ### Getting Started with Relevance AI Canonical: https://www.launchpadhq.io/platform/relevance-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/relevance-ai.json Last reviewed: 2026-09-17 Tool: Relevance AI (https://www.launchpadhq.io/platform/relevance-ai): starting price Free, from See pricing; free tier yes; data as of not dated Describe a job in plain English and get an AI specialist that does it — no flowchart, no code. You open a builder and describe, in plain English, one job — who this agent (AI that takes actions on its own rather than only answering) is and how it should behave. You pick what it is allowed to reach: your email, your files, the web. Then you chat with it, see what it did, and adjust the wording. What it is: - Relevance AI sells what it calls an "AI Workforce." Instead of drawing a flowchart of steps, you write down who a worker is and what it is allowed to touch, give it a name, and let it work out the steps itself. Several of these specialists can then be grouped into a Workforce that handles a bigger job between them. - That framing cuts both ways. Describing a job is far easier than diagramming one, which is why people who bounced off flowchart tools often get further here. But you give up knowing the exact route in advance. - The company is Australian — the site footer reads "© 2026 OnSearch Pty Ltd T/A Relevance AI" — and it has moved on from where it started, which was a product for searching messy data. A separate marketplace holds over a thousand ready-made agents you can copy into your own account rather than starting from a blank box. Before you start: - There is no free plan any more, and this is very recent. Relevance's pricing documentation now reads "The Free plan is retired and closed to new signups" — a change merged into the vendor's own documentation on 11 September 2026. The cheapest way in is the Pro plan, cheaper paid annually than month to month; current prices are on its Launchpad page. You pay by card and sign up yourself; no demo call is required, despite "Book a demo" being the loudest button on the homepage. Note that Relevance's own marketplace site still advertises "Start for free." The pricing docs are the binding source; the marketing has not caught up. - Two separate meters run while you work, and Relevance's word for one of them is confusing. Actions are counted each time one of your agent's tools runs — so "action" here means a unit of spend, not the "then" half of an automation the way most tools use it. Vendor Credits are the AI model bill itself, passed through with no markup. Pro includes a monthly allowance of both; run out of either and you buy top-ups rather than being cut off. Unused Actions from a top-up carry forward; the ones included with your plan reset when it renews. - You choose where your data physically lives when you sign up — US (N. Virginia), EU (London) or AU (Sydney) — and the docs say that choice cannot be changed afterwards without contacting support. If you have any obligation to keep data in a particular region, decide before you click. - If you already pay for an AI service of your own, you can plug your own key in and skip the Vendor Credits meter entirely. The docs list this as available on every tier. It does not help with Actions, which are charged either way. Security: - A flowchart tool does what you drew. A Relevance agent is handed a set of tools and decides for itself which to reach for on any given run, so the same agent can take two different routes on two different days. What you approve is a capability, not a sequence, and the strongest steering you have is what you write in the agent's instructions — including the things you forbid. Security: What that means in practice: - Write your prohibitions down. "Do not send email to anyone, only show me the draft" belongs in the agent's own instructions, because there is no step you can simply leave out. - Give an agent the narrowest account you can, and prefer ones that read over ones that send or delete. - Run it against a test folder or label a few times and read what it actually did, not just whether it finished. Security: - On the company's side, the position is better than average and worth knowing: Relevance says it is SOC 2 Type II compliant (an audit of how a company handles customer data), encrypts data in transit and at rest, and states plainly that it does not use your data to train its models unless you have signed a specific partnership agreement. Model requests go to vendors including OpenAI and Anthropic; Relevance says nothing is stored or trained on in that process. That is a clear commitment rather than silence — but it is still a third party processing whatever you route through it. Best at: - Jobs you can describe but could never diagram — "read these support emails, work out which are about billing, and summarise each one" is one sentence here and a branching chain anywhere else. - Reusing one specialist across several jobs. An agent you wrote once is a thing with a name that other agents and Workforces can call on. Okay at: - Simple two-step plumbing. "New form row → append to a sheet" works, but you are paying agent prices for something a basic connector tool does more cheaply and more predictably. - Exact timing. Each plan has a limit on how many tasks run at once, and anything over it is queued until capacity frees up rather than failing outright. - Long single jobs. A tool run currently stops at 15 minutes, and an agent times out 15 minutes after your last message to it, so very long tasks need to be broken up. Avoid: - Anything where you must know in advance exactly what will happen. Non-determinism is the design, not a defect — if a wrong route on one run out of fifty is unacceptable, this is the wrong shape of tool. - Learning what automation is in general. With the free plan gone, your first afternoon of clicking around and getting it wrong now costs money, and a tool with a standing free tier is a cheaper classroom. - Trusting a marketplace agent because it is listed. Some of them depend on third-party services you must supply your own key for, which carry their own separate bills. Starter: Copy a working agent before you write one When you open the marketplace → then clone an agent that is near what you want into your project, and read its instructions before changing a word. What you inherit here is prose, not wiring. A Zapier template hands you a configured chain with nothing written down to read, because there the logic is the diagram — so reading someone else's finished instructions is the fastest way to learn how specific yours need to be. Change one sentence, run it, and see what moved. Starter: An inbox triage agent that is not allowed to reply The refusal line is the load-bearing part. On Zapier you keep an agent from sending mail by simply not adding a send step; here the agent picks its own tools, so a prohibition written into the instructions is the control you actually have. It is also the cheap shape: a read-and-report agent costs a fraction of one that writes. Once you trust it, change the last line to have it draft replies without sending them. Prompt: You are my inbox triage assistant. Each morning, read the unread email in my inbox and sort it into three lists: needs a reply from me today, can wait until later this week, and needs no reply. For each item in the first list, write one sentence saying what the sender wants. Do not send, reply to, archive or delete anything. Post the three lists back to me as a message. Starter: A scheduled research agent that files what it finds When a schedule fires each weekday morning — your trigger, the "when" that starts an automation → then the agent searches the web for mentions of a name you give it, and writes one row per find into a spreadsheet. This is the card that makes Relevance's bill visible before it surprises you: one scheduled run that searches, reads two pages and writes a row is four Actions, not one, so a daily agent is roughly eighty Actions a month against Pro's included allowance. Zapier's per-task counter and Lindy's per-task credits meter the job; this one meters every tool the agent chose to pick up. Start it weekly — you can always shorten the interval once you have seen a month of real consumption. Starter: A second specialist, so you can see what a Workforce is for Relevance's unit is a named worker you can hand to other jobs, so this one summariser gets reused by the agent above and by whatever you build next. Lindy's assistants and Zapier's Zaps are both built around one flow doing one job end to end; the reason to accept this platform's unpredictability is that the pieces are reusable rather than single-use. Give it a deliberately bad page and see whether it uses that last line honestly. Prompt: You are a research summariser. You will be given a web page or a block of text. Write three bullets covering what it says, who it is aimed at, and anything that looks time-sensitive or out of date. Then write one line saying what you could not determine from the text. Do not search for anything else and do not guess. Pitfalls: - "It worked when I tested it" means less here. Run anything consequential several times on harmless data before pointing it at the real thing. - The meter counts tools, not runs. A loop over twenty items multiplies every Action inside it, and that is where a month's allowance disappears without anyone noticing. - Almost every guide and review you find is now out of date on price. Anything written before September 2026 describes a free tier that no longer accepts signups. Check the pricing page yourself before planning around a number you read elsewhere. Where to next: - Workflow & Automation: https://www.launchpadhq.io/discover?category=workflow-automation - Sales, Marketing & SEO: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai ### Getting Started with Replicate Canonical: https://www.launchpadhq.io/platform/replicate/guide Data: https://www.launchpadhq.io/data/v1/guides/replicate.json Last reviewed: 2026-09-08 Tool: Replicate (https://www.launchpadhq.io/platform/replicate): starting price See pricing; free tier yes; data as of not dated A catalog of thousands of ready-to-run AI models — image, video, audio, text — that you can run from a form in your browser and pay for by the run, with no subscription. You browse a directory of AI models (the AI "brain" that does the actual thinking), open the one you want, and its page gives you a form — type a description, upload a photo, set a couple of options. Press Run, wait a few seconds, and the result appears underneath. Change an input, run it again. What it is: - Replicate is a directory of AI models you rent by the run. Thousands of them — image generators, video models, upscalers, transcription, old-photo restoration, text models — each on its own page with a form, a price and a Run button. You don't open an account with the lab that built each model; you open one here and run any of them. - Two things shape what it feels like. Models come in two grades: "official" ones Replicate maintains, which stay running and charge a flat price per image or per thousand words, and community ones uploaded by their authors, charged for the seconds of hardware time they use. And Replicate was acquired by Cloudflare in late 2025 — replicate.com still runs under its own name and was still shipping changes into 2026, but its long-term direction is now Cloudflare's call. Before you start: - The account gate is the first surprise. Replicate's sign-in page currently offers only "Sign in with GitHub" — there is no email-and-password option, so signing up means creating a free GitHub developer account if you don't have one. It takes a couple of minutes, and it stops more non-developers than anything else on the site. You also need to be 18. - "Free" here is a fixed number of runs, not a monthly allowance. A "Try for free" collection lets you run a small set of selected models a limited number of times once you have an account; after that you're asked to set up billing. - You pay by the run out of a balance you top up, so there's no subscription to cancel — but budget a little real money if you want to explore properly. Official models publish a flat per-run price on their page; community models bill for however long the hardware runs, which is harder to predict. Worth knowing up front: Replicate removed its self-serve monthly spend limit in July 2025, so there's no toggle to cap yourself. - You do not need to write code. The code path is there when you want it; nothing on day one requires it. Security: - The thing worth knowing is where your runs end up. Replicate stores the inputs, outputs, files and logs from what you run, and the two paths are not the same: runs you make in the browser are kept indefinitely until you delete them, while runs made through the API (a way for programs to talk to each other without a person clicking) are removed automatically after an hour. The convenient path is the one that remembers everything. Security: Three habits that fit how it actually works:: - Treat your run history as a filing cabinet, not a scratchpad. - Read the model, not just the platform. Most models here were uploaded by other people, and some forward your input to another company's service to run it. Replicate's terms hand you the rights to what you generate — but each model carries its own licence on top, and that is the one that governs commercial use. - Don't send anything you'd hesitate to hand a third party. Replicate's terms take a broad licence over what you submit in order to run the service. Fine for a photo you'd post publicly; poor for a passport scan or a client's contract. Best at: - Trying one specific model — a named image, video, upscaling or transcription model — without opening an account with the company that made it - Narrow, unglamorous jobs nobody built a consumer app around — face restoration, upscaling, background removal, reading text out of a scan Okay at: - Chatting — text models do run here, but a chat window is a better place to talk to one Avoid: - Building a habit around a community model — its author maintains it, not Replicate, and its inputs can change between versions - Assuming you can cap your spending in settings — the self-serve monthly spend limit was deprecated in July 2025, and getting one now means asking support Starter: Start in the 'Try for free' collection Runs one of a small set of selected models — image generators, upscalers, photo restorers — a limited number of times without setting up billing. It's the one corner of the site where the Run button works before a payment method does. Open a model from search instead and you get a short run of goes and then a billing prompt. Starting here means you find out whether the output is any good before you decide to fund an account. Bookmark the collection page — it's easy to wander into the wider catalog and lose track of which models were the free ones. Starter: Prefer models labelled 'Official' while you're learning Narrows your choices to the models Replicate maintains itself. The two grades are billed on different clocks. An official model tells you it costs about four cents an image before you press anything; a community model charges for however many seconds the hardware ran, which you can't know in advance. Learning what things cost is far easier on the side of the catalog where the price is a number rather than a stopwatch. Starter: Run one model in the browser before writing a line of code Uses the form on the model page — type or upload, press Run, see the result on the same page. Each model page generates its form from that model's own inputs, so the browser covers the whole catalog rather than a demo subset, and those runs stay in your account history where you can see what you ran and what it cost. Starter: Go looking for the model nobody made an app out of Searches the community catalog for the narrow jobs — sharpening a blurred face in an old photo, enlarging a small image, pulling text out of a scan, cutting out a background. These are research models their authors published and never wrapped in a product, so there's no website to sign up to; this catalog is often the only place a non-developer can press Run on one. It's also the clearest reason to be here rather than at a curated model service, which carries the popular models and not the long tail. Run counts are shown on the model cards. One that thousands of people have used is a safer first try than one with a handful. Pitfalls: - Scrolling past the price. It's on the model page — per output, or per second of hardware — and it's the easiest thing to skip on the way to the form. - Judging a model by a slow first run. One that hasn't been used lately has to start up first, which can take minutes. You aren't charged for that wait. - Treating an output link as storage. Files from runs made through code are deleted after an hour — download anything you want to keep. - Expecting 'free' to come back. The free runs are a one-off allotment on selected models, not a monthly allowance. Where to next: - AI APIs & Developer Services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - Local & Open-Source AI: https://www.launchpadhq.io/discover?category=local-open-source-ai ### Getting Started with Replit Agent Canonical: https://www.launchpadhq.io/platform/replit-agent/guide Data: https://www.launchpadhq.io/data/v1/guides/replit-agent.json Last reviewed: 2026-09-08 Tool: Replit Agent (https://www.launchpadhq.io/platform/replit-agent): starting price $20/mo ($18/mo billed annually) including $20 toward more powerful models; free tier yes; data as of 2026-10-14 Describe an app in ordinary sentences, get a working one you can click — and learn where the money goes. You open a project in your browser and type what you want built, in ordinary sentences. An agent — AI that takes actions on its own rather than only answering — writes the files, runs the app and shows you a live preview. You click around, tell it what looked wrong, and it goes again. What it is: - Replit Agent builds working software from a description you type, inside Replit's workspace in your browser. Nothing is installed and you are never asked to open a terminal. The current version is Agent 4, released March 2026. - As it works it saves checkpoints — snapshots of the whole project at each finished piece of work. - It is genuinely aimed at people who don't code. But what comes out is a real app holding real data, and putting that in front of other people is a bigger step than making it. Before you start: - Sign up at replit.com and you land on the free Starter plan: Agent credits — the platform's unit of spend — that refresh daily up to a monthly cap, plus one published app, all in the browser. - Starter's ceiling is features as much as spending. It runs only Replit's "Lite build" mode, for quick, focused changes; the fuller build, Plan Mode, outside connections and anything that isn't a web or mobile app need Core, Replit's main paid plan. - Publishing is billed separately, as a small monthly charge plus per-request fees, and Starter's one free published app expires after 30 days unless you re-publish it. - Set a spending limit before your first real build, under Settings → Account → Usage → "Manage limits". Nothing caps you by default, and auto-reload buys another credit pack on its own when you run low. Security: - Everything happens on Replit's machines: what you type, the code it writes, whatever your app stores. Replit's privacy policy, updated August 2026, says it collects "code, project files, text, commands and prompts" and uses that data "to improve the accuracy of our machine learning technologies such as code generation." No consumer plan has a switch to turn that off; the granular controls are an Enterprise feature. Security: Four things to get right before anyone else sees it: - Keep passwords and access keys in Replit Secrets, never in the chat window — and per Replit's own checklist, never in client-side code, local storage or cookies - Publish privately if it isn't meant for the public; private publishing works on Starter and Core, and stops unauthorised requests before they reach your app - Confirm every part of the app that touches data requires a login — Agent wrote that code, and checking it is your job - Read Replit's security checklist in full before anything with a sign-up form goes live Security: - One detail is a real safeguard. Every Replit app keeps two separate stores of information: a development one Agent works in, and a production one behind your published app that Replit's docs say Agent can't touch. That split exists because it had to — in July 2025 an Agent run deleted a customer's live database during a code freeze, which Replit's leadership acknowledged publicly. Best at: - Turning a plain-English description into something clickable within the hour - Fixing its own mistakes once you say exactly what you saw go wrong - Building the unglamorous parts — a sign-up form, somewhere to save entries, a page listing them back — without you naming a technology Okay at: - Anything beyond a small web or mobile app on the free plan; slides, data apps and animations are Core-only - Larger or older projects, where Replit steers you toward Max, its most capable and most expensive mode Avoid: - Treating the free published app as permanent hosting — Starter's one free app expires after 30 days and must be re-published by hand - Leaving auto-reload on before you know your own pace; it buys another credit pack without asking again - Other people's real information in an app you haven't reviewed — Agent wrote the code protecting it, and nothing has checked that code but Agent Starter: Start with one screen, not one app Replit prices a request by the effort it takes, rather than the flat per-checkpoint fee it used to charge, and asks you to confirm before paid work starts. Something you can describe in four sentences keeps that confirmation cheap and stays checkable. Swap in your own idea, then delete half of it before you send. Prompt: Build one web page where I can add a task with a due date, see the list sorted by date, and tick items off. No login, nothing saved between visits yet. Plain, readable styling. Starter: Make it write the plan first Replit's built-in Plan Mode requires Core, so on free Starter, asking for the plan inside the message is the only way to see what Agent intends before it edits files. If a step surprises you, say so before you tell it to go. Prompt: Before you change any code: write out what you'd build, in what order, what counts as finished, and what you're deliberately leaving out. Don't build anything yet. Starter: Get the live version's data in on purpose A Replit app keeps its building data and its live data in two separate places, and the live one starts empty. A published app looking blank is the system working as documented, not something you broke. Move reference data you'd hate to retype; leave the test junk behind. Prompt: The published app is empty. Explain where the entries I made while building went, then tell me how to get the categories — not my test entries — into the live version. Starter: Practise rolling back before you need to Agent History keeps a checkpoint at each finished piece of work and restores any of them, which makes "undo that" a cheaper first move than "now fix it". Do it once on something trivial, so the panel isn't unfamiliar during a real mess. Prompt: That last change made things worse. Show me the checkpoints from this session and what changed in each, then roll back to the one before the layout change. Pitfalls: - Entries made while building do not appear in the published app, which reads a separate, empty store until you copy things across. - Cost tracks the work a request takes, not the number of messages. Replit dropped its flat 25-cents-a-checkpoint fee in 2025, so a debugging spiral costs more than the message count suggests. Where to next: - AI Coding & Development: https://www.launchpadhq.io/discover?category=ai-coding-development - Workflow & Automation: https://www.launchpadhq.io/discover?category=workflow-automation ### Getting Started with ResearchRabbit Canonical: https://www.launchpadhq.io/platform/research-rabbit/guide Data: https://www.launchpadhq.io/data/v1/guides/research-rabbit.json Last reviewed: 2026-09-15 Tool: Research Rabbit (https://www.launchpadhq.io/platform/research-rabbit): starting price Free; free tier yes; data as of not dated Start from one paper you trust and let the citation trail lead you to the rest. You search for a paper you already trust and add it to your library. ResearchRabbit fans out from it, showing related papers as a list beside a map of dots and connecting lines. You skim, keep the ones worth keeping, and search again from the bigger set — each round widens the net. What it is: - ResearchRabbit is a discovery tool for academic literature. You give it a few papers you already know are relevant, and it recommends more by following citations (the earlier work a paper refers to, and the later work that refers back to it) across a database it currently describes as more than 310 million articles — a figure that has moved upward since its 2025 relaunch, so treat it as approximate. - It is not a chatbot. There is no box to ask a question in, and it does not summarise papers or write anything for you — the vendor's help centre says ResearchRabbit "does not use large language models to power its core recommendation algorithms." What it does instead is arithmetic on a citation network. - Since May 2025 it has been owned by Litmaps, the New Zealand company behind the literature-mapping tool of the same name; ResearchRabbit's own site now carries a "© 2026 Litmap Ltd" footer. The product was rebuilt and relaunched in October 2025, and its paid tier arrived with that relaunch. Before you start: - Free to use after a sign-up at researchrabbit.ai — no card required. The pricing page currently lists the free plan as free forever, and the help centre explains the arrangement in the vendor's own words: "We're able to offer the Free Tier thanks to the income we receive from subscribers to ResearchRabbit+." That is the company's own current wording for its commitment, not a guarantee anyone else is making — it is funded by paying users rather than by advertising. - The free-tier ceiling that will actually stop you is the cap on seed articles — seed articles being the papers you hand the tool to search out from. The free cap is generous for finding your way into a topic and tight for a full systematic review. ResearchRabbit+ (billed annually or month to month, with discount codes offered for 100-plus countries, so what you pay may differ) raises that cap several times over and adds advanced search filters, multiple projects and alerts. The free plan is also reported to cap you at a single project and a small number of authors per search, but the pricing table spells out only the seed-article limit, so treat the other two as likely rather than confirmed. - Bring one paper you already trust — a title, or the link to it. The tool is only as good as what you seed it with, and starting from a bare keyword gives it far less to work from than starting from a paper. - It only knows what its sources know. By the vendor's own account the catalogue comes from Semantic Scholar, Crossref and OpenAlex, reached through parent company Litmaps; a university library evaluation names a slightly different set and says the provenance of further sources is unclear. Either way it is strong on journal articles and conference papers, and thin on books, theses, government reports and other work that was never indexed as a journal article. Security: - ResearchRabbit sits at the gentler end of the AI-privacy spectrum, and for a structural reason rather than a promise: you are not pasting text into it. What it holds is a record of which published papers you saved, how you grouped them, and any notes you wrote — not your manuscript, not your data, not your unpublished work. Security: What it does hold about you: - Your account email, and the papers in your library and collections - Your notes, labels and colours — yours to write, but stored on their servers - Whatever you imported from Zotero: the reference records, not the attached PDFs - Any collection you share by link — a public link means exactly that, anyone holding it can read the collection Security: - Two honest caveats. Your collection is a map of what you are working on, which is more revealing than it first looks; if the topic itself is confidential, keep it out of a shared collection. And we could not read ResearchRabbit's current privacy policy or terms — both are served through an embedded viewer that needs a live browser — so we cannot tell you their effective date or what they say about sharing your data, and whether your library or your activity feeds the recommendation models is not something we could confirm from its public pages either. The help centre states only that large language models do not power the core recommendations. Read the policy yourself at researchrabbit.ai before you put anything sensitive in. Best at: - Finding papers a keyword search misses, by following citations outward from one you already have - Getting oriented in an unfamiliar field quickly — the map shows which papers everything else hangs off - Working backwards from a recent paper to the foundational ones it rests on - Keeping a growing reading list organised in collections and subcollections, with notes and colour labels - Handing a reading list to a supervisor or co-author as a shared link Okay at: - Starting from keywords alone — it will do it, but it is built to start from papers - Author-based discovery — a route through authors exists, though a reviewer who tested the rebuilt tool rated it the least useful of the recommendation types - Telling you whether a paper is any good — it shows you how connected a paper is, which is not the same thing as how sound it is Avoid: - Systematic reviews that need a reproducible, documented search. A university library evaluation found results can differ depending on whether you search by title or by identifier, and that you cannot search across all its underlying sources at once — both are problems for a method section you have to defend. - The map's vertical axis is not a quality ranking. It plots citation count, which is partly just an age effect: old and fashionable papers ride high, and a strong 2026 paper sits at the bottom because nobody has had time to cite it. - Using it as your only search. Following citations works outward from what is already connected, so a genuinely isolated or brand-new paper is precisely the kind it will not surface. Starter: Seed it with one paper you already trust Search for a paper by its exact title, add it to your library, and run a search from it. You get a list of recommendations down one side and a map beside it, where the horizontal axis is time, oldest to newest, and the vertical axis is how heavily a paper has been cited. Your own papers show as filled dots; recommendations show as hollow ones. Connected Papers builds one graph per origin paper and hands it to you finished. ResearchRabbit instead keeps your seeds in a library you keep adding to, and re-runs the recommendations against the whole growing set — so the first seed you pick shapes every round after it, in a way it does not in a tool where you simply generate another graph. Add two or three seeds rather than one. The recommendations tighten noticeably once the tool can see what your papers have in common. Starter: Walk backwards, then forwards, from the same paper With a paper selected, switch between Earlier Work and Later Work. Earlier Work is that paper's own reference list — what it was built on. Later Work is everything published since that cites it. One paper, two opposite directions in time. The ranking behind these two views is arithmetic, not taste: the vendor's help centre says recommendations are calculated from the most influential articles in the local citation network, "where influence comes from both the number and quality of citations." That makes a well-cited older paper's Later Work the fastest honest way to find out whether its conclusion actually survived — something a similarity-based map like Connected Papers, which does not separate direction at all, cannot tell you. An empty Later Work usually means the paper is too recent to have been cited yet, not that it was ignored. Starter: Use Similar Work when the citation trail goes cold Switch to Similar Work. Unlike Earlier and Later Work, this view does not appear to follow citations at all — a reviewer who examined the rebuilt tool describes it as matching papers on their title and abstract, though the vendor does not spell out the method. Either way it can reach work in a neighbouring field that never cited yours and never will. Knowing which of the three views follows citations and which does not is the difference between using ResearchRabbit and clicking around it. Litmaps and Connected Papers both fold similarity into a single blended score; ResearchRabbit keeps the routes on separate buttons and makes you pick one — more work, and more control over what you are actually asking for. If Earlier and Later Work keep returning the same clique of papers, that is the signal to switch to Similar Work. Starter: Pull your Zotero library in Under Profile, then Account Settings, then Preferences, link your Zotero account — it uses a one-time permission screen rather than a key you copy. Then use the Zotero button on the Library page to choose a Zotero collection, pick which articles come across, and say which ResearchRabbit collection they land in. It is a manual import you re-run when you want it, not a live sync, and it runs one way. To get papers back out to Zotero you export a citation file from ResearchRabbit and import it on the Zotero side. The vendor says a two-way sync is in development and has not given a shipping date; until it ships, whichever side you last edited is the side that is correct, and nothing tells you which that was. Import one narrow Zotero collection rather than your whole library — the help centre warns that large Zotero libraries take a while to load. Starter: Save a collection and share the link Group what you have found into a collection, then share it, either with named collaborators or as a public link. Collections and subcollections, notes, stickers and colour labels are all on the free plan, and the pricing page puts no cap on how many you make. Someone opening your shared ResearchRabbit collection sees not just the list but the route — which papers you seeded and what came back. That is a different artefact from emailing a bibliography, and unlike the 50-seed search ceiling, it is the part of the product the free plan does not meter at all. Pitfalls: - Seeding with a review article floods you. Reviews cite hundreds of papers, so Earlier Work comes back enormous and undifferentiated. Seed with primary research papers and let the reviews surface as recommendations instead. - It is very easy to keep exploring and never start reading. Every paper opens three more routes, and the design rewards that. Decide how many rounds you will do before you open it. - The catalogue lags publication. A paper that came out last month may not be in there yet, and the citations to it certainly will not be — so recency is the one thing you should not trust it on. - A paper appearing in the map is not a paper you can read. ResearchRabbit surfaces the record; it does not tell you whether the full text is open, paywalled, or behind a subscription your institution lacks. - "Recently Found" is a holding pen, not a collection. Saved papers land there by default, and if you never move them into real collections you end up with one long undifferentiated list — which defeats the organising that is half of what the tool is for. Where to next: - More research tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Working with PDFs: https://www.launchpadhq.io/discover?category=document-pdf-processing - AI chatbots: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Restb.ai Canonical: https://www.launchpadhq.io/platform/restb-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/restb-ai.json Last reviewed: 2026-09-12 Tool: Restb.ai (https://www.launchpadhq.io/platform/restb-ai): starting price See pricing; free tier yes; data as of not dated AI that reads property photos and writes down what it sees — built for real-estate companies, with one way in for individuals. Photos go in, facts about the property come back. There is no chat box and nothing to write. Software hands it a picture — a kitchen, a roof, a street view — and it returns labels: which room this is, what features it has, what condition the place looks to be in. What it is: - Restb.ai's AI does one thing: computer vision (software that looks at a photo and works out what's in it), trained specifically on property pictures. Point it at a listing's photos and it names the rooms, spots features and damage — the company says more than 500 distinct details — and rates how the place looks. Since May 2026 it has been owned by Clear Capital, an American property-valuation firm, which said it would keep the Restb.ai brand. - Almost nobody buys it as a person. Its customers are the businesses behind property listings — the shared listing databases estate agents post homes to (in the US, an MLS), property portals, appraisal firms, insurers and lenders — and it reaches them as an API (a way for programs to talk to each other without a person clicking). - There is one exception, and it is how an individual gets in: Restb Lens, a web tool from the same company, pitched at appraisers, lenders and valuation reviewers rather than the general public. Before you start: - The main product is not something you sign up for. Restb.ai's site has no log-in, no sign-up and no self-serve access — the only buttons are "Try our demos" and "contact us", and its FAQ says plainly that it does not have an app. The pricing page promises "accessible pricing" and prints no numbers; the FAQ gives the shape instead: a fixed monthly fee covering an allowance of requests, an overage fee for each one beyond it, and separate pricing for large batches of property data. - If you are a US estate agent, check before you buy anything — you may already have it. Restb.ai sells to MLSs, which switch it on for their members, and by the company's own April 2025 count that reached more than 800,000 agents and brokers across North America. No vendor page says whether those features are covered by your dues or billed on top, so ask your board rather than assuming. - The one path that does not go through a salesperson is Restb Lens. Its published plans start with a single-user plan that caps monthly property searches, rising through pricier tiers with more seats and more searches, and the site advertises a two-week free trial you request by form — "We'll send you a promo code right away. No credit card needed." That trial is the only free look, and we could not confirm from outside whether it is still being offered. - It is built around American property. Restb Lens runs on nationwide property data, and its condition and quality ratings follow the 1-to-6 scales Fannie Mae and Freddie Mac use in US appraisals. The photo labelling does not look country-specific — Restb.ai has European customers and was founded in Spain — but the scores, the recent-sales comparisons and the compliance rules are US-shaped. - Restb Lens needs no coding; the API side genuinely does — Restb.ai's FAQ describes clients integrating it themselves. If you only want to see what AI makes of a photo you own, a general chatbot will describe one at no cost; property data is what brings you back here. Security: - What you hand this tool is photographs of the inside of someone's home — usually your client's, not yours. You are not typing your own secrets, you are forwarding other people's property. Security: What the paperwork actually says: - Restb.ai's terms and conditions, last updated 2 February 2026 and the version in force today, say the client "shall retain all rights and ownership of such User Content" and that Restb.ai "will not use User Content in any way other than in connection with providing the services". - The same terms say that when an agreement ends, Restb.ai "shall cease any and all use of Your Data and will destroy or return it to you". - The FAQ puts it more strongly: "We don't have access to your data and do not control it. All responses are returned in real time." Those two documents do not quite agree — a contract promising to destroy or return your data is describing data it holds. The terms are the binding document, so read them as the floor and the FAQ as the friendlier summary. - Those terms were revised three months before Clear Capital bought the company, and have not been reissued since the sale. If data handling matters to your work, put that question to them in writing. Security: - Restb.ai sells a photo-compliance tool for flagging listing images that break privacy and fair-housing rules — faces, people, things nobody meant to publish. Take that as evidence that listing photo sets routinely contain exactly that. Best at: - Naming what is in a property photo — room type, features, damage, architectural style - Doing it at volume, applying the same judgement to every image in a listing set - Turning photos into ratings an American appraisal form already has a box for - Catching photos that should not go live — watermarks, duplicates, faces Okay at: - Being bought by one person — the API side is not sold to individuals at all - Property outside the US — the photo labels travel, the scores do not - Explaining itself — you get a rating and a confidence number, and nothing public describes showing you which photo drove the score Avoid: - Treating a condition score as an inspection — it grades what was photographed, so an unphotographed room or a problem behind a wall is not in the number - Judging people, households or neighbourhoods from listing imagery — in US housing the line between describing a property and describing who lives in it is a legal one Starter: Find out whether your MLS already switched it on Next time you enter a listing, watch what happens once the photos upload. Restb.ai's MLS product fills listing fields in from the images rather than appearing as a separate app, so if room types and features arrive already ticked, there is a decent chance it is this. Restb.ai does not sell to agents at all — it sells to the MLS platform vendors its own customers page names (Rapattoni, FBS, ICE, Cotality), who put it in front of members. So check your listing-entry screen, not an invoice or an app store. If nothing fills itself in, ask your MLS's support desk which AI features are on for your board — this varies by MLS, not by agent. Starter: Run the free trial against a house you know inside out Request the two-week Restb Lens trial, then look up a property you have actually stood inside — your own home, a place you sold — and compare the condition and quality ratings it returns with what you already know about the building. The ratings map onto the American appraisal scales noted above, so this is one of the few AI outputs a beginner can mark against a published external standard rather than a feeling. Starter: Walk the demos before you talk to anyone Restb.ai's demo page links a separate demonstration for each product — image tagging, property condition, image captions, photo compliance, watermark detection, duplicate detection and visual similarities. This is not one product with a price; it is ten separately sold solutions, and a quote depends entirely on which you name. The demos are listed as open to visitors, but we could not confirm from outside whether any of them ask for an email first, or whether they only show the company's own sample photos. Starter: Ask them to run your own photos Restb.ai's FAQ offers "complimentary PoCs" — free trial runs the company sets up for you — "where you can use our visualization tools to see the impact of our solutions without writing any code", and its demo page carries a form headed "Want to test our AI with your images?" It is the only route the company documents for getting your own images scored without a developer — openly a sales step, and described as costing nothing. Pitfalls: - Two products, one name. "Restb.ai" is the business product; "Restb Lens" is the web tool you can actually sign up for. Searching for Restb.ai pricing lands you on a page that talks about affordability and prints no figures — that is the real page, not a broken one. - Assuming that because it arrived through your MLS, you control it. You get whichever features your board switched on; nothing public describes a setting you can change or an upgrade you can buy for yourself. - It meters by property, not by hours. The Lens search allowance noted above is the real ceiling. The vendor's own user guide never defines what counts as one search, so treat re-runs and report exports as possible charges until someone tells you otherwise. - Trusting a score attached to stale photos. A listing whose pictures predate a renovation — or a flood — scores the house that was photographed, and nothing in the output tells you when the shutter clicked. Where to next: - Finance & Real Estate AI: https://www.launchpadhq.io/discover?category=finance-real-estate-ai - AI APIs & Developer Services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Runway Canonical: https://www.launchpadhq.io/platform/runway/guide Data: https://www.launchpadhq.io/data/v1/guides/runway.json Last reviewed: 2026-10-14 Tool: Runway (https://www.launchpadhq.io/platform/runway): starting price $15/mo monthly or $12/mo billed annually; free tier yes; data as of 2026-10-14 Turn a text prompt or a still image into short AI-generated video — a filmmaking tool you drive by describing what you want. You describe a clip in words, or give it a still image to animate, and it renders a few seconds of video. You watch the result, then either adjust the prompt and generate again or describe a specific change to the clip you already have. What it is: - Runway is an AI video generator (with image and audio tools alongside). You give it a prompt, or a starting image, and it produces a short video clip; you can also direct edits by describing the change you want. It's used by hobbyists making social clips all the way up to professional studios, so the ceiling is high — but the on-ramp is simple enough for a first-timer. Before you start: - The free plan is a one-time sample, not a monthly allowance: a small credit deposit that never expires (official FAQ, 2026-10-14), and clips carry a Runway watermark. The paid plans — Standard, Pro and Max — are monthly, each cheaper billed annually. - Budget for iteration. AI video rarely nails it first try; plan on multiple takes per usable clip, and know that credits mostly don't roll over (only the top plan rolls a little forward). - Web-app credits and API credits are separate pools — don't assume buying one covers the other. Security: - Runway is unusually creator-friendly on rights: it grants commercial-use rights on every tier, including the free plan, and you retain ownership of what you generate, with no attribution to Runway required. The practical catch on free is the watermark, which makes those clips fine for testing but not for client delivery — that clears on the paid plans. Security: Where the real risk lives — your prompt, not the license: - If your prompt references a copyrighted character, a real brand, a trademark, or a real person, the output can carry legal risk no matter what Runway's license says. The license covers Runway's grant to you; it can't grant rights Runway doesn't hold. - You're responsible for the rights to anything you upload as a starting image. - Runway's license doesn't override the rules of wherever you publish (a platform can still take a video down). Security: - Generations are processed on Runway's cloud. As with any generator, treat outputs as drafts and check them — AI video still produces artifacts, and "it looked right in the preview" isn't a rights clearance. Best at: - Turning a still image into a short animated clip - Generating stylized or cinematic B-roll from a description - Directed edits — telling it what to change instead of restarting - Fast concepting for social and short-form content Okay at: - Long or perfectly consistent sequences (short clips are the sweet spot) - Exact prompt adherence (expect iteration) - Photorealistic humans without artifacts (improving, not solved) Avoid: - Prompts built on copyrighted characters, brands, or real people - Client delivery from the watermarked free tier - Assuming credits stretch far — a few HQ seconds go quickly Starter: Animate a still image Image-to-video is Runway's most reliable, most beginner-friendly move and shows the quality immediately. Start from your own photo; describe the one motion you want most. Prompt: Take this photo and add gentle, natural motion: drifting clouds, a slight camera push-in, subtle movement in the foreground. Keep it realistic and calm. Starter: A short text-to-video clip Specific, static-ish scenes with one clear motion are what the models do best — a fair first read on text-to-video. Keep it to one subject and one motion; complexity is where clips fall apart. Prompt: A slow cinematic shot of a coffee cup on a wooden table by a rainy window, warm morning light, soft focus, gentle steam rising. Starter: Direct an edit Directed editing — describing the change rather than regenerating from scratch — is Runway's standout, and worth trying early. Make one change at a time so you can see what each instruction does. Prompt: Same scene, but change the time of day to sunset and add a person walking past in the background. Pitfalls: - The free plan doesn't refill. It's a one-time credit deposit with a watermark — a demo, not a workflow. - Credits burn by the second, times your retries. A minute of finished video can cost far more than a minute of credits suggests. - The license is generous; your prompt is where the legal risk is. Copyrighted characters and real people in prompts are the trap. Where to next: - Video Creation & Editing: https://www.launchpadhq.io/discover?category=video-creation-editing - Image Generation & Editing: https://www.launchpadhq.io/discover?category=image-generation-editing ### Getting Started with Sakana AI Canonical: https://www.launchpadhq.io/platform/sakana-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/sakana-ai.json Last reviewed: 2026-09-13 Tool: Sakana AI (https://www.launchpadhq.io/platform/sakana-ai): starting price See pricing; free tier yes; data as of not dated Japan's frontier AI lab, and the free Japanese chatbot it built — which its own help page says only works from inside Japan. You open a page in your browser, type a message into a box (this is called a prompt — the message you type), and an answer comes back in the thread. You keep typing follow-ups in the same conversation. What it is: - Sakana AI is a Tokyo research company founded in 2023 by David Ha, Ren Ito and Llion Jones — one of the three authors' names behind the paper that started the current wave of AI. For its first two years it was known for papers rather than products, and that reputation is out of date. - It now runs three things you can actually use: Sakana Chat, a free chatbot; Sakana Translate, a Japanese–English–Chinese translation page attached to it; and Sakana Fugu, a paid service for programmers that spreads a single request across several different AI models (the AI "brain" that does the actual thinking) instead of using just one. - The reason to come here rather than to ChatGPT is Japanese. The chatbot runs on Namazu, a model Sakana tuned specifically for Japanese writing and Japanese business context. Before you start: - Sakana Chat's own help page says it currently works only from inside Japan — 「現在Sakana Chatは、日本国内からのみご利用いただけます」— and its terms of service reserve the right to block access by IP address. If you are reading this from outside Japan you may simply not be able to sign in, and there is no waitlist to join. Our Text & Conversational AI category is where to find a chatbot you can definitely reach; nothing else on this page depends on you getting past that gate. - The interface is Japanese and the answers come back in Japanese. You can type English at it, but this is a tool built for Japanese first, and treating it as an English chatbot wastes the only thing it does better than the alternatives. - You will not need to pay for the chatbot. It is free, and unusually there is no paid plan to upgrade to — the pricing page states plainly that Sakana Chat has no subscription and that buying a developer plan does not raise its limits. The FAQ says it is free "currently" and that any change would be announced first. - Sakana's FAQ says you do not strictly need an account to send a message, but registering — with a Google account or an email sign-in link — is what saves your conversation history and raises the cap on how many requests you can make in a given period. - The two paid products are separate purchases with separate signups. Fugu, the developer service, has a monthly plan or is billed per use; Marlin, the long-running research agent, is sold only to businesses and billed in credits, with each report using a large batch of them. Neither is currently sold in the EU or EEA. Security: - Where your conversation physically sits is a documented fact here, and it is worth reading precisely rather than taking on trust. Sakana's privacy policy, in force since 19 May 2026, says chat history is stored on Google Cloud Firestore located in Japan and that it does not transfer personal data to third parties outside Japan except where Japanese law permits. If your reason for choosing a Japanese tool is that you want Japanese data residency, that is written down rather than implied. Security: - What is not private by default is training. The terms say Sakana uses your content for machine-learning training and improvement, and that it provides a way to opt out. The opt-out lives in the settings screen under プライバシー (Privacy) — which means it is an account setting, so if you are using the chatbot without registering there is no settings screen in which to turn it off. Security: Two clauses in the terms worth knowing before you rely on it: - You may not use the output to build a rival model, including by distillation or extraction — the terms name that specifically, so anything you generate here is off-limits as training data for a model of your own. - Using it to deliver work that legally requires a qualified professional, without that professional's oversight, is prohibited outright rather than merely discouraged. Security: - Everyday use is not the risk. The value of these documents is that the boundaries are drawn in writing, where a free product usually leaves them to inference. Best at: - Japanese writing, reading and rewriting — this is what the model behind it was built for - Translating between Japanese, English and Chinese, including phrasing that carries cultural weight rather than just vocabulary (Sakana's own August 2026 evaluation claims it beat competing models more than half the time; that is the vendor's own test) - Checking and improving a Japanese draft you have already written, rather than generating one from nothing - Everyday questions that need a current web search folded into the answer Okay at: - English-language work — it will do it, but dozens of tools do it better - Coding and data work — real, but this is not where its strength is - Long documents — file attachments were added in August 2026 and are newer than the rest of the product Avoid: - Generating material to train or fine-tune your own AI — a licence problem rather than a quality one - Any language outside Japanese, English and Chinese in the translation page — those three are the whole supported set, so Spanish or French are not a matter of quality but of absence - Building anything you need to still work next year on the free chat specifically — there is deliberately no paid plan to fall back on if that changes Starter: Proofread a Japanese draft instead of generating one Paste the draft into the translation page and switch it to 添削 (Proofread) mode first. DeepL returns one rendering and stops. Sakana Translate's proofread mode marks its changes against your own draft rather than replacing it, and its third mode takes a follow-up question about the rendering it just produced in the same panel. Paste English and set the target to Japanese to see the same three modes work the other direction. Prompt: [paste your Japanese paragraph here] Starter: Ask it something Japan-specific Roughly: "List five differences in business custom a small Japanese company should watch for when trading with an overseas customer for the first time, briefly." Namazu was post-trained for Japanese and Japan-specific business context and released as its own API in August 2026. Ask the same question in English and compare — the gap between the two answers is the clearest read on what the Japanese tuning bought. Prompt: 日本の中小企業が初めて海外の顧客と取引するときに気をつけるべき商習慣の違いを、5つ挙げて簡潔に説明してください。 Starter: Make it produce a chart rather than describe one Roughly: "Search the web for current information, then summarise the result as a table and a chart, on the topic of how generative-AI adoption in Japan has changed over time." The 13 August 2026 update added running Python in a sandbox and previewing what it produces, and it landed in the free chat. Prompt: 最新の情報をウェブで調べたうえで、結果を表とグラフにまとめてください。テーマは「日本の生成AI利用率の推移」です。 Starter: Fugu, the part that works from outside Japan Fugu reaches programmers through an API (a way for programs to talk to each other without a person clicking) that is deliberately shaped like OpenAI's so existing code needs only the address and key changed. There is no chat box; the console is for keys and monitoring. Fugu is an orchestration layer sold as if it were a model — billed per token, with Fugu Ultra costing several times as much as Fugu Max, priced by how hard you want it to work rather than by which model you picked. Starter: Marlin, the eight-hour report machine you will read about Marlin takes a strategic research question and works on it autonomously for up to eight hours before returning a long structured report, and it is sold only to companies, organisations and sole traders — not to individuals. It is worth knowing what it is because it is the Sakana product that shows up in business coverage. Marlin bills by the report rather than by the seat, which is a different purchase shape from a monthly assistant subscription. Pitfalls: - The three Sakana properties are three separate signups. chat.sakana.ai is the free chatbot, console.sakana.ai is the paid developer platform and marlin.sakana.ai is the business research agent — an account or a subscription on one does nothing on the others. - translate.sakana.ai is not a separate product. It redirects to chat.sakana.ai/translate, which means Sakana Chat's terms — including the Japan-only statement — govern the translation page too, even though the translation interface itself will show you English. - The character limit on the translation box is not the number you will read about. Sakana's August 2026 post describes handling around 5,000 characters, while the counter on the page itself showed a lower figure when we looked — watch the counter rather than the announcement. Where to next: - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai - Research & Academic Tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Sakana AI's product and research blog: https://sakana.ai/blog/ ### Getting Started with Salesforce Einstein Canonical: https://www.launchpadhq.io/platform/salesforce-einstein/guide Data: https://www.launchpadhq.io/data/v1/guides/salesforce-einstein.json Last reviewed: 2026-09-09 Tool: Salesforce Einstein (https://www.launchpadhq.io/platform/salesforce-einstein): starting price $500/mo; free tier no; data as of not dated The AI that already reads your customer records — under a name Salesforce keeps changing. You are not opening a separate AI app. You work inside Salesforce as usual — a customer record, a list of leads — and the AI appears as a button on the record or a panel beside it. You click, or type a request in plain English, and an answer or a draft comes back in the same window. What it is: - Einstein is the name Salesforce gives the AI built into its CRM (the system a sales team keeps its customer records in). It is not one product: it covers older scoring and forecasting features that still carry the Einstein name, the Einstein Trust Layer that governs how your data reaches an outside AI model, and — under a newer name — the generative and agent side of the platform (AI that takes actions on its own rather than only answering). - Since October 2025 Salesforce brands that newer half Agentforce 360, having previously called it Einstein GPT and then Einstein Copilot. The rename is not cosmetic for a beginner: much of what is written about "Einstein" describes something Salesforce now sells under a different name, and Salesforce's own training material still sits at addresses like build-ai-assistants-with-einstein-copilot. - What every part of it has in common is that the AI reads your own account, contact and deal records rather than the open internet — which is the reason a company would let it near a customer's details at all. Before you start: - You can get in without talking to a salesperson. Salesforce's Free Suite is free for up to two people, and Salesforce says it is "completely free with no expiration date, as long as you log in at least once every 60 days". It includes the click-based AI that summarises a record and drafts an email. That is a real starting point, not a demo. - The Einstein-branded predictive features are the expensive half. Sales Cloud Einstein — lead scoring, opportunity scoring, Einstein Forecasting — is documented as available in Performance and Unlimited Editions and "for an extra cost in Enterprise Edition". Those need a paid plan and usually a sales conversation. Nothing in the starter list below depends on them. - There are two other free doors: a 30-day trial of the full paid product, and a free developer account that Salesforce says "doesn't expire as long as you keep using it" and that now includes Agentforce. The developer account is a practice space rather than somewhere to run a business, but it is currently the cheapest way to see the agent builder. - If you do end up paying, plans are priced per user per month, rising steeply from Starter Suite to the top editions, and the agent side is metered separately in Flex Credits — credits are the platform's unit of spend, each thing you do costs some — bought in blocks, so every customer conversation an agent handles has a cost. Current prices are on its Launchpad page. - In a company Salesforce, the licence is not the hard part; learning Salesforce is. Trailhead, which Salesforce calls "the free online learning platform from Salesforce", is where the vendor itself sends beginners, and it hands you a free practice environment to click around in. Security: - The sensitive material here is not the message you type; it is the customer database the AI is already standing on. Salesforce's published answer to that is the Einstein Trust Layer. Security: What Salesforce commits to, in its own words: - Nothing is retained by the model provider. Salesforce describes zero data retention as a policy where "the prompts and generated responses are never stored or used to train the underlying third-party large language models", and says it has "agreements in place with LLM providers, such as OpenAI" containing those commitments. - Your data is not training material. Its Agentforce privacy FAQ states that "Salesforce does not currently use Customer Data to train generative AI models." The word *currently* is Salesforce's, not ours, and it is worth keeping. - Masking has a documented gap. Personal details are meant to be swapped for placeholder text before anything leaves — but Salesforce's own documents say "Data masking is currently disabled for Agentforce to improve performance and accuracy of agents." If you are building an agent rather than clicking a summary button, that particular protection is not in play. Security: - Two practical consequences. It is your employer's Salesforce, so an administrator decides which of these features exist for you at all. And anything an agent does, it does with real permissions against real customer records — which is why Salesforce prices agent activity per action rather than per person. Best at: - Writing the email you were going to write anyway, from facts already on the record — the name, the stage the deal is at, when you last spoke. - Catching you up on a customer nobody has touched in months, in one click on the record instead of reading its whole history. - Filing itself: Einstein Activity Capture pulls emails and meetings onto the matching records. Okay at: - Scoring and forecasting. Genuinely useful once there is years of closed business to learn from, and close to meaningless when there are only a few dozen finished deals. - Answering customers without a person watching. How well it goes depends almost entirely on how good the help documents you point it at are, not on the AI. - Being picked up quickly. The AI part is the easy part; Salesforce itself is the learning curve. Avoid: - Building anything on Einstein Prediction Builder. Salesforce is retiring it and tells customers to migrate 60 days before their subscription end date. - Reading the free two-person Suite as a preview of the agent platform. The agents the marketing is about are metered separately and sit on paid plans. - Documenting a process around a product name. Einstein GPT and Einstein Copilot were both replaced inside roughly two years, and every screenshot and internal note built on them had to be redone. Starter: When a deal has gone quiet, summarise it before you call On an account or opportunity record -> click the summary button and get where the deal actually stands, in a paragraph. HubSpot's built-in AI summarises a record too — the difference is which reader can reach it. Salesforce ships this particular button in Free Suite, so it is the one piece of the Einstein family you can use without a licence conversation, and it reads the record layout your own administrator has already customised rather than a generic one. Run it first on a customer you know well. That is how you find out what it quietly leaves out. Starter: When a reply is overdue, let it draft and you edit On a contact, lead or case -> ask for a draft email and it writes one using that record's own details. Set beside HubSpot's AI email writer, which starts from a blank composer, this one is invoked from the record — and Salesforce documents that what you get changes with your plan. Free Suite gives a fixed Draft with AI click; Starter Suite and above add an assistant you can ask for the draft in your own words. Keep the first sentence yours. The drafts are competent and slightly interchangeable, and the opening line is where that shows. Starter: When the same question keeps arriving, build an agent for it Point Agentforce Builder at your help articles -> it answers that question for customers without a person in the loop. The reason to practise agent-building here rather than in another CRM is the free developer account described above — a permanent, no-cost practice environment, which HubSpot has no equivalent of. Salesforce's documentation describes the builder step by step; we have not driven it ourselves, so treat the click path as their account of it rather than ours. Feed it three real past questions, including the awkward one. The awkward one is the answer. Starter: When email and meetings pile up, stop logging them by hand Connect a mailbox -> Einstein Activity Capture files emails and meetings onto the matching customer records automatically. HubSpot logs email automatically as well. The Salesforce-specific number is the retention split: Salesforce documents that the Standard version keeps roughly three to six months of captured activity on the timeline, while the paid Sales Cloud Einstein version keeps 24 months. Your plan, not your habits, decides how far back a colleague can see. Point it at one mailbox and check what actually lands on a record before switching a whole team on. Pitfalls: - "Einstein" is not one switch. Some features come on with your plan, some are a paid add-on, some are metered separately. Before believing an article, work out whether it is describing something you actually have. - The pricing pages and the help pages use different plan names. Salesforce's public pricing currently lists Free Suite, Starter, Pro, Core, Advanced and Max, while many of its help articles still say Professional, Enterprise, Performance and Unlimited. They overlap, but do not assume an article's "Enterprise Edition" is the price you were quoted. - Flex Credits are consumption, not a seat. An agent left pointed at a busy inbox spends money while nobody is watching it. - You probably cannot switch any of this on yourself. The first move is finding out who administers your company's Salesforce rather than hunting through the settings menu. - It summarises the record you have, not the one you meant to keep. A half-filled opportunity produces a confident-sounding summary of very little, and nothing in the output tells you which fields were empty. Where to next: - Gentler AI inside business tools you can sign up for yourself: https://www.launchpadhq.io/discover?category=ai-plugins-business-software - Connect a simpler CRM to the rest of your apps: https://www.launchpadhq.io/discover?category=workflow-automation - Just want something to talk to first?: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Scale AI Canonical: https://www.launchpadhq.io/platform/scale-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/scale-ai.json Last reviewed: 2026-09-08 Tool: Scale AI (https://www.launchpadhq.io/platform/scale-ai): starting price Custom; free tier no; data as of not dated The company that builds the training data behind the big AI labs. You almost certainly can't buy it — but its free public leaderboards tell you which AI model is currently best at the job you actually have. You open Scale's public leaderboards in a browser, pick a skill you care about — coding, reasoning, voice, safety — and read a ranked table of which AI models (the AI "brain" that does the actual thinking) did best at it. Then you go use whichever one won. What it is: - Scale AI doesn't sell you an AI. It sells the raw material the AI companies need: enormous volumes of carefully labelled, human-checked training data, plus the tests used to grade the results. Its customers are AI labs, large enterprises and governments, on negotiated contracts. If you came looking for something to sign up for and type into, this isn't it. - What Scale gives away is useful, and it's why this page exists. Its research arm, Scale Labs, publishes free public leaderboards at labs.scale.com ranking the well-known models against each other on specific named tests — agentic coding, scientific reasoning, legal and finance work, voice, honesty under pressure. No account, no card, nothing to install. It is one of the better free answers to the question every beginner has and nobody answers well: which one of these should I actually be using? - Read those rankings knowing the company behind them is mid-upheaval. Meta took a large minority stake in June 2025 and founder Alexandr Wang left to run AI at Meta; Scale cut about 14% of staff a month later, and reporting at the time said OpenAI wound down its work with Scale while Google, its largest customer, planned to move on. Francis deSouza, previously COO at Google Cloud, became CEO in August 2026. Alive, funded and shipping — but not the company it was two years ago. Before you start: - Nothing here is something you can buy by clicking. Scale is sales-led: the pricing page lists an Enterprise tier with no prices and a "Book a demo" link, and that demo page is a form to book an intro call. - The free part needs no account. The leaderboards at labs.scale.com are public web pages you read like any article. That's where a beginner should spend their time here. - There's one self-serve exception, and it probably isn't what you want. The pricing page still advertises a pay-as-you-go "Self-Serve Data Engine" billed to a card, with the first 1,000 labelling units and first 10,000 uploaded images free. What that buys is a console for labelling a dataset you already own — useful if you're training your own model, useless if you wanted something to chat with. - One Scale surface an individual can sign up for: Outlier, its platform for paid freelance contributors who train and grade AI models. That's a job, not a product, and it's covered further down. Security: - The usual warning barely applies, because on the path this page recommends you don't type anything in. Reading a leaderboard is ordinary web browsing — no account, no prompt box, no upload. There is nothing to leak. Security: Two places data does change hands:: - Outlier, if you apply to contribute. Onboarding asks for government ID, a résumé and a LinkedIn profile, and the work means reading and grading real AI conversations. That's an identity-and-employment decision, not a signup. Apply from outlier.ai, reached via Scale's own site — not a link someone messaged you. - The enterprise product, if your employer ever buys it. Handing over training data means human contractors read it; that is the service, not a side effect, and it's governed by a negotiated agreement rather than a checkbox. Scale's terms and subprocessor list are at scale.com/legal. Best at: - Answering "which model is best at this specific thing" — these are separate named tests, not one overall score - Benchmarks that look like real work: the Remote Labor Index grades agents (AI that takes actions on its own rather than only answering) Okay at: - Picking a model for ordinary everyday chat — the tests skew hard and technical, and the gaps at the top are small - Being a neutral referee, for the reason in the next column Avoid: - Treating the rankings as independent scorekeeping. Scale sells data and evaluation services to several of the labs on its own leaderboards. It publishes a safeguard — it says it won't license recent data drawn from the live leaderboard's distribution — and it's worth understanding both that the safeguard exists and why it's needed. - Coming here to try the actual product. There is no trial of Scale Data Engine and no free tier of the GenAI portfolio. Starter: Start with the Remote Labor Index Shows how often AI agents finish an entire real freelance job — the kind someone paid a human to do — to a standard a client would accept. Most rankings score models on exam-style questions. Its own history is the useful part: the best agent completed roughly 2.5% of those projects at launch in late 2025, and the top of the board now sits near 16%. Don't read the winner's name — read the failure breakdown. Most misses are quality and incompleteness, not the model being unable to start. Starter: Pick the leaderboard that matches what you'd actually use AI for Takes you from one global "best model" number to the board for your job — agentic coding, scientific and legal reasoning, image and video understanding, voice. A single overall ranking averages things you'll never do. Scale's board is a shelf of separately named tests, each with a page describing what it tested, so "best at refactoring code" and "best at finance questions" are different answers with different winners. Open a safety or honesty board for the same model while you're there — one tests whether it tells the truth under pressure. A model can lead one and lag the other, and that gap is the thing to notice. Starter: Read Showdown — then read who did the voting Ranks models by which answer real people preferred, side by side, without knowing which model wrote which. The votes come from Scale's paid contributor network — 80+ countries and 70+ languages — cast inside Scale's own apps as part of their work, and the results filter by country, language, age and education. That's a different population from an open arena anyone can walk into, and it's why a model can top one board and not the other. Filter to your own country and language first — the regional differences are the interesting part. Starter: If you'd rather work in AI data than buy it, look at Outlier Scale's freelance platform, where people with subject expertise are paid to write problems for AI models, grade their answers, and set the standards those answers are marked against. It also explains the rest of this page: the Showdown votes and much of the data Scale sells come from this workforce. The bar is degree-level expertise plus English. Niche expertise is worth more here than general availability. Pitfalls: - Reading "top AI platform" lists as "things I can use." Scale is a supplier to AI companies; the honest translation of its place on those lists is "top AI company," which is a different thing. - Chasing whoever is first on a board. Gaps at the top are small and the order changes monthly; a model you already pay for and that sits in the top few beats a new subscription for two points. Where to next: - AI APIs & Developer Services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Scenario Canonical: https://www.launchpadhq.io/platform/scenario/guide Data: https://www.launchpadhq.io/data/v1/guides/scenario.json Last reviewed: 2026-09-12 Tool: Scenario (https://www.launchpadhq.io/platform/scenario): starting price Free, from $29/mo; free tier yes; data as of not dated The AI image generator built for game art — where the goal is forty things that match, not one thing that dazzles. You open a panel on the left, pick which AI model (the AI "brain" that does the actual thinking) will draw for you, type a description of the picture you want, and press Generate. The pictures appear on the right. You change a word, generate again, and keep going until one is close. What it is: - Scenario is an AI image generator built for people making game art — characters, props, icons, background tiles, menu screens — rather than for people making one striking picture to post. - Its distinguishing feature is that you can train it on your own artwork: give it reference images sharing one style and it builds a private model of that style, so afterwards a short prompt (the message you type) matches everything else you have made. Scenario's guide puts a workable style set at roughly 10–20 images, and 5–15 for a single character or object. - You pick it over a general image generator when the problem is "these forty icons need to look like a set," not "I want one great picture." It also makes video, 3D meshes and audio from the same workspace, but images are where to start. Before you start: - Free to try without a card. Sign up at scenario.com and you get a daily allowance of credits — the platform's unit of spend, where each thing you make costs some (Scenario's help pages also call them Compute Units). They reset daily rather than piling up, and what a single image costs depends on the model, so a free day buys a modest number of pictures. On paid plans the allowance renews on your billing date and unused credits do not roll over. - The custom-style training Scenario is known for is not on the free tier, and not on the cheapest paid tier either. Its plan comparison table currently puts "Train custom models" on the Pro plan, while the cheaper Starter plan lists everything except training. By Scenario's own published figures a single training run costs far more credits than an image, so even setting the plan gate aside, a free day's allowance could not pay for one. - What will actually decide this for you is not which plan to pick but whether you have a consistent body of your own artwork to train on. Without that, Scenario is a capable general image generator that bills you per picture, and there are free ones in the same category on this site. Security: - Scenario asks you to upload your own artwork, which makes what it then does with that artwork the important question — and Scenario's public pages answer it two different ways. Its marketing pages say your data stays yours and is never used to improve other models. Its privacy policy, last updated 16 July 2026, says of ordinary accounts: "To operate, secure, and improve the Services and our AI models, we may use your content and aggregated, de-identified usage data." In plain words: on a website sign-up, Scenario reserves the right to use what you upload and make to improve its own models. The promise not to do that is written for enterprise customers with a signed contract, and no opt-out is described for anyone else. Where a marketing page and a privacy policy disagree, the policy is the one that binds. Security: Three things to settle before you upload anything: - Your prompts and generated assets are kept while the account is open and deleted within 60 days of closing it, per the same privacy policy. - Whether your free-tier work is visible to anyone else is not stated anywhere public. Scenario's pricing table lists "Private generations" as a feature of the paid Starter plan and shows no free column at all, which is suggestive but not an answer. Check the setting once you are signed in rather than assuming either way. - Whether you may sell what the free tier makes is contradictory on Scenario's own pages. The pricing page's FAQ says "Free plan outputs are for personal and evaluation use only." The help centre says all generations from Scenario's ready-made models "are cleared for commercial use." The terms and conditions in force today — v2.1.1, last updated 16 July 2026 — say "You own your Generated Assets" and draw no line between plans, while separately letting Scenario impose feature restrictions on free accounts. If money depends on the answer, the paid plans are the ones that state a commercial licence in writing. Best at: - Producing many images that belong to the same set rather than one showpiece. - Working from pictures as well as words — you can attach reference images and slide how strongly each one pulls the result. - Carrying an asset past the picture stage: background removal, enlarging and image-to-3D-mesh all sit in the same workspace on the same credit balance. Okay at: - One-off images. The ready-made models are the same families other tools use, but you are spending per picture on machinery built for consistency. - Video and 3D generation, which the vendor's own cost guide places in a higher credit band than images, so experiments there run down a free day fast. Avoid: - Training a style model on artwork you did not make. Scenario's FAQ puts the responsibility on you to hold "full rights to the training data you upload," the terms bar uploads that infringe someone else's rights, and the indemnity clause leaves any claim with you. Cloning an artist's portfolio is the wrong use of the platform's main feature, and the one it makes easiest. - Realistic pictures of real, identifiable people. The terms prohibit generating recognisable depictions of real individuals without their consent, sexual or not. - Throwing a big untidy folder at the trainer. Scenario's guidance is that a small, well-curated set beats a large varied one, with the variety in subject and camera angle while the style stays fixed — the reverse of what most people upload first. Starter: One icon, framed so the next thirty match Scenario prints the credit cost of the run on the Generate button before you press it, and that number changes the moment you pick a different model above the prompt field — so a small icon shows as a cheap experiment before you commit. Change only the noun — "shield", "map scroll", "iron key". Repeating the framing words verbatim is what turns separate pictures into a set. Prompt: A single fantasy health potion bottle, game inventory icon, centred on a plain flat background, soft rim light, clean silhouette, no text, no border. Starter: The same sentence through three different models The model selector sits directly above the prompt field and Scenario's library runs to hundreds, so one fixed sentence is how you find which one draws the way your project needs — and the choice is not cosmetic, because a style model you train later sits on top of it and inherits its habits. Keep the sentence identical between runs. Edit the words and the model at once and you have learned nothing about either. Prompt: A stout dwarven blacksmith standing at a forge, three-quarter view, waist up, painterly, warm firelight. Starter: Turn a picture into a game asset Takes an image you have already generated and converts it into a 3D mesh or a short video, from a menu on the image itself. For a game the 2D concept is rarely the deliverable, and most image generators stop at the picture. Scenario keeps picture-to-3D in the same workspace on the same credit balance rather than sending you to a second tool with a second bill. Converting to 3D works best from the image you would have drawn for the job anyway: one object, plain background, no dramatic lighting hiding its shape. Starter: Train the style, then write less Turns a folder of your own artwork into a private model that draws in that style. Scenario's documented steps: crop the images square, let it caption them, choose which ready-made model to build on top of, press Start Training. It emails you when the run finishes and shows progress under a Recent Tasks icon. Scenario's prompting guide says custom models want the opposite of what you just learned: shorter descriptions, shaped like the captions used in training, because the model weights the earliest words most. Carrying a long ornate description over from a ready-made model is the usual reason a freshly trained style looks worse than what you trained it on. If the results look like an averaged blur rather than your style, the variety is in the wrong place — see the last Avoid item. Starter: Ask for the size you are going to ship The Dimensions control sits in the same panel as the prompt and costs nothing extra to set, while enlarging an image afterwards is a separately billed action that can cost more credits than the picture did in the first place. Swap "stone dungeon floor" for any surface you need repeated — grass, timber decking, cracked ice — and keep "seamless" and "tileable"; they do most of the work. Prompt: A seamless stone dungeon floor texture, top-down view, even lighting, no shadows, no objects, tileable. Pitfalls: - Credits are spent on the attempt, not the result. A batch of four that all miss costs what four keepers cost, so generate one at a time while you are still hunting for the wording. - Judging Scenario on a free session. The free tier gives you the ready-made models, not the trained-style machinery described above — so "this is just another image generator" is a fair verdict on the free tier and an unfair one on the product. - Forgetting which ready-made model your own style was built on. Its habits — how it handles faces, edges, text — persist through everything the trained model makes afterwards. Write down which one you picked. - Treating the model library as the hard part. The four or five framing words you repeat in every description change your output far more than switching between two models trained on similar pictures. Where to next: - Image generation and editing: https://www.launchpadhq.io/discover?category=image-generation-editing - Gaming and creative AI: https://www.launchpadhq.io/discover?category=gaming-creative-ai - Video creation and editing: https://www.launchpadhq.io/discover?category=video-creation-editing ### Getting Started with Scholarcy Canonical: https://www.launchpadhq.io/platform/scholarcy/guide Data: https://www.launchpadhq.io/data/v1/guides/scholarcy.json Last reviewed: 2026-09-15 Tool: Scholarcy (https://www.launchpadhq.io/platform/scholarcy): starting price Free, from See pricing; free tier yes; data as of not dated Turn a research paper into a summary card you can actually skim. You drag a PDF onto the page, paste a link, or click a browser button while a paper is open. Scholarcy reads the document and hands back a card: the main finding, the key claims, the numbers, the references. You scroll it, jump to the bit you want, and swap the summary to a different level of detail. What it is: - Scholarcy is a summarizer built specifically for academic papers. You give it a paper; it gives you back a Summary Flashcard (Scholarcy's name for the card it builds — one paper, one card, split into sections you can jump between). - It is not a chatbot. There is no message box and nothing to write — you import a document and pick from a short list of menus. That makes it far narrower than a general AI assistant, and far quicker at the one thing it does. - There are two front doors. The free Article Summarizer lives at library.scholarcy.com/try and summarizes on the spot. Scholarcy Plus is a paid Library at library.scholarcy.com where imported papers stay put, with notes, highlights and collections. Before you start: - The free tool needs no card, and its entry link goes to a try page rather than a signup page, so you should be able to drop a PDF in and see what comes out before deciding anything — though you may be asked to sign up after the first one. Scholarcy Plus is a monthly or yearly subscription, and starts with a one-week free trial that bills automatically at the end unless you cancel. Scholarcy's terms say it does not refund subscription fees unless legally required, so cancel before the week is up if you are only testing. - How much you get free is stated three different ways on Scholarcy's own site, so verify it in the app before you plan around it. The pricing comparison table, the FAQ further down that same page and the Article Summarizer page's own comparison table each give a different free limit — one a flat total, two a daily cap, and the two daily caps don't match. All three were live on 15 September 2026. The vendor's pages disagree with each other, so take the free tier as small — a handful of summaries a day at most, enough to evaluate the tool and not enough to run a literature review through — and confirm the current number once you are in. - File size caps at 11MB on a free account and 50MB on a paid one. Scholarcy is built for one paper or one book chapter at a time, not whole books — its own example is a 400-page textbook reduced to a 400-word summary, one word per page. Split anything long into chapters before you upload. - Language support is partial in a way that matters. Scholarcy says it summarizes articles in most European languages but does not publish the list, so whether yours is covered is something you will have to test with one paper. Only some of the output stays in the original either way: the highlights, the summary, the comparative analysis and the full text come back in the paper's own language, while the plain-language Synopsis and every enhanced summary mode are rewritten in English regardless of what you uploaded, because of the outside services Scholarcy routes them through. If you are reading a Spanish or German paper and want a Spanish or German summary, the most readable outputs will not be. Security: - What happens to your PDF is the question worth asking here, and Scholarcy answers it concretely: its FAQ states it never stores imported documents for long-term use and deletes the PDF once the flashcard is built, and it says it stores data on third-party servers in Europe. Its privacy policy — effective 1 January 2020, last revised 7 February 2024 — is the version in force today. Security: Worth knowing before you upload: - The browser extension reads the text or the URL of the page you are on; that is how it produces a summary, so while it is active it sees what you are reading. - The Chrome Web Store listing discloses that the extension handles location and website content, and states the data is not sold to third parties or used for unrelated purposes. - Parts of the summary are generated by outside services Scholarcy does not own, so your text does not stay inside Scholarcy alone. - Neither the terms of service nor the privacy policy says whether your uploads are used to improve the service. They are silent on it, which is not the same as a promise that they are not, so treat silence as silence. - Scholarcy says it uses "end-to-end encryption methods" but does not define what that covers; it is not the same claim as end-to-end encrypted messaging and should not be read as one. Security: - Uploading a paywalled paper you legitimately have access to is fine by Scholarcy's account: it describes itself as a summarizing and management tool rather than a distribution service and does not keep the PDF. What you may do with the result is narrower than people assume — the terms grant you a perpetual, royalty-free licence over the output for personal use only and bar commercial distribution, and your library's or publisher's own licence still governs the source paper. Best at: - Deciding in two minutes whether a paper is worth reading properly - Pulling the claims, numbers, figures and tables out of a dense results section - Turning a paper's reference list into a reading list — it links each citation out to a freely readable version where it can find one - Working through a stack of papers you have already gathered, where you want the same questions asked of each one Okay at: - Book chapters — good one chapter at a time, poor on anything book-length - Online articles and YouTube videos — supported, but the extraction is tuned for academic structure, so a news piece or a talk yields a thinner card than a paper does Avoid: - Trusting how confident the card sounds. Scholarcy's own page says its output is factually correct and traceable to the source; its terms of service disclaim any warranty that the service will meet your requirements. Both can be true, because the thing that goes missing is rarely a fact — it is the hedge around it. "The effect was small, in one sample, and needs replication" compresses into a flat claim, and the flashcard gives you no signal that a qualifier was dropped. Open the sentence a claim came from before you repeat it. - Citing or quoting a paper you have only met as a flashcard. The card is an index into the paper — it tells you where to look. It is not a source in its own right, and a quotation lifted from a summary is a quotation from Scholarcy. - Looking for papers with it. Scholarcy only ever works on documents you hand it; it does not search, and it will not tell you which papers exist. If you do not already know what you need to read, this is the wrong end of the process. Starter: Summarize a paper you already know well Open the free Article Summarizer at library.scholarcy.com/try, drag in a PDF of a paper whose conclusions you could already state out loud, and read the card against your own memory of it. Scholarcy's card is broken into fixed, named sections — highlights, synopsis, claims, figures, references — rather than one block of prose, so a paper you know lets you audit each section separately and learn which ones you can rely on. Against a paper you have not read, a dropped caveat is invisible. Then try one from outside your field and notice how much harder the card is to sanity-check. Starter: Find the Enhance menu before you decide the summary is wrong On a generated summary, click Enhance and pick a reading style: School level, General reader, Bulleted list, Undergrad level, One liner or Researcher. You can also choose Specific topics and select up to five keywords to point the summary at. Six named styles and five keywords, chosen from menus, are the whole of Scholarcy's steering. Check what you actually get, too: Scholarcy's pricing table lists enhanced summaries as a Scholarcy Plus feature while its own FAQ explains Enhance as though any user can reach it. The two vendor pages disagree, which is not resolvable from outside, so the app is the tiebreaker. Scholarcy describes these as reading styles rather than lengths, which suggests the shorter ones drop categories of detail rather than merely compressing them — run one paper at two styles and compare before you rely on the short one. Starter: Mine the references, not just the summary Scroll to the references section of the card. Scholarcy tries to link each cited work out to a version you can read for free, so you can open the papers this one is built on without going back through a subscription wall. Chasing down freely readable copies of the exact works cited is a step Scholarcy runs as part of building the card — its own extension listing describes linking to open versions of each citation. Tools that suggest "related papers" are searching a database for things like your paper; this is the actual bibliography of the paper in front of you. Start from a recent review article — its reference list is a ready-made syllabus. Starter: Install the extension and summarize without downloading anything Add the Scholarcy extension from the Chrome, Edge or Firefox store — Safari is listed as coming soon — then click it while a paper is open in your browser. It returns the card without you leaving the page. The extension and the upload path are not interchangeable, and knowing which one failed saves you a wasted conclusion. The extension works from whatever the page will give it, so a subscription-only PDF that summarizes perfectly well after you download it can come back empty through the extension. When that happens, download the PDF and upload it instead of assuming the paper cannot be summarized. The extension appears to have some free allowance of its own, since the Chrome Web Store listing sells "unlimited extension use" as a paid benefit — but Scholarcy does not publish what the free limit is, or whether it draws on the same allowance as the web summarizer, so watch what it lets you do. Starter: Line several papers up side by side (Plus only) With Scholarcy Plus, import several papers into the Library and generate a Literature Matrix — a grid that puts the papers in rows and what each one did and found in columns. Free accounts cannot create one. Paying turns on this one feature, and the honest test is simple: the matrix is built only from papers you imported yourself, so it is a filing system rather than a discovery one. If you already have the stack, it earns its money; if you are still assembling it, nothing here will help you find the next paper. Pitfalls: - An empty card is a known outcome, not something you did wrong. Scholarcy keeps a dedicated help article for imports that finish with a blank record. The usual causes: the PDF is a scan, so its pages are pictures of text with nothing to extract; the URL is behind a subscription wall and returns nothing; or the file is too big or too image-heavy. For a scan, run it through text-recognition software first — Scholarcy does not do that step for you. - On the free tool, closing the tab loses the summary. There is no library holding it and saving flashcards is a Scholarcy Plus feature; free users can export one flashcard at a time, and paid users up to 100 at once. Export as you go, or plan to re-run the paper. - Read the highlight-colour key once. Scholarcy colour-codes the terms, claims and findings it pulls out and documents the scheme in its help centre, but nothing in the interface tells you that the colours mean different things. Until you look it up the card reads as undifferentiated highlighting and you lose most of what it is signalling. - The marketing numbers on Scholarcy's own site do not agree with each other — 600,000 users on one page, 400,000 on another. It is not a reason to distrust the summaries, but it is a reason to confirm any number you are planning around inside the app rather than taking it off a landing page. Where to next: - Document and PDF tools: https://www.launchpadhq.io/discover?category=document-pdf-processing - Learning tools: https://www.launchpadhq.io/discover?category=education-learning-ai - AI chatbots: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with SciSpace Canonical: https://www.launchpadhq.io/platform/scispace/guide Data: https://www.launchpadhq.io/data/v1/guides/scispace.json Last reviewed: 2026-09-15 Tool: SciSpace (https://www.launchpadhq.io/platform/scispace): starting price Free, from $20/mo; free tier yes; data as of not dated Upload a research paper, ask it questions, and get answers that point back at the page. You upload a paper as a PDF, or search for one, and a chat box appears alongside it. You type a question about the paper and the answer comes back with small numbered links you click to jump to the sentence it came from. Highlight any passage to get a plainer version of it. What it is: - SciSpace is an AI research assistant built around reading papers. You either point it at a PDF you already have and ask questions about it, or search its index of published papers and pull what each one found into a side-by-side table. - You will meet two old names. It was Typeset.io until 2022 — typeset.io still redirects here — and its in-paper reading assistant was called Copilot before it became Chat with PDF. Tutorials and reviews written even a year ago use menu names that no longer exist, so check a walkthrough's date before you conclude a feature was removed. - It is run by PubGenius Inc., and it sits alongside your library rather than in place of it. Before you start: - Signing up is free and takes no card. The Basic plan is the free tier at scispace.com, and it is where everyone starts. - The free ceiling is worth understanding before you invest an afternoon in it. Basic includes a small monthly credit allowance for SciSpace Agent tasks — *credits* are the platform's unit of spend, each thing you run costs some — and unused ones expire at the end of each billing cycle rather than rolling over. An agent here means AI that goes off and takes actions on its own rather than only answering. Separately, SciSpace's plan table marks chats with PDFs, paraphraser outputs, citation generations, topic searches and AI detections as "Limited" on Basic without publishing the numbers behind that word — while its individual feature pages describe several of those same tools as simply "free". The two do not agree, SciSpace does not reconcile them anywhere public, and there is no documented reset schedule for those caps the way there is for the monthly credits — so you will most likely find out where they sit by reaching one. - Exports are the one hard wall on the free plan. Basic is listed as "No Exports allowed" — no CSV, Excel, BIB, RIS or XML out of any tool, including the comparison table you just spent an hour building. Anything you want to keep has to be copied out by hand until you pay. - Paying for Premium (cheaper if billed annually) turns the "Limited" rows into unlimited ones. Deep Review — the deeper literature-analysis mode most reviews write about — sat on the pricier Advanced plan rather than on Premium, in the most recent plan table we were able to read, which dates from January 2026; SciSpace's live table is drawn by the browser and its own Deep Review help article names no plan at all. Treat the per-tier rows as current in shape, check the numbers on the pricing page before you pay, and confirm which tier Deep Review is on before upgrading for it specifically. Security: - Read the feature page and the binding policy in that order, because they do not say the same thing. SciSpace's Chat with PDF pages promise that uploaded files are encrypted at rest, travel over an encrypted connection, are "never used for model training", stay private to your account, are auto-purged after 30 days of inactivity, and can be deleted from your dashboard on demand. That is more than most tools in this category commit to in writing. Security: - The policy says less. SciSpace's privacy policy — effective 10 April 2024, and still the version in force today — does not carry that no-training promise for the platform as a whole. It describes using data to "research, develop and improve our AI models" in the context of its synthetic audio and video features, and it names advertising partners and data brokers among the categories of party it shares personal data with. We could not find anywhere SciSpace reconciles the two, and where a feature page and a policy differ the policy is the one you would be held to, so read the no-training line as scoped to Chat with PDF specifically rather than to your whole account. Security: Worth deciding before you upload: - Whether you actually hold the rights to the PDF — a paper licensed to you through your institution is not always yours to put on a third party's server - Whether unpublished work belongs here at all: manuscripts under review, grant drafts and anything covered by a confidentiality agreement are a conversation with your supervisor or research office first - Deleting files when you are done with them, rather than letting the 30-day inactivity purge do it, since "inactivity" restarts every time you open the file Security: - For a published, open-access paper none of this is a real risk. The care is entirely for work that is not yours or not out yet. Best at: - Getting oriented fast in a paper you did not choose — what it did, what it found, and what it quietly did not test - Explaining one specific paragraph, equation, table or figure in plainer words without leaving the PDF - Turning a research question into a first list of candidate papers, then into a table of what each one says - Working in a language you do not read well — SciSpace supports asking and answering in 75+ languages Okay at: - Scanned and photocopied PDFs. It has to read the text off the image first, and SciSpace puts that at 90%+ accuracy, rising to 98% on clean high-resolution scans — its own figures, which we found no independent test of, and which still leave a real margin on an old scan - Telling you how much it actually has. SciSpace's homepage currently claims 280M+ papers while its own Literature Review page claims 200M+, both checked on the same day. That is a wide gap for one product, neither figure is dated, and either is better taken as a rough order of magnitude than quoted - Writing. There is an AI Writer and a paraphraser, but this is where the tool and your institution's rules on AI-assisted work meet, and the rules win Avoid: - Taking a number out of an extracted comparison table as though it were the paper's number. SciSpace's own guidance for its writing tools is "verify every citation against the source" — that applies just as hard to a value it lifted out of a results table, where a units error or a pulled-from-the-wrong-row error looks identical to a correct answer - Using the AI detector to establish that someone used AI. SciSpace's own help page says manual verification is recommended for a final judgement, and that human writing containing "excessive AI-like structures" can be flagged - Assuming its search is complete. SciSpace's published benchmark claims Deep Review returns more *relevant* papers per query than Elicit or Consensus — that is a relevance claim, not a coverage one. Nothing the vendor publishes says it finds everything, and a review you will have to defend still needs a database search behind it Starter: Ask the paper the question you actually have SciSpace publishes both halves of this: a long paper is split into chunks for it automatically, with no length limit on the text itself, and the answer is written from the chunks nearest your question. Aim, not length, is what you control — a question pointed at one thing matches one passage, while "summarise this" matches everything and averages it. Ask the same question again with "and what did it not test?" appended. The gaps are usually in the limitations section nobody reads. Prompt: What did this paper actually measure, how many participants or samples, and what was the single main result? Quote the sentence each part of your answer comes from. Starter: Highlight the paragraph that lost you Select any confusing passage inside the PDF with your cursor. Alongside the plainer wording, SciSpace offers papers related to what you highlighted. Highlighting a method you have never met is also how you find the paper that introduced it, because those recommendations come out of SciSpace's own index of published papers rather than out of the PDF in front of you — which is the half a standalone PDF chat tool cannot do. Highlight a method name rather than a conclusion. Conclusions are where you already know what you think. Starter: Make it read the table and the equation SciSpace names equations and tables as things you can select and have explained, not just prose — its own FAQ calls this out for maths, physics and chemistry papers. Then open the table yourself and check two rows against what it said. Two is enough to tell you how much to trust the rest. Prompt: Explain the main results table row by row in plain language. For each row, say what was compared, what the number means, and whether the difference is one the authors themselves call significant. Starter: Search the literature as a question, not as keywords SciSpace's Literature Review page gives a documented instruction most people never read: if you are asking a question, put a question mark on the end and the results improve. It is the rare search tip that comes from the vendor rather than from a blog guessing. Swap in your own question. Keep it to one comparison — a question with two "ands" in it returns a worse list than two separate searches. Prompt: Does intermittent fasting improve blood pressure in adults with hypertension? Starter: Build the comparison table, then notice where it stops From a set of results or your own uploaded PDFs, add columns for the things you care about — sample size, method, outcome, limitations — and let it fill the grid across every paper at once. The column cap is the free plan's real shape: Basic stops at 5 columns including the ones SciSpace suggests for you, Premium goes to 50. Five is enough to learn whether the extraction is trustworthy on your topic and not enough to finish a review, which is a useful thing to discover in twenty minutes rather than at the end. Make one column "page number for this claim". It costs you a column and turns the table into something you can check. Pitfalls: - PDFs only, 100 MB each. SciSpace says Word, PowerPoint and video support is planned but not here, so a slide deck or a .docx has to be converted before it will go in. - It answers from the document in front of it, not from the field. Ask about a section your paper does not have and you can still get back something that reads like an answer. - Credits drain faster than the allowance sounds. SciSpace's own help page explains that one Agent run fans out into dozens of sub-tasks and shows a user spending nearly a month's allowance on two runs — so the Agent is the thing to spend deliberately, while Chat with PDF and Literature Review do not draw on credits at all. - Free answers come from SciSpace's standard AI engine; the higher-quality one is a paid unlock listed on the plan table. If an answer feels shallower than the paper deserves, the plan is a likelier cause than the paper. - The paraphraser's limit on how much text it will take per run sits behind a collapsed FAQ we could not read, so you may need to check that one once you are signed in. Where to next: - Research and academic tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Document and PDF tools: https://www.launchpadhq.io/discover?category=document-pdf-processing - Learning with AI: https://www.launchpadhq.io/discover?category=education-learning-ai ### Getting Started with Scite Canonical: https://www.launchpadhq.io/platform/scite/guide Data: https://www.launchpadhq.io/data/v1/guides/scite.json Last reviewed: 2026-09-15 Tool: Scite (https://www.launchpadhq.io/platform/scite): starting price Free, from $20/mo; free tier yes; data as of not dated Checking whether a scientific paper has been backed up or shot down by the research that came after it. You type a paper's title or a research question into Scite's search box. Back comes the paper plus a breakdown of how later papers cited it, with the actual sentences quoted. You click a sentence to jump to the paper it came from, filter down to just the disagreements, and search again from there. What it is: - Scite is a search tool for scientific papers that does one unusual thing: rather than only counting how many times a paper has been cited, it reads the sentence around each citation and sorts it into supporting, contrasting or mentioning. - Scite calls these Smart Citations. A paper with 400 citations that are nearly all passing mentions is a different animal from one with 40 citations where six later studies contradicted it — and a raw citation count hides that difference completely. - It also has Scite Assistant, a chat box where you ask a research question in plain English and get an answer with links to the papers it drew from. - Scite was bought by Research Solutions in November 2023 and now trades as "Scite, LLC, a Research Solutions Company." It is still sold directly to individuals; the acquisition did not move it behind an institution-only door. Before you start: - This is not free in the way most things on this site are. Scite's free "Connect" plan gives you a small monthly allowance of credits (the platform's unit of spend — each thing you do costs some) for a connector that pipes Scite's data into ChatGPT, Claude or Gemini. It does not get you into Scite's own site: the pricing table lists Connect as having no Assistant and no search. Scite's Assistant page separately says you start a free trial "to save and access your chat history", which reads as though you might get a question or two in without one — the two pages disagree, so assume you will need the trial. - Paid access starts with Basic, billed a year at a time. Scite's pricing page shows only the annual rate; paying month to month costs more, but the page does not print that figure. Pro adds bigger collections plus patent, clinical trial and grant data. There is a 7-day free trial and you create an account to start it. - Before paying, check whether your university or employer already subscribes — Scite sells institutional plans through a sales team, and library access is how many people get in. Scite's pricing page also points students and academics at its support address for a discount. - The coverage gap matters more than the price. Scite indexes around 317 million articles but holds only about 41 million of them in full text, and a Smart Citation can only be pulled from full text it holds. Its published partner list — 35-plus publishers, last updated June 2026 — includes Wiley, SAGE, Cambridge University Press and the Royal Society of Chemistry, but does not name Elsevier, Springer Nature, Taylor & Francis, IEEE or Oxford University Press. Whole fields will look thinner on Scite than they are. Scite says it adds partners over time, so a publisher missing today may be covered later; check the coverage page rather than trusting a list you read once. - The browser extension is a separate, free thing and runs on Chrome, Firefox, Edge and Safari. Scite describes it as a free extension; you may need to sign in to a free account once it is installed, which is worth checking on the day. Security: - Scite's privacy policy, effective 26 March 2026, says in so many words that Scite does "not use Customer Data to train, fine-tune, retrain, or otherwise improve" its AI systems, and its terms repeat the promise. For a tool where your searches reveal what you are working on before you publish it, that is worth more than it sounds. Security: What still applies: - Your searches, Assistant conversations and any documents you upload sit on Scite's servers, not your machine - The same policy says account data is kept for up to 10 years after you close the account - Deletion goes through Research Solutions customer support rather than a button in your settings, as far as the public policy describes Security: - Scite's terms also limit what you may do with its classifications: you may not use its Classifier Results to build anything that competes with or substitutes for Scite's own products, and you may not create derivative works from them. If you were planning to bulk-export supporting and contrasting counts into a database or a paper of your own, read that clause first — noting that those terms carried no visible effective or last-updated date when we checked on 15 September 2026, so which version governs is not something the public page will tell you. Best at: - Sanity-checking a paper you are about to build on — one glance at whether anything contrasts it - Finding where a literature disagrees with itself, by filtering one paper's citations down to contrasting only - Catching retractions (a journal formally withdrawing a paper) and editorial notices on work you have already collected Okay at: - General literature searching — it works, but the coverage gap above shapes what comes back - Papers published in the last few months — nothing has cited them yet, so there is nothing for Smart Citations to classify Avoid: - Treating "contrasting: 0" as a clean bill of health. Far more often it means nobody has cited that paper inside the full text Scite holds — which, for the publishers named above as missing, is a great many papers. - Reading the supporting and contrasting labels as verdicts. A machine learning model (the AI "brain" doing the work) assigns them from a single sentence, and Scite's own help documentation describes a route for users to flag a misclassification for two experts to review. That route exists because the labels are sometimes wrong. Starter: Start with a paper you already half-trust Search Scite for a paper you have cited before or are about to cite, and read the supporting, mentioning and contrasting split at the top of its report before you read anything else on the page. Reading the split before the abstract inverts the usual order — you find out whether later work pushed back before you have formed a view of the paper. A Google Scholar count cannot answer that question at any point in the process. Try it on a paper you already know is contested and see whether the split matches what you know. Starter: Filter down to just the contrasting citations On any report, set the classification filter to contrasting only, then read the quoted sentences that remain. Contrasting is by far the smallest of the three buckets — on Scite's own published paper it sits under one percent of citations — so this one filter turns thousands of citations into a readable handful. Starter: Ask the Assistant a question instead of typing keywords Scite Assistant answers out of papers it holds in full text, including paywalled ones covered by its publisher deals, so asking for both sides at once is asking it to do the single thing its index is built for. Swap in your own question. The clause worth keeping is "and any studies that contradict them." Prompt: What is the current evidence on whether intermittent fasting improves insulin sensitivity in adults without diabetes? Give me the strongest supporting studies, and any studies that contradict them, with citations. Starter: Put the badges on the search you already use Install the Scite browser extension, then run a normal PubMed or Google Scholar search; Scite's badges appear alongside the results, and a panel appears at the side when you open a paper. It injects Scite's classification into PubMed and Google Scholar results — pages Scite does not own — so the split reaches you inside the search tool you already use, with no second lookup. Starter: Run a finished reference list through Reference Check Upload a manuscript PDF to Reference Check; Scite reports how each reference in it has been cited by others, flagging any that have been retracted or picked up editorial notices. A retraction is the failure nobody catches by rereading their own bibliography, because the entry still looks fine; this checks the list you already wrote rather than helping you build a new one. Scite's feature pages describe Reference Check but its pricing table does not say which plan includes it, so expect to need the trial or a paid plan. Pitfalls: - Mistaking the Assistant for the citation data. They are two different things on one site: the Assistant writes you an answer, while Smart Citations hand you raw classified sentences. The second is what Scite is actually distinctive for, and it is the one people skip. - Assuming "supporting" means somebody replicated the original. Scite's help documentation says a citation can count as supporting when the conclusions line up, even where the method, model or conditions were different. - Reading Scite's marketing figures as coverage. Its Assistant page describes "317M+ full-text articles", which does not square with the article-versus-full-text split on Scite's own data-coverage page. The coverage page is the specific one; trust it. - Judging the paid product by a public report page. One currently opens to anyone with the three counts visible, but the pricing table lists article Smart Citation reports as a Basic-plan feature — so how much of a report you can read without paying may already have narrowed and may narrow further. Treat the public page as a preview rather than the product. Where to next: - More research tools: https://www.launchpadhq.io/discover?category=research-academic-tools - AI chat assistants: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Semantic Scholar Canonical: https://www.launchpadhq.io/platform/semantic-scholar/guide Data: https://www.launchpadhq.io/data/v1/guides/semantic-scholar.json Last reviewed: 2026-09-15 Tool: Semantic Scholar (https://www.launchpadhq.io/platform/semantic-scholar): starting price Free; free tier yes; data as of not dated A free, non-profit academic search engine — with AI features that cover far less than the search does. You type a research topic or a paper title into a search box and get a ranked list of papers. Open one and you see its summary, who cited it, and what it cited. Follow those links outward, save what matters, and the connections between papers become the thing you are actually browsing. What it is: - Semantic Scholar is a free academic search engine run by Ai2, the Seattle non-profit AI research institute founded in 2014 by the late Microsoft co-founder Paul Allen. Semantic Scholar itself launched in 2015. - Its index spans all fields of science, assembled from partnerships with 50-plus publishers and scholarly societies, from data providers such as PubMed, arXiv and Springer Nature, and from papers found on the open web. - The AI in it is not a chatbot. It ranks results by meaning rather than exact keyword matches, writes one-sentence paper summaries, classifies why one paper cited another, and recommends new papers based on what you save. - It is mature rather than fast-moving. Its own release-notes page carries no entry newer than early 2024 — and the dates on that page are internally inconsistent, so read that as approximate rather than exact — while Ai2's newer research-agent work went into a separate product, Asta, launched in August 2025 and built on Semantic Scholar's own data. Nothing public says whether Ai2 intends to keep developing both. There is no shutdown notice and the service runs normally, so the honest position is that Semantic Scholar is stable and well-maintained rather than actively growing. Before you start: - Nothing here costs money and nothing is being sold to you. Ai2 is a non-profit, and Semantic Scholar has no paid tier, no credits and no card field anywhere — so the will-I-need-to-pay question has a one-word answer, which is no. - You can search, open papers and read summaries without signing in at all. A free account — email, Google, or your institution — adds three things: saved library folders, email alerts, and the research feed. If your university subscribes to journals, signing in through your institution (Semantic Scholar uses OpenAthens for this) also surfaces full text you already have access to. - The AI extras cover far less of the index than the search does. The one-sentence summaries Semantic Scholar calls TLDRs are currently limited to computer science and biomedical papers; topic pages to computer science; and Semantic Reader, its in-browser paper reader, to papers from arXiv (a free site where physics, maths and computer-science papers are posted, often before formal publication). Search everything; expect the extras on a slice of it. - Ask This Paper — the one place on the site where you type a question instead of keywords — is documented by Semantic Scholar in its FAQ and its December 2023 release notes, but paper pages block automated visits, so this page cannot confirm the box still renders. Semantic Scholar also describes it as tested only on English-language papers and available on a limited set of them. Treat it as something to look for on the paper you open rather than something guaranteed to be there. Security: - This is one of the least worrying tools on this site. Ai2 is a non-profit, there is no card on file, and Semantic Scholar's FAQ states it does not collect sign-in information such as your username or password — accounts are handled by Google, by your institution, or by a reset link. Ai2's privacy policy, last updated 19 February 2025, is the document that governs. Security: - The disclosure worth knowing is that two features send text to a third party. Semantic Scholar's FAQ says Ask This Paper queries OpenAI's API with the paper's content alongside your question, and that the on-click term definitions in Semantic Reader ask ChatGPT. The specific model named on that FAQ page looks out of date, so the durable fact is that your question leaves Ai2 for OpenAI rather than anything about which system answers it. Those are the two places it happens and nowhere else on the site. Security: Two things to know about your own trail here: - Author pages are built from publicly listed paper details whether or not you ever create an account — yours may already exist. - The research feed is trained on what you save, so your library folders are a standing record of your interests held on Ai2's servers. Security: - Reading here exposes far less than pasting an unpublished manuscript into a chatbot. The papers and the citations between them are already public; that is the whole premise of the site. Best at: - Following a citation trail in both directions — forward to who cited a paper, backward to what it cited - Finding the paper you half-remember from a fragment of its title - Judging whether a paper mattered, using citation counts and its "highly influential citations" flag - Standing up a recurring email alert on an author, a paper or a topic Okay at: - Precise lookups — it ranks by meaning, which helps when you don't know the right keywords and gets in the way when you know the exact phrase you want - The newest papers, where indexing and paper details catch up after publication Avoid: - Systematic-review-grade searching. Semantic Scholar's FAQ states its search supports quoted phrases but not the AND/OR/wildcard syntax (it calls these boolean operators and wildcards) that a reproducible search strategy is written in. - Any literature that lives in books. The same FAQ says book coverage is very limited and patents are not included, so a search here under-represents fields that publish as monographs rather than as journal articles. Semantic Scholar publishes no figure for how completely arts, humanities and law are represented either way, so run one search here against a subject database you already trust before relying on it for those. - Reusing data pulled through the free API commercially without reading the licence. Ai2's API License Agreement, last updated 17 May 2023 and in force today, says the underlying data carries its own licences such as CC BY-NC, which excludes commercial use. Starter: Ask one paper a question in its own words Unlike Elicit, which answers from a pile of papers at once, this is scoped to the single paper on screen and returns supporting sentences lifted from that paper's own text — so the answer arrives with the evidence for checking it attached. Semantic Scholar offers starter questions like "What is the baseline condition?" — try one of those first to see the shape of a question it handles well. Prompt: What are the key results of this paper, and what did the authors say their method could not do? Starter: Walk the citation graph both ways Open any paper and move between its references — what it cited — and its citations — who cited it. Then filter the citing papers by whether Semantic Scholar classified the citation as background, method, or result extension. Google Scholar will tell you a paper was cited 4,000 times and stop there. Semantic Scholar labels the intent behind each citation, so you can skip the 3,900 that mention it in passing and go to the ones that built on the method. Sort the citing papers by most influential to find where this paper was central rather than a footnote. Starter: Teach the research feed in one sitting Save five papers you consider relevant into a single library folder, mark three suggestions "not relevant" in the feed for that folder, and toggle the research feed on. Those numbers are Semantic Scholar's own stated minimum for its recommender — five positive signals, three negative — and its recommendations are drawn only from papers published in the last three months, which makes the feed a current-awareness tool rather than a search. Come back tomorrow. A newly created feed produces nothing on the day you make it, and refreshes daily after that. Starter: Read an arXiv paper in Semantic Reader Open a paper hosted on arXiv and choose Semantic Reader instead of the plain PDF. Citations become cards you open in place instead of scrolling to the reference list, and terms with a dotted underline expand into a definition. Semantic Reader is reading the same citation index as the search engine rather than only the file in front of it, so a cited paper's own one-sentence summary appears in the margin without you leaving the page. The skimming highlights — AI-drawn Goal, Method and Result overlays — are available on most English-language arXiv computer science papers, on desktop only. Starter: Notice what else is running on this data Connected Papers, Litmaps and a number of other research tools are built on Semantic Scholar's API (a way for programs to talk to each other without a person clicking). Most of its endpoints answer without any key, and a key is free on request. Scopus and Web of Science sell access to exactly this kind of data; Ai2 gives it away, which is why Semantic Scholar turns up under other products' hoods rather than only at its own address. If you are not building anything, the useful takeaway is that several tools you already use are reading this same index — and inherit its gaps. Pitfalls: - The size of the index depends which page you read. The search bar currently says about 238 million papers, the API overview still says 214 million, and the About page says "over 200 million". Treat any single figure as approximate. - Indexed does not mean readable. Semantic Scholar links out to publishers. If a paper is paywalled you may land on a page asking you to buy it, and for some entries there is no link at all — just the title and its details. - The generated text is wrong in quiet ways. Semantic Scholar's own FAQ warns that errors in its AI-generated features will mostly be subtle rather than obvious. Read the abstract before you cite the summary. - Author pages assemble themselves. Papers get attached to the wrong person and one researcher can end up split across several pages. Claiming your page is how that gets fixed. - Its metrics are not the same as anyone else's. Citation counts here differ from other sites because the index differs, and Semantic Scholar explicitly says its h-index should be treated as a single indicator and not used for comparative analysis or research assessment. Where to next: - Other research and academic tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Working with PDFs and documents: https://www.launchpadhq.io/discover?category=document-pdf-processing - AI chatbots for general questions: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Semrush's AI Visibility Tools Canonical: https://www.launchpadhq.io/platform/semrush-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/semrush-ai.json Last reviewed: 2026-09-17 Tool: Semrush AI (https://www.launchpadhq.io/platform/semrush-ai): starting price $129.95/mo; free tier no; data as of not dated Finding out whether ChatGPT and Google's AI answers mention your business — and what it costs to keep watching. You type a website address, or a list of prompts (the short questions people type into an AI), into a box and hit go. Semrush runs them and hands back a report: a score, a count of times your brand came up, and which of your pages got linked. You change the list and re-run. What it is: - "Semrush AI" is not a product you can buy. Semrush is a long-running SEO suite — SEO meaning the work of getting your website to show up in search results — and it has bolted AI features onto that suite under several different names. - The one that matters is the AI Visibility Toolkit: it measures whether AI assistants name your brand when someone asks about your market, and whether they link to your site. - Semrush has been an Adobe company since 28 April 2026, when Adobe completed its acquisition. The product, the pricing pages and the signup flow all still operate under the Semrush name. - Everything else with "AI" on it — AI writing, AI review replies, AI press releases — lives in a different toolkit with its own separate subscription. Buying one does not get you the others. Before you start: - Start with the free checker, not the pricing page. Semrush's AI Search Visibility Checker takes a website address and gives you a report three times a day, with no account and no card — the vendor states this in the tool's own FAQ. That is enough to learn whether this category tells you anything you did not already know. - What you pay for after that is a quota and a schedule, not access. The AI Visibility Toolkit is sold on its own, covering one website address, or bundled with the search tools from the Starter plan up, cheaper on an annual commitment. Current prices are on its Launchpad page. - A free Semrush account (no card required) gets you one demo project, a small daily number of reports across the main tools, and what Semrush's knowledge base calls "a limited set of AI Visibility features." It does not say which features, and we did not sign in to find out — expect to discover the edges yourself. - You need a website or a brand name already out in the world for any of this to return something; there is nothing here to poke at with a made-up question. To see what AI can actually do, a text tool is the better first stop — this one measures a business rather than teaching you. Security: - What you hand Semrush is different from what you hand a chatbot: your own website, your competitors' names, and the questions you think your customers are asking — a strategy document sitting on Semrush's servers. Security: - Semrush's AI Functionality Terms are unusually clear on training: "Semrush does not train or develop its models, and does not allow its third-party providers to train or develop its models on User Input." The same document grants Semrush the right to "transfer and make User Input available to third-party artificial intelligence service providers," naming OpenAI products, Stable Diffusion and MidJourney — so what you type does leave Semrush. The terms carry no published effective date; the parent Terms of Service that incorporate them are dated 25 August 2026. Security: Worth knowing before you type: - Client names and unannounced products go into that prompt list the same way your own brand does - The privacy policy (last modified 13 October 2025) sends data to the United States and to group companies elsewhere, under standard EU contract terms - Semrush’s output carries no accuracy warranty at all — see the triad for what the terms disclaim Best at: - Answering what a search-ranking tool cannot: whether an AI assistant says your name at all - Running the same questions on a fixed daily schedule, so a change over a month is a measurement rather than an anecdote - Putting your brand and a named competitor against the identical prompt list, side by side - Separating *mentions* (your name appeared) from *citations* (an AI linked to your site) — only the second sends anyone to you Okay at: - Explaining why a number moved. It reports the movement; working out the cause is your job - Telling you which AI platforms are actually covered — Semrush's own pages disagree with each other, see the pitfalls - Writing the content you would fix any of this with. That lives in the separate Content Toolkit Avoid: - Feeding Semrush's output into another AI tool. Its Terms of Service (25 August 2026), §3.3(r), forbid using Semrush outputs "as inputs/prompts into, or for the purposes of developing, training, improving, fine-tuning, testing, or otherwise enhancing" any AI or machine-learning system, carved out only for Semrush's own official integrations. Pasting a Semrush report into a chatbot to summarise it is the exact move the clause names. - Treating what comes back as accurate because you paid for it. The AI Functionality Terms disclaim, in capitals, all warranties that outputs will be "FREE FROM ERRORS, BIAS, OR OMISSIONS, OR THAT THE OUTPUTS WILL BE UNIQUE, ACCURATE, RELIABLE, OR FIT FOR ANY PARTICULAR PURPOSE," and license output to you "only to the extent Semrush has such rights." - Buying the entry SEO plan because its card lists "Monitor custom prompts." Semrush's own knowledge base says that plan "doesn't include the full AI visibility feature set — no dedicated prompt-tracking quota, AI-ready Site Audit, or per-domain AI brand performance." Card and knowledge base contradict each other; the knowledge base is the more specific, and prompt tracking with a stated quota starts one tier up. Starter: Check your own site before you pay for anything Open Semrush's free AI Search Visibility Checker, type your website address in, and click Check AI Visibility. You get a score out of 100, a count of brand mentions, a per-platform breakdown, and the pages of yours AI answers link to most. The vendor's published three-a-day, no-card allowance makes a first look free rather than trial-shaped. Tested against Ahrefs' brand-monitoring tooling: swap the name in and the sentence fails, because it is anchored to an allowance Semrush publishes on this page. Run a competitor's address on one of your three daily checks — the gap between the scores is more informative than either number alone. Starter: Give Prompt Tracking real questions, not keywords Semrush's documentation says it accepts these one per line or comma-separated and runs them unchanged every day — which is why they must read like something a person would type, not like the keyword fragments an SEO tool normally wants. The quota is the product: 25 prompts on the standalone tier, 50 on Starter, 100 on Pro+, 200 on Advanced. Tested against typing the same four lines into ChatGPT yourself: that is one non-repeatable reading; the metered daily re-run is the difference. Swap the bracketed name for a competitor's on one line — same question, same day, is the comparison the score is built from. Prompt: best project management software for small teams project management tools compared 2026 easiest project management app to learn is [your brand] good for a five-person team Starter: Find out what people actually ask before you guess Prompt Research takes a bare phrase and returns hundreds of real prompts sorted by intent, against what Semrush describes as a database of 317M+ AI queries, with a daily cap on how many prompts you can pull. Tested against a keyword tool, Semrush's own free Keyword Tool included: those return search terms people type into Google, not the sentences people type into an assistant. Feed the best five straight into your Prompt Tracking list — that is the intended handoff between the two reports. Prompt: standing desk Starter: Read the citations column, not the score On any report, find Mentions and Citations and treat them as two separate numbers rather than one headline. Semrush publishes them as distinct counters on the same free report and says citations are "how you turn AI search visibility into real referral traffic" — so a high score built on mentions with no citations means the AI is describing you to people who will never reach you. Tested against a rank tracker, Semrush's own included: ranking tools have no equivalent split, because a blue link is a mention and a citation at once. Pitfalls: - The plan names changed and most guides have not caught up. Semrush One launched 29 October 2025 with Starter, Pro+ and Advanced, replacing Pro, Guru and Business. Stale references to the old names still appear in the text bundle of Semrush's own live pricing page, and in almost every third-party pricing article. Read the pricing page itself, on the day. - Adobe's acquisition came with a customer FAQ of the same date: "no immediate changes to your current services, agreements, or points of contact" — a statement about that moment, not a commitment about next year. Treat the current pricing as current, not settled. - Your own spot-check will not match the report, and that is expected: Semrush's FAQ says answers "change based on factors such as model updates, shifts in training data, or differences in how the same query is phrased," and that it runs prompts "in a controlled environment" so its numbers stay comparable. Do not open ChatGPT, get a different answer, and conclude the tool is broken. - Which AI platforms are covered depends on which Semrush page you read. The AI Visibility pricing page lists ChatGPT, Google AI, Gemini and Perplexity; the Prompt Tracking documentation lists ChatGPT Search, Google's AI Mode and Gemini, no Perplexity; the free checker's FAQ lists ChatGPT and AI Overviews as live with Gemini and AI Mode "coming soon," under a headline claiming both are covered. Three vendor pages, three lists — check the page for the tool you are buying. - The toolkits bill separately and it adds up fast. A Semrush plan covers search and AI visibility; Content, Local, Social, Advertising and AI PR are each their own subscription on top — including the Content Toolkit, where the AI writing lives. Where to next: - More marketing and SEO tools: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai - AI chatbots to try first: https://www.launchpadhq.io/discover?category=text-conversational-ai - Research tools: https://www.launchpadhq.io/discover?category=research-academic-tools ### Getting Started with Shopify Magic and Sidekick Canonical: https://www.launchpadhq.io/platform/shopify-magic-sidekick/guide Data: https://www.launchpadhq.io/data/v1/guides/shopify-magic-sidekick.json Last reviewed: 2026-09-09 Tool: Shopify Magic / Sidekick (https://www.launchpadhq.io/platform/shopify-magic-sidekick): starting price See pricing; free tier yes; data as of not dated Shopify's own AI, already switched on inside your store's dashboard — it writes the product page, fixes the photo, and answers "how do I do that?" You work in your Shopify admin, the dashboard where you run your store. Where you would normally type — a product description, an email subject line — a generate button sits in the toolbar; you describe what you want and edit what comes back. A purple glasses icon opens a chat panel that answers questions and proposes changes you approve. What it is: - Shopify Magic is the umbrella name for a set of AI features built into the admin itself rather than sold separately: a generate button wherever you write text, background and lighting editing on product photos, content for your store's design template and marketing emails, plus quieter things like summarised app reviews and automatic descriptions of customer groups. - Sidekick is one of those features and the only one you hold a conversation with. Shopify describes it as "an AI-enabled commerce assistant in your Shopify admin," trained on Shopify's own features and able to see your store's data. - Shopify's marketing site and its recent Editions releases now lead almost entirely with Sidekick, while "Shopify Magic" survives mainly in the help centre as the label for the whole set — shopify.com/magic currently serves a page about Sidekick that does not use the word Magic at all. Both names still resolve to real, current things; expect to hear Sidekick more. Before you start: - The AI is free; the store is not. Shopify states its AI tools are "available for free, regardless of your subscription plan" — no separate signup, no add-on, no credit balance to top up. What you pay for is Shopify itself. It currently advertises a short free trial and then a few months at a token price, after which you move onto a monthly plan — Basic, Grow or Advanced, each cheaper if you pay for a year. Current prices are on its Launchpad page. Let the trial lapse without picking a plan and Shopify's help centre says your store is paused until you choose one: your products and settings are kept, but the admin closes, and every tool on this page is inside the admin. - There is nothing to install or switch on: the generate buttons are already sitting in the text boxes, and Shopify's instruction for the assistant is exactly this: "Open your Shopify admin and click the purple glasses icon. That's it. No setup, no configuration." - Some of it is desktop-only, which matters if you run your shop from your phone. Shopify's help centre says text generation "is not supported on mobile devices," and that Sidekick "isn't currently available on mobile web admin or via the Shopify app on tablet devices, including iPads." It does work in the phone app, and has an optional voice mode there. - Shopify says its generally available AI features are available "to all plans and in all languages," and that Sidekick "adapts to your store's language settings." It also warns that language support "for some features, such as those in early access, might vary" — so if something described here is missing or English-only for you, that is the likeliest reason. Security: - Sidekick is useful precisely because it can read your live store — orders, customers, what sold and when — and that reach is also the whole of the risk. Its help centre says Shopify "doesn't use any merchant's store-level data to power Shopify Magic for other merchants," that where your data improves your own experience "that data isn't shared with other merchants," and — the same wall from the other side — that Sidekick "can't share private information or data from other Shopify merchants with you." Security: Four things worth knowing before you lean on it:: - Your staff see only what their permissions already allow. Shopify says every staff member uses Sidekick "based on their admin permissions" and that it "respects access controls, so team members only interact with data and features they're authorized to use." Someone who can't open your payouts page can't ask the assistant to read it out either. - It asks before it acts. Shopify's documentation says Sidekick "is never allowed to make changes to your shop without your approval, and will only present options for you to review and approve." That approval step is the entire safety mechanism. - Generated images carry an invisible watermark (a mark on generated output identifying it as AI-made). Shopify says it "doesn't claim ownership over images merchants create using media generation, nor restrict where they can use those images," and that the watermark doesn't restrict commercial use. - It remembers between conversations. Shopify's changelog announced in December 2025 that Sidekick now remembers your preferences and retains context from past chats. Best at: - Filling the empty boxes when you are setting a store up — product descriptions, page copy, email subject lines — from a product title and a few real facts about the thing - Answering "how do I do this in Shopify?" without leaving the admin, because Shopify describes the assistant as "trained to know all of Shopify's features" rather than answering from the general internet - Making a phone photo look deliberate — replacing a background with a solid colour, or describing the lighting and scene you wanted instead - Fetching a number you would otherwise build a report for; Shopify's changelog added payments and web performance data to what the assistant can read Okay at: - Longer jobs you would rather hand over. Shopify's Spring '26 release describes the assistant working in the background and carrying out multi-step tasks, but that is release-note language rather than documented behaviour — so treat it as a fast drafter, not something you leave running - Sounding like you. Shopify's own guidance is to state "the tone, length, format, and language of the output" in each request; it does not learn your house voice between products, so the voice is whatever you remembered to ask for that time Avoid: - Talking to your customers. Shopify's documentation is explicit: Sidekick "can't talk to your customers or handle customer support conversations on your behalf." It is an assistant for the shopkeeper, not a shop-front chatbot - Retouching images that are already placed inside your store's design template or your product list — Shopify says it "can't make edits to images within a store theme or your product list," so image work happens on the file, in the file editor, before it is placed - Rewriting a whole catalogue in one go. Shopify documents generation from inside an individual product's own description box and documents no bulk path, so a 400-product rewrite is an afternoon of clicking rather than a button — check for yourself before planning around it Starter: Draft the product description you have been avoiding When you are adding a product and the description box is empty → then click the generate icon in the toolbar and give it the product's real features, any keywords, and the tone you want. Swap-tested against Squarespace AI, which will also write you product copy. What is specific to Shopify is what it prints beside the button: that you are "responsible for the accuracy of all of the content that you publish to your store," and that the generator "can include things such as product benefits even when you didn't list any explicitly." Doing one product first, slowly, teaches you exactly what it invents. Feed it the three facts a customer actually asks about — what it is made of, what size it is, what it is for — and one tone word. Shopify's help centre says the prompt (the message you type) can carry "product features, keywords, desired tone, or any other instructions," and the generator already has the product title for context. Starter: Rescue a phone photo of your product When the only decent shot of a product has your kitchen counter in the background → then open the image in the admin's file editor and describe the background or lighting you want instead. Wix AI ships image tools too; what Shopify documents is mechanical: the editor "creates 1 AI-generated scene at a time," unused scenes are deleted when you close it unless you save them, and output lands at about 1 megapixel scaled to your image's aspect ratio. That is sized for a product listing, not for print — useful to know before you plan a whole shoot around it. Try the plain solid-colour background replacement before anything ambitious. It is more predictable than a described scene, and it is usually what a listing thumbnail actually needs. Starter: Ask the assistant where the setting lives When you are hunting for something — a discount code, a shipping rate, a domain → then open Sidekick and describe what you want in one sentence rather than searching the help centre. A general assistant like Wix AI answers Shopify questions from whatever it read on the internet, which goes stale with each Editions release. Shopify's version can see your store's own settings, so the answer arrives as a proposed change in your admin rather than as directions to a menu you then have to find. Ask it to make the change, not just explain it — the worst outcome of asking is a proposal you decline. Starter: Ask what actually happened last week When a week ends → then ask for a plain-language summary of how the shop did, or which products are running low. This is where the swap test against Squarespace AI fails hardest: the assistant reads your own order and payments data from inside the admin, so it answers from your actual sales. An outside tool can only tell you which report to open. Ask a follow-up rather than one enormous question. Shopify's own guidance is to treat it as "a working session that you refine over time rather than a single perfect request." Pitfalls: - It will invent product benefits. On a product page that is not a style problem — a claim about what something is made of or what it does is read literally by customers and, depending on what you sell, by regulators. Read every generated description before you save it. - Descriptions come out interchangeable. Shopify notes that "the same request won't always produce the same result," so you will not get literal duplicates — you will get a shop where every description opens the same way and reaches for the same three adjectives. Rewriting the first sentence of each one by hand fixes most of this. - The approval click becomes reflex fast. Slow down specifically where a wrong approval costs money or reaches customers: prices, discount codes, shipping rates, and any page that is already live. - Don't confuse this with the AI pointed at your shoppers. Shopify also builds and supports AI that faces the buyer — store chat, agentic storefronts, apps from the app store. Advice you find online about "Shopify AI" frequently means the other kind, and the settings do not overlap. - The buttons move. Shopify ships AI changes on a fast release cycle — the assistant gained background tasks, app generation and memory in a single December 2025 release, and app integrations the following spring. A guide written a year ago may describe a layout that no longer exists; Shopify's changelog is where to check what is current. Where to next: - Sales, marketing and SEO tools: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai - Image generation and editing: https://www.launchpadhq.io/discover?category=image-generation-editing - AI built into business software: https://www.launchpadhq.io/discover?category=ai-plugins-business-software ### Getting Started with Slack AI Canonical: https://www.launchpadhq.io/platform/slack-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/slack-ai.json Last reviewed: 2026-09-07 Tool: Slack AI (https://www.launchpadhq.io/platform/slack-ai): starting price $10/user; free tier yes; data as of not dated The AI that already knows what your team said last week — because it lives inside Slack. Slack AI isn't a separate app you sign into — it sits inside the Slack you already use for work. You ask a question in the search bar, or pick a summary from a conversation's menu, and it answers in sentences about your own projects rather than the internet. What it is: - A set of AI features built into Slack itself: channel and thread summaries, a daily recap, a search that answers in sentences instead of listing messages, notes from huddles (Slack's quick audio calls), and Slackbot as a personal assistant. - It's for people who live in Slack and lose real time scrolling — back from holiday, joining late, or in twenty channels they can't keep up with. - Why this rather than pasting messages into a general chatbot: it searches your workspace's own archive, which no outside tool can see — and only the part you could already read: its features "only use Slack data that members have access to at the time of request." Before you start: - There is no free tier. Conversation and thread summaries plus huddle notes currently come with every paid plan; the rest needs Business+ or higher — search answers, daily recaps, file summaries, translations, message explanations, canvas generation and the full Slackbot assistant. - You almost certainly can't switch it on yourself: AI is controlled by workspace and organisation Admins, so an ordinary member's first step is asking whoever administers the workspace whether it's on. - You won't be paying personally; your employer does. Slack has withdrawn the old per-seat AI add-on — it "is no longer available to purchase on the Slack website" — and moved existing customers onto plan tiers at their first renewal after August 2025. It also notes that the price "won't change if you turn all AI features off." Pro and Business+ are priced per user per month and cost less on annual billing; Enterprise+ is priced by sales. - First step once it's on: open a busy channel, click the three-dots icon at the top, choose Summarize channel. Not there? Your plan or your admin hasn't enabled it. Security: - Slack states that your messages and files are not used to train LLMs (large language models — the kind of AI that reads and writes text), and that the models it uses "are deployed inside Slack's cloud environment, so model providers do not have access to your data." Its security pages are blunter: "Customer data is never used to train third-party LLMs." Security: Worth knowing before you rely on it: - Permissions cut both ways: a recap can hand you a tidy digest of a sensitive channel you were added to months ago and forgot. - Outputs aren't all temporary — summaries and search answers are currently ephemeral, recaps persist around 90 days, and workflow-generated summaries follow your organisation's retention policy. - Your admin can switch AI off, or restrict who gets it, at any time. Security: - Slack is owned by Salesforce, which sells separate AI products that plug into Slack; this page covers only the AI inside Slack. Best at: - Catching you up after time off — a long thread compressed into a minute's reading - Answering "where was this decided?" and pointing at the source messages - Turning a huddle into written notes nobody had to take Okay at: - Summarising disagreement — it captures what was said, flattens who was pushing back - Translating between languages; fine for gist, not for anything contractual - Explaining internal jargon it has seen used in context Avoid: - Treating a summary as the record of what was agreed - Anything living in a channel or DM you're not a member of - Assuming it caught the sarcasm, or the reversal three messages later Starter: Catch up on one busy channel after time away Opens the channel, clicks the three-dots icon at the top and chooses Summarize channel. Its summaries link back to the messages they were built from, so one works as an index into two weeks of scrollback rather than a replacement for it. Start with your busiest channel, then open the threads it flagged. Starter: Ask where a decision was made Type a plain question into the Slack search bar and get an answer in sentences, with the messages it drew from. Search answers sit behind the Business+ gate, so this one feature decides whether the upgrade is worth it. Ordinary search hands you a list; this hands you the answer plus its evidence. Click through to the source messages it cites, every time. Starter: Turn on notes for your next huddle Starts a huddle from the channel and turns on notes. The notes are posted back into the channel that led to the call, so the decision and its context stay together rather than scattering into someone's notes app. Say at the start that notes are being taken — it changes what people say. Starter: Summarise a file someone dropped in a thread Opens a file that has already been shared in a thread and asks Slack AI to summarise it from there. It reads a file already in your workspace, so nothing is downloaded or uploaded to an outside tool — for internal documents that's often the difference between allowed and not. Use it to triage, never as your only read of anything you'll sign. Pitfalls: - Trusting a summary of a nuanced discussion. Summaries capture what was talked about, not how strongly people felt — twenty messages of quiet disagreement compress into one confident sentence. - Treating silence as evidence. It can't see private channels or DMs you aren't in, so "Slack AI didn't mention it" never means it didn't happen. - Quoting an AI answer as if it were the source. "Slack AI said the deadline was the 14th" is not a citation. Click through and quote the message. Where to next: - More AI inside the business software you already use: https://www.launchpadhq.io/discover?category=ai-plugins-business-software - Tools that capture and summarise meetings: https://www.launchpadhq.io/discover?category=meetings-notes ### Getting Started with Soundraw Canonical: https://www.launchpadhq.io/platform/soundraw/guide Data: https://www.launchpadhq.io/data/v1/guides/soundraw.json Last reviewed: 2026-09-15 Tool: Soundraw (https://www.launchpadhq.io/platform/soundraw): starting price $16.99/mo; free tier yes; data as of not dated Build a royalty-free backing track by clicking tags instead of describing a song — then pay to take it away. You open the generator and click tags for mood, genre and theme, then set a length and a tempo. A list of finished tracks appears and you play them. Pick one and an editing screen opens where you toggle instruments, raise or lower the energy of each section and trim the length, rebuilding the track as you go. What it is: - Soundraw makes royalty-free instrumental music — royalty-free meaning you pay for a licence rather than paying again each time the track is used. It is aimed at people who need something playing under a video, podcast, stream or game, not at people who want to write a song. - It does not generate vocals or lyrics. Its help centre is direct about this: Soundraw "specializes in generating instrumental tracks and does not support the creation of songs with integrated vocals or lyrics." If you want singing, you want a different tool in this category. - The other thing that sets it apart is what it is trained on. Soundraw says its AI learns only from music its own in-house producers wrote and recorded, which is why it is willing to promise a worldwide commercial licence on everything you generate. That claim is the whole product, and it is the reason the licence terms below are worth reading properly. Before you start: - You can generate and listen to full tracks at soundraw.io without signing up at all. A free account adds one thing: the ability to favourite a track and find it again later, under Favorites in the account menu. - The paywall is the download, not the music. Soundraw's help centre puts it plainly — creating songs is free, but "downloading and using the tracks require a subscription to one of our plans", and it says elsewhere that "there is no free trial of our license." So the honest answer on cost is: you will be paying before any file reaches your computer. - Which plan depends on what you are making. Creator, a monthly plan that costs less per month billed annually, gives unlimited MP3 downloads for background music in your own projects. The Artist tiers cost more and are for musicians who want to release tracks; only the top two include WAV and stems — the separate instrument files, drums and bass and melody apart, so you can rework them in music software like GarageBand or Ableton — and the two cheaper Artist tiers cap how many tracks you can download each month. One oddity to check once you are signed in: the pricing page's stems answer mentions "vocals (if any)" among the files, which sits awkwardly beside the help centre's statement that Soundraw does not generate vocals at all. - Annual plans are hard to leave. The terms of service (updated 12 June 2025) say fees already paid are not refunded, and that ending a yearly plan early means paying the remaining unexpired fees up front. Start monthly and switch later if it earns it. Security: - Soundraw is unusual at the input end: you type nothing and you upload nothing. What the company holds is ordinary account data — its privacy policy lists name, date of birth, address, phone, email and card number, plus usage and device information — and that policy is undated, which is worth knowing when you rely on it. The part that actually deserves your attention is the licence. Security: What the licence gives you, and what it takes back: - Soundraw's own explainer (19 February 2025) says that when you cancel you "do not have to take down any of your existing content" and can keep monetising projects that already use a downloaded track. You just stop being able to generate new ones. - There are named exceptions, and they are the uses people build businesses on. The licence page says that if you use Soundraw tracks as downloaded for meditation, lo-fi and relaxation content, "you can only keep your content published while your SOUNDRAW subscription is active." The cancellation explainer extends the same rule to background-music and lo-fi channels and to business hold music. - Registering a Soundraw track with Content ID — YouTube's automatic copyright-matching system, and the equivalents run by distributors — is prohibited outright. The licence agreement puts it in capitals. - You are licensing, not buying. Soundraw's licence agreement treats the recordings as Soundraw's and splits the underlying composition of anything you derive from them 50/50 between you and Soundraw. Security: - Note the tension before you plan around it: Soundraw's pricing FAQ says "Anything you make while you're subscribed stays licensed for life, even if you cancel later," while the licence page carves out those content types. The licence page is the more specific document and the one to act on — but if a meditation channel or a hold-music contract is the reason you are here, ask Soundraw in writing which one governs you. Best at: - Hitting an exact runtime: you set the length before generating and the arrangement is rebuilt around it, ending rather than fading - Auditioning a dozen directions in a few minutes without writing a word - Genre blends built as a combination of tags rather than as a phrase Okay at: - Landing on a specific reference sound: tags get you the neighbourhood, not the address - Long pieces — five minutes is currently the per-track ceiling, so anything longer means stitching two tracks together in an editor - Music that has to carry a moment on its own rather than sit underneath one Avoid: - Scoring video you do not own — films, TV, adverts — which the licence page lists as not covered; that needs a separate licence you have to contact Soundraw for - Selling a Soundraw-derived track on a stock-audio library; the licence names that as not allowed - Putting an unmodified Soundraw beat out as your own single. Even on an Artist plan, distribution to Spotify and the rest is permitted only if you change the beat first — "the final, modified song needs to sound clearly different to the original, downloaded track" Starter: Make a 30-second podcast intro Open the generator, set the length to 30 seconds, and pick one mood tag and one genre tag — Happy and Acoustic is a safe pair. Play the top three results, keep one, then in the editing screen mute the drums on the opening section so the track starts quietly and lands when your voice comes in. Muting a named section of a finished arrangement is not the same as fading a file — the intro drops out cleanly and the rest plays exactly as generated. Suno hands back a finished mix with no sections to reach into. Change the mood tag and regenerate — the length and the muted opening stay put. Starter: Blend two genres on purpose Select two genre tags at once — Hip-Hop with Orchestra, or Trap with Lo-Fi — and generate. Then remove one tag, generate again, and listen to how much of the result that single tag was carrying. Genre here is a filter rather than a word in a sentence, so a blend is repeatable: the same two tags return the same territory every time, which is not true of the same phrase typed into Suno twice. Try a pairing that should not work. The mismatches are where you learn what each tag actually controls. Starter: Decide whether it is worth paying for, before you pay Make a free account, spend two sessions favouriting five candidates, and play each one against your rough cut in the browser. Only subscribe once one of them clearly wins. Soundraw's help centre doesn't say how long an unpaid account keeps its favourites, so don't lean on one you couldn't bear to lose. Because only the download is metered (see Before you start), the entire audition happens unpaid — which makes the favourites list, not the generator, the thing to stress-test before you subscribe. Favourite the near-misses too. They are useful for working out which tags to change. Pitfalls: - The tags narrow a catalogue rather than compose from scratch. Stack too many at once and you get a short list of near-identical options; start with two or three and add one at a time. - On the Artist tiers the monthly download allowance is spent at the moment you download, not when you generate. Audition in the browser and download only the version you have settled on. - Editing happens before the download, not after. Once the MP3 is on your machine, changing the arrangement means going back to Soundraw and downloading again — which costs another download on a counted plan. Where to next: - Music Generation: https://www.launchpadhq.io/discover?category=music-generation - Video Creation & Editing: https://www.launchpadhq.io/discover?category=video-creation-editing - Voice & Speech: https://www.launchpadhq.io/discover?category=voice-speech ### Getting Started with Speak Canonical: https://www.launchpadhq.io/platform/speak-language-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/speak-language-ai.json Last reviewed: 2026-09-12 Tool: Speak (Language AI) (https://www.launchpadhq.io/platform/speak-language-ai): starting price $13.99/mo; free tier yes; data as of not dated Speaking practice in a new language with an AI that listens to you and scores what you said — phone only, and paid after a seven-day trial. You open the app on your phone, it shows a short line and plays it aloud, and you say it back into the microphone. It scores how close you got and moves you on. What it is: - Speak is a language-learning app built around one thing: getting words out of your mouth. - It's made by Speakeasy Labs. OpenAI's models (a model is the AI "brain" that does the actual thinking) power the conversation side, and on 10 September 2026 Speak announced a limited rollout of live tutor conversations built on OpenAI's newest voice model. - It teaches English to speakers of fifteen languages — Korean, Japanese, Traditional and Simplified Chinese, Spanish, French, German, Italian, Portuguese, Polish, Vietnamese, Hindi, Thai, Turkish and Indonesian — and teaches Spanish, French, Japanese, Korean, Italian and Simplified Chinese to English speakers. Speak's help centre, updated 3 July 2026, states the catch plainly: non-English courses exist only if your native language is English. - Why pick it over Duolingo, which is free: Duolingo lets you go months tapping multiple-choice answers without ever saying a full sentence, and Speak doesn't. Before you start: - There is no free tier. Speak's own site says the seven-day free trial is followed by a required paid subscription, and its help centre (updated 27 May 2026) says you add a payment method to start the trial and must cancel at least a day before it ends or you're charged automatically. Both Premium and the pricier Premium Plus can be paid monthly or yearly, with yearly working out much cheaper — Speak says prices vary by region and promotion and that the accurate figure is the one shown in the app. Refunds follow the store: purchases made through the Apple App Store can only be refunded by Apple, while for Google Play and purchases made on Speak's own site, Speak offers a full refund within 7 days and none after 30. - Speak's help centre says the app runs on iOS 16 or later and Android 8 or later, on phones and tablets, and that Speak is "not available on desktop (PC)". You can sign up and pay on the website, but the lessons happen on the handset. - You'll be talking out loud for most of a session, so decide where you'll do that before you start paying for days. There are quieter fallbacks — a typing mode inside the free-form conversations, a listening-only review mode, and short quiz-style lessons — but the quiet modes are a way to keep a streak alive, not a way to use the app. - Premium is the tier to start on. Per Speak's help centre both plans include the conversation practice, the tutor and the scripted course; Premium Plus adds a personalized study plan, targeting of your frequent mistakes, and removes daily caps on custom lessons. Speak's own two pages disagree about what a Premium subscriber gets from the "Made For You" custom-lesson screen — its plan comparison article says a small daily number, its feature article says that screen is Premium Plus only — so check which is true during the trial rather than after. Getting set up safely (Install the app published by Speakeasy Labs, and get to it from speak.com's own App Store and Google Play links rather than by searching a store. Check the publisher name on the listing before you tap install: the App Store also carries a different app called "Speak AI - Language Learning", published by Meta Innovation Limited, and a separate company at speakai.co sells a meeting-transcription product called "Speak Ai". Neither is this.): - Expect the app to ask for your microphone on first run, and say yes. Speak's own troubleshooting page for broken voice recognition is almost entirely about microphone permissions — on iOS, the microphone toggle and Screen Time restrictions on Microphone or Speech Recognition; on Android, permissions set to Deny, and Bixby Voice on Samsung devices. - Your "it worked" checkpoint is finishing one lesson and watching it mark a line you said out loud. If it scores everything as silence, that's the permission, not your accent. Security: - Speak's product is your voice, so the recording is the data. Its privacy notice, last updated 06/05/2026, says the app requests your microphone and uses that audio not only to score your pronunciation but to "develop, create, improve, optimize, maintain, and otherwise enhance its automatic speech recognition and voice transcription technology". Its European disclosure names OpenAI and Microsoft among the recipients of audio information. We could not find an opt-out for that improvement use anywhere in the notice. Security: Worth knowing before you record: - The scripted lessons are everyday phrases, so most sessions contain nothing about you. The part that changes is the free-form conversations, where you improvise about your own job, family, health or travel. - Speak's terms of service give it a "fully paid, royalty-free, transferable, perpetual, irrevocable, non-exclusive, and worldwide" licence to your content, and name "your name, voice, and likeness" in it. - We found no self-serve deletion route in the privacy notice. It says only that data is deleted or stripped of what links it to you once keeping it is "no longer reasonably necessary". If you want a specific recording gone, expect to email support. Security: - The sensible habit is a boring one: practice ordering coffee, not rehearsing what you'll tell your doctor. Best at: - Making you produce whole sentences out loud, repeatedly, with nobody around to hear you get them wrong - Short sessions — the unit of work is a handful of spoken lines - Pronunciation feedback the moment you've spoken — though Speak's help centre says its dedicated Pronunciation Coach is only for people learning English, while the plan table on speak.com lists the same feature with no such restriction, so the two vendor pages don't agree Okay at: - Grammar. The tutor will answer grammar questions, but the app's unit of progress is a spoken line, not a rule you've understood. - Reading and writing, which barely feature — expect to end up able to say more than you can spell. - Anything outside the language pairings listed above. There's no workaround, because the restriction is on which course exists at all, not on a setting. Avoid: - Assuming a feature you read about reaches your language. Speak's help centre describes its longer Tutor Lessons as currently available to English speakers learning Japanese, Korean and Italian, and calls the live tutor conversations a limited rollout for English and Spanish learners. - Treating the score as a verdict on your sentence. It measures how close your sounds landed to a target line — which is not the same as telling you that what you built was something a person would actually say. Starter: Finish one full scripted lesson before you judge anything The core loop scores you line by line: it marks the words it thinks you got and the ones it thinks you missed, and you go again until it's satisfied. Duolingo's speaking exercises are optional and skippable; here the score is the lesson. If it keeps failing you on one word, say that word on its own rather than repeating the whole line. Starter: Set up a roleplay for a conversation you're actually dreading In the free-form conversation section you pick a scenario and talk your way through it with a character that answers back and offers hints and corrections as you go. Speak's help centre describes a "Create Your Own" option where you specify a scenario, but the article never shows that screen, so this is the documentation's description rather than something we've seen working. You could get a roughly similar conversation out of ChatGPT's voice mode for nothing. What ChatGPT won't do is run it at two deliberately different speeds: Speak's help centre says its immersive version listens continuously and lets you cut in mid-sentence like a real conversation, while its chat version stays strictly turn-based. Start with something that has a fixed script in real life — a pharmacy counter, a check-in desk — before you try open small talk. Starter: Keep the conversation alive when you can't say it out loud This is the one place in Speak you type rather than speak: its help centre's own steps say that in Free Talk you tap the keyboard icon to switch to Typing Mode and type your responses instead of speaking. A repair phrase is the right thing to put in that box. Type it once, then say it out loud the next three times you need it. Prompt: Sorry, I didn't catch that. Could you say it again more slowly? Starter: Ask the tutor why you were marked wrong Speak Tutor is the app's general-purpose helper: its help centre says you can ask it about grammar, vocabulary or pronunciation, run conversation practice, play word games, and ask it to build you a speaking lesson on a topic you name. It lives at what the help centre calls a chat entry point at the bottom of the Home screen and the Practice tab — whether you type your question there or speak it isn't shown in the documentation. A lesson score tells you that you missed a word, never why. The tutor is the only part of Speak that answers "why". Ask about the sound, not the rule. "Why didn't it hear the ending of this word?" gets you further than "explain the past tense." Starter: Spend the trial week testing the method, not the app Speak's help centre says the seven-day trial gives you full access to everything in the plan you picked. So aim it at the hardest part — the free-form conversations — rather than the scripted lessons. Seven days is genuinely enough to learn whether being made to talk actually works on you. Do it on consecutive days — what you're testing is whether you'll still do it on day four. Pitfalls: - Installing it for a child on the strength of the store rating. The App Store rates the app 13+ and Google Play rates it 3+, but Speak's terms of service say use of the service by anyone who is a minor — "under the age of 18 in most jurisdictions" — is "strictly prohibited" unless a parent or legal guardian supervises it and accepts the terms on the child's behalf. The terms are the binding document; the store rating isn't. - Cancelling a day late and expecting the money back. Speak's help centre is explicit that cancelling stops the next renewal and does not undo one that has already gone through. - Pausing to think in the middle of a sentence. Speak describes its newest live tutor conversations as listening continuously for when you pause, finish, or need help mid-answer — which means a long hunt for a word can be read as your turn ending. - Counting on offline practice. Speak's own fix-list for broken voice recognition includes switching between Wi-Fi and mobile data, which points to the listening happening on its servers rather than on your handset. We found no vendor statement either way about using lessons with no connection, so don't buy it for the flight. - Mistaking the certificate for proof of level. Speak will issue an "enrollment certificate" on request, but per its help centre it certifies that you bought a Premium membership and are using the app — not that you reached any standard. Where to next: - Voice & Speech: https://www.launchpadhq.io/discover?category=voice-speech - Education & Learning AI: https://www.launchpadhq.io/discover?category=education-learning-ai - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Speechify Canonical: https://www.launchpadhq.io/platform/speechify/guide Data: https://www.launchpadhq.io/data/v1/guides/speechify.json Last reviewed: 2026-10-15 Tool: Speechify (https://www.launchpadhq.io/platform/speechify): starting price $29/Month; free tier yes; data as of 2026-10-14 An app that reads web pages, documents and files out loud to you. The free version is basic, and the paid trial bills you automatically if you don't cancel in time. You open a web page, document or file in Speechify and press play. A computer-generated voice reads it aloud, and the words on screen are highlighted as they're spoken. You can slow it down or speed it up, change the voice, or (in the phone apps) tap any word to start reading again from there. What it is: - Speechify is a reading app that turns written text into speech: articles, PDFs, Word documents, ebooks, emails. You listen instead of reading, or listen and read along at the same time. It runs on iPhone and iPad, Android, Mac, Windows, the web, and as a Chrome or Edge browser extension (a small add-on to your web browser). - Its founder, Cliff Weitzman, writes on Speechify's own About page that he is dyslexic and built the first version to get through his reading for college. The company lists "individuals with reading challenges like dyslexia" among its users, alongside students and professionals. Speechify's terms also say plainly that it is "not a health care or education provider." It is a reading tool. It isn't a treatment or a learning programme. - Since early 2026 Speechify has been pitching itself as a "Voice AI Assistant": it also turns your speech into text (it calls this Voice Typing), summarises pages, and answers questions about what you're reading. Speechify's pricing page lists those extras under the paid Premium plan. The free plan is listed as covering text-to-speech features only. - Choose it over the free read-aloud button built into your browser or phone if you want the same reader, library and settings on every device you own. Whether that's worth paying for comes down to whether you can live with the free voices, which is the first thing to find out. Before you start: - There's a free plan, and it's genuinely limited. Speechify's pricing page lists it as speeds up to 1.5x, 10 voices it describes as "robotic sounding", and text-to-speech features only. That's enough to find out whether listening works for you. Speechify's terms also say free access is "provided at our sole discretion" and is "not guaranteed," so it could change. - The paid trial and the free plan are different things, and the trial bills you. Speechify's Terms of Service (last updated 1 July 2025) say the free trial of Premium is "currently three (3) days unless a different period is stated at sign-up." Unless you cancel before it ends, it "automatically converts to a paid Subscription Plan" and charges the payment method you gave at sign-up. After that it renews for the same length each time. If you only want the free plan, skip the "Try for free" offer on Premium. If you take the trial, set a reminder for the day before it ends. - Premium is currently $29/month on monthly billing, per Speechify's pricing page. The page also offers yearly billing it says will "save 60%" but doesn't show the yearly dollar price, so you'll only see it at checkout. If you buy through the iPhone or Android app, Apple or Google handles the subscription, and Speechify's terms say those sales are final. - Refunds are narrow. The terms say fees for the current term "will not be refunded, in whole or in part" and there are no partial refunds. Speechify's separate refund policy covers charges made in error or through account takeover, and an unused yearly plan, only if you contact them within 7 days of the charge (the 7 days include any trial period). Cancelling stops the next charge. Speechify's help article says you keep Premium until the end of the period you've paid for. Security: - Speechify's privacy policy, updated 22 July 2026, says the text and documents you upload are 'transferred to, processed, and stored in the United States,' and that it uses information to 'improve the features, algorithms, and usability' of the service. The policy does not describe a way to opt out of that. It also says Speechify does not sell or rent your personal data. Security: Before you load your first document:: - Speechify's terms say what you upload 'is not considered to be confidential' and give Speechify an 'irrevocable, perpetual' licence to it that continues after you stop using the service. You still own your material. Even so, a contract, medical letter or someone else's private message is better read by a tool that keeps it on your device. - To delete your account, the privacy policy says to use the Settings page and, if you pay, cancel your subscription first. You can also ask for a deletion or a copy of your data through a support ticket. - If the account is a school's (Speechify for Education), the policy says your school administrator and parents may be able to see and delete it. Best at: - Getting through long reading you'd otherwise put off: articles, handouts, PDFs, with the text highlighted as it's read so you can follow along. - Picking up where you left off on another device. - Turning a web page into something you listen to on a walk or a commute, by pasting its link into the phone app. Okay at: - Sounding natural on the free plan, whose voices Speechify itself calls 'robotic sounding'. - Being a writing or research assistant. The summaries, chat and Voice Typing are newer than the reader they're attached to. Avoid: - Voicing anything you plan to publish or sell. Speechify's terms say its services, apart from its separate Voice Over Studio, 'are not intended for your commercial use.' - Starting the Premium trial casually 'just to look': it turns into a paid plan by itself (see Before you start). Starter: Have a web page read to you in Chrome Install the Speechify Chrome extension, open an article you've been meaning to read, click the Speechify icon next to the address bar, then click the blue play button. It reads from the top of the page. The free voices are the ones Speechify's own pricing page calls 'robotic sounding', so the useful test is a full article, not a sentence: can you listen to these voices for ten minutes? If yes, the free plan may be all you need. If no, you've found the real reason to consider Premium before paying for it. Try two or three of the free voices on the same paragraph. They differ more than you'd expect. Starter: Save a page to your phone and jump back by tapping a word Copy a web page's link, then in the Speechify phone app tap + (on iPhone choose 'Paste Website Link', on Android 'Web Link'), paste it, and open the new item in your library. Speechify's help article says that once a page is in your library you can 'tap on any word to start reading from that spot', so a sentence you missed is one tap away instead of scrubbing back through audio. That's what makes it practical for reading you need to understand, not just get through. Use the read-along highlighting on the hardest paragraph: listen and read at the same time, then try listening alone, and see which works better for you. Starter: Load a document you actually have to read Add a PDF, Word file or ebook (.doc, .docx, .pdf or .epub) to your Speechify library and play it, instead of a web page. Speechify's founder describes the gap he built it for as textbooks, handouts, PDFs and emails that have no audiobook. A real document from your week, with page headers, footnotes and columns, tells you whether the reader copes with your material, which a clean demo article won't. If the file sits in Google Drive, Dropbox or OneDrive, check once you're signed in whether the app can open it from there, or save a copy to your device first. Starter: Premium only: ask about the page you're reading Speechify's pricing page lists 'AI Summaries & Chats' and a Voice AI Assistant you can 'talk to … about anything, including any website or book' under Premium. The point is that you ask about the text already open in the reader, rather than copying it into a separate chatbot (an AI you type questions to). If you're on the Premium trial, try this before day three. Otherwise skip it: nothing here is needed to use the reader. Prompt: Summarise this page in five plain sentences, then tell me which part is hardest to understand and explain it more simply. Pitfalls: - Mistaking the trial for the free plan. Speechify's own blog spells it out: 'A trial is not the free plan.' Buttons saying 'Try for free' on Premium start the trial that bills you (see Before you start). The free plan has no time limit. - Cancelling the wrong place. If you subscribed on the web, cancel on the web: profile picture, then Settings, then Cancel Subscription, per Speechify's help article. If you subscribed in the iPhone or Android app, cancel it in your App Store or Google Play subscriptions. Deleting the app does not cancel anything. - Judging it on the voice you hear in ads. Marketing clips use the premium voices, so judge the free plan on its own voices. Where to next: - Voice & speech tools: https://www.launchpadhq.io/discover?category=voice-speech - Document & PDF tools: https://www.launchpadhq.io/discover?category=document-pdf-processing - AI chat assistants: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Spellbook Canonical: https://www.launchpadhq.io/platform/spellbook/guide Data: https://www.launchpadhq.io/data/v1/guides/spellbook.json Last reviewed: 2026-09-13 Tool: Spellbook (https://www.launchpadhq.io/platform/spellbook): starting price $99/user/mo; free tier yes; data as of not dated AI that marks up contracts inside the Microsoft Word document you already have open. You open a contract in Microsoft Word as usual and Spellbook sits in a panel down the right-hand side. You click a button or type an instruction, wait while it reads the whole document, and its edits and comments appear marked up in the text itself. You accept the ones you want, ignore the rest, and ask again. What it is: - Spellbook is a contract review and drafting tool for lawyers, built by Dialog Enterprises Inc. and sold to law firms and in-house legal teams. Rather than being a website you upload a document to, its main form is an add-in — a panel that installs inside Microsoft Word and works on the file you already have open. - A beginner would pick it over a general chatbot for one reason: it writes its suggestions back into the document as redlines (edits marked up so the other side can see exactly what changed) and margin comments, so a review ends with a marked-up contract rather than a list of findings to retype. - There is a second, separate product called Associate that works across several documents at once; the two are sold together as one licence. Before you start: - The requirement that decides everything else is Microsoft Word. Spellbook's Microsoft Marketplace listing declares support for Word on Windows, Word on Mac and Word in a browser; no iPad or phone version is listed. That listing is publisher-supplied metadata rather than a written statement from Spellbook, and it currently reaches back as far as Word 2013 on Windows and Word 2016 on Mac, so if you are on an older perpetual Office licence rather than a Microsoft 365 subscription, treat it as worth testing during the trial before you commit. - If you don't have Word, there are two other routes and neither is identical. Associate runs in an ordinary browser at associate.spellbook.legal and there is a desktop app; a Google Docs add-in also exists, but Spellbook's own install guide says the Chrome extension is what lets it insert tracked changes, it is Chrome-only, and without it you get a "Tracked changes unavailable" banner. - The trial form refuses consumer email addresses — its own error text says "please enter your business email (not gmail, yahoo, etc)" — and asks what kind of legal professional you are, though "No, but I'd love to try Spellbook" is one of the answers. The terms contract with a business entity, which does include a one-person firm. - You will need to pay to keep using it. There is no free tier and no published price: the pricing page carries no numbers at all, says pricing is "structured around the number of team members on your license," and routes every enquiry to a demo, so you cannot budget for this before a sales call. What you can get without one is a 7-day trial from spellbook.com/trial. Free access is offered separately to academic institutions through Spellbook's academic partnerships programme. - Anything you build during the trial is on a clock. Spellbook's help centre says that when the trial ends your projects become inaccessible unless you upgrade, and tells you to download any files you need before it expires. Getting set up safely (Microsoft's own add-in store, reached from inside Word — there is no Spellbook installer to download.): - Install it from inside Word itself: Home tab → Add-ins → More Add-ins → search "Spellbook" → Add. The publisher on the listing is Dialog Enterprises Inc. — that is the right one. Treat any search result offering a "Spellbook setup" file as something to close. - Expect it to fail on a work laptop before you conclude the product is broken. Spellbook's own guidance is that Word's optional connected experiences must be switched on (Windows: File → Options → Trust Center → Privacy Settings; Mac: Preferences → Privacy), and that on a work or school account your IT administrator may have blocked Office add-ins entirely. Firms that hit this deploy it centrally through the Microsoft 365 admin centre instead, which Spellbook documents. - After installing you sign in and enter a licence key, so the install and the account are two separate steps — finishing the first one does not mean you have access yet. Security: - The thing worth being clear-eyed about is what the panel is doing while you work: the Microsoft Marketplace listing states, in Microsoft's own words for this add-in, that it "can read and make changes to your document" and "can send data over the Internet." A review is your client's draft contract leaving your machine. Security: Where the document actually goes: - To Spellbook's own cloud first — its security page says it uses cloud providers with data centres in Canada and the US to store and process customer data. - Then out to third-party AI companies to do the reading. Spellbook names OpenAI and Anthropic and says it has negotiated zero-retention agreements with both, so that "customer data included in requests and responses with these LLMs is not persisted and exists only in memory in order to process a request." - Its Terms of Service (updated 27 July 2026, in force as at 13 September 2026) put that in binding form: data sent to those providers "will not be used to train, improve, or develop the AI models," and Spellbook agrees to treat your documents as confidential and not disclose them to other customers or third parties. Security: - Read that carefully rather than as reassurance. The promise Spellbook makes in writing is about what the outside AI companies may do; its security page does not separately address whether Spellbook itself trains on customer data, and documents you add to its Library or connect through SharePoint or iManage are stored by Spellbook by design. Files uploaded to the Ask feature are a stated exception — its help centre says those are deleted automatically after three days. Security: - One check before a real client file goes through it: Spellbook's Marketplace listing does not currently show a Microsoft 365 App Certification or publisher attestation, so a firm's IT team assessing it will be reading Spellbook's own trust portal rather than a Microsoft-run audit. Best at: - Taking a first pass over a contract you already have open and marking it up in place - Turning house positions into something reusable — Playbooks and Benchmarks convert your standards into pass/fail rules that everyone on the account then runs against the same contract types - Answering "what does this clause actually say" about the document in front of you Okay at: - Drafting a whole document from nothing: the feature works, but it is built to fill a blank file - Heavily formatted files — Spellbook's own guidance is that photos, tables and complex formatting affect how accurately its tools read a document - Working in other languages: its Ask feature page claims review and chat in 140+ languages, but there is no published plan or feature table anywhere on the site to check that against, so treat it as a claim to test during the trial Avoid: - Sending out AI-written markup where somebody is entitled to know a machine wrote it. Spellbook advertises that redlines are inserted under your own name in Word — Spellbook's own marketing line for it is "Shhh.... We won't tell them it was AI" — which means the tracked-changes record attributes the edits to you. Check your firm's policy, your client's expectations and any court's standing order on disclosing AI use before that document leaves your outbox. - Treating its output as finished work. Spellbook's Terms of Service put the obligation back on you in terms: the customer "will not solely rely on Output as constituting formal legal advice, and will ensure, when appropriate, that any Output is reviewed or vetted accordingly by a duly licensed and qualified lawyer." - Reaching for it for a one-off personal document — a lease, a freelance contract, a private sale. Nothing here fits a person reviewing their own paperwork. Starter: Run a first pass on a contract you already have open Spellbook's help centre documents the flow like this: in the Spellbook panel, click Review, then Comprehensive Review. Before it starts, set who you are acting for in the Representing Party dropdown and confirm the jurisdiction, which it fills in from the document but you can change. Tick whether you want general risks, proofreading or both, choose how much markup you want, then Start. Suggestions come back attached to specific passages; click Apply to insert one, or Apply All. The review is aimed before it runs — you set the side you act for and the jurisdiction, so the markup argues a position rather than describing the document neutrally. Starter: Write your own review instruction and save it for next time Custom Review runs a typed instruction as a review pass across the whole document; Spellbook documents a bookmark icon at the top of the prompt that saves that instruction into a personal list you reopen, edit or delete, which is how a one-off phrasing becomes something you run on every deal. Prompt: Review this agreement as though I act for the supplier. Flag every clause that lets the customer terminate for convenience, every indemnity that isn't capped, and anything that survives termination. For each one, suggest replacement wording that improves my client's position. Starter: Ask about the clause in front of you and follow the citation back Ask answers from the document open in Word and, per Spellbook's documentation, returns numbered reference bubbles you click to jump to the passage it used; the same box takes follow-ups, and toggles let you pull in your own Library, Spellbook's curated legal sources, or a web search. Prompt: Explain in plain English what this indemnity actually obliges my client to do, and point me to the exact wording each part of your answer depends on. Starter: Ask whether a term is normal, not just whether you like it Open Benchmarks from the panel's "more" menu. Spellbook's documentation describes it working out what kind of contract it is looking at, scoring how well the document covers the standards for that type, and listing what's missing. Click "Go to" to find the spot, "Show Fix" to see proposed language. This runs against a library of currently more than 2,000 pre-built standards rather than against your own precedent, so it answers a different question from the review. You can also generate a custom standard by uploading a model document. Starter: Draft from your own precedents instead of from a blank page The Draft box has a dropdown that switches the same typed request between "AI Generated" and "My Library" — the second searches clauses from documents you uploaded, with filters by folder and contract type — and Auto-Adjust then rewrites the chosen clause to match the defined terms and style of the contract you are in before you insert it at your cursor. Prompt: A limited liability clause favouring my client, capped at the fees paid in the previous twelve months, with carve-outs for breach of confidentiality and IP infringement. Pitfalls: - The Spellbook icon greys out on .doc files. Spellbook's help centre says it only works on .docx, because it uses Word features that older files don't expose — File → Save As, choose .docx, and it comes back. This looks exactly like a broken install and isn't one. - It cannot read a PDF. Spellbook's own answer is to convert the PDF to .docx first, or to use Associate, which does take PDFs — so the counterparty's scanned execution copy is not something the Word panel will touch. - There are size limits that bite on long agreements: currently around 4 MB per document in the Word and Google Docs add-ins, and around 20 MB per file in Associate. Ask's document uploads are capped at 20 files at a time. - "Insert All" in the full-document drafting flow overwrites everything currently in your Word file — Spellbook's own instructions say to run that feature on a blank document. Insert at Cursor is the safe button when you only want one clause. - Two vendor pages disagree about what happens when the trial ends, and the difference is money. Spellbook's public Refund Policy says "All monthly Spellbook subscriptions automatically renew after the trial period ends" and that it does not issue refunds once that has happened. Its help centre says a self-serve trial started on the website "will automatically expire after 7 days," that you "will not be charged automatically," and that only a sales-assisted trial may convert into a paid subscription. Assume the refund policy governs unless someone at Spellbook tells you in writing which trial you are on, and ask before the seventh day. - Bookmarks and links will look inconsistent: the marketing site is now spellbook.com while the help centre and the Associate app are still on spellbook.legal. That split is Spellbook's own, not a sign you are on a fake site. Where to next: - Legal AI tools: https://www.launchpadhq.io/discover?category=legal-ai - Document & PDF processing: https://www.launchpadhq.io/discover?category=document-pdf-processing - AI plugins for business software: https://www.launchpadhq.io/discover?category=ai-plugins-business-software ### Getting Started with Squarespace AI Canonical: https://www.launchpadhq.io/platform/squarespace-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/squarespace-ai.json Last reviewed: 2026-09-09 Tool: Squarespace AI (https://www.launchpadhq.io/platform/squarespace-ai): starting price $16/mo; free tier yes; data as of not dated Answer a questionnaire about your business and get a whole website back — then spend your real time editing what it wrote. You start by answering a few questions about your business — what it does, who it is for, how you want it to sound. A complete site comes back, with pages, pictures and words already in place. From then on you edit it: click any block of text, ask for a rewrite, and swap out whatever you do not like. What it is: - Squarespace AI is a name Squarespace itself uses — it is what the company calls the writing tool built into its website editor. It sits inside a wider bundle the company brands Design Intelligence: a builder that drafts an entire site from a questionnaire, an image generator, writers for product listings and discounts, and a checker that tells you whether ChatGPT and Gemini mention your business. - None of it is sold separately: you get it because you have a Squarespace website subscription, and it lives in the same editor as everything else. - Against Wix, the split that will actually reach you is pricing: Wix has a permanently free plan you can sit on indefinitely; Squarespace has no free plan at all — a short free trial, then a paid subscription, cheapest when paid annually. - The editor is tighter in exchange. Squarespace gives you fewer places to put things than Wix's free-form canvas, so an AI-built Squarespace page has a smaller range of ways to come out looking wrong. Before you start: - The trial is the whole entry barrier. It needs no card, and Squarespace will extend it once by seven days if you ask. During the trial your site stays private, search engines are told not to index it, and Squarespace says trial content is marked for permanent deletion if you never upgrade. So you should be able to build the entire thing with AI for nothing — Squarespace's help centre says a trial includes nearly all platform features, without spelling out the AI text tools specifically — and you just cannot show it to anyone until you pay. - There are four tiers, which Squarespace now calls Basic, Core, Plus and Advanced. Older accounts still show the previous names — Personal, Business, Commerce — because the new plans have been rolling out in phases since early 2025, so the tier list in a help article may not match the one in your account. - The writing is unmetered; the pictures are not. Squarespace runs AI credits (the platform's unit of spend — each thing you make costs some) on its image tool and its AI-search checker, and its help centre says the text tools and the built-in assistant do not consume them. A trial site gets a small number of credits. Paid tiers range from a one-time starter amount at the bottom to a larger monthly allowance at the top, and you can buy extra packs, which roll over and expire a year after purchase. - First move: start a trial at squarespace.com and let the builder generate before you touch anything. Regenerating costs nothing, so it is cheaper to throw the first site away than to start hand-editing a version you already dislike. Security: - Almost everything you make here is meant to be published under your own name, so the sharper question is not what happens to what you type but what goes live, and what it was made from. Security: What Squarespace's own terms and settings say: - You own both ends of it. Squarespace's AI Terms (currently effective 25 June 2026) say that as between you and Squarespace, you own what you type in and what comes out. - But you may not type in personal information. The same terms bar it outright — not as advice but as a restriction. Worth remembering the moment you are tempted to paste a real client's details in to get a better product description. - The AI doing the writing is not Squarespace's. Its help centre says Squarespace "has partnered with Open AI, Anthropic, and Google to support how Squarespace AI works," and its terms allow those providers to use the output under their own policies. What you type leaves Squarespace. - Squarespace pre-emptively disclaims the output. Its terms say generated content "may be inaccurate, incorrect, erroneous, strange, incomplete or not fit for use or purpose," and put the checking on you before you publish. - AI crawlers can read your finished site unless you stop them. Under Settings, a Crawlers checkbox adds known AI crawlers to your site's exclusion file. It is off by default — Squarespace says it does not want to "potentially impact your site's traffic by excluding it from chat answers and sources" — and it warns that ticking it "doesn't guarantee" crawlers comply, and does not remove anything already taken. Best at: - Producing a complete, coherent site in one sitting, with every page populated rather than left as placeholder text - Rewriting text you have already written — shortening, lengthening or simplifying a paragraph in place, without leaving the page you are editing - The dull-but-necessary writing: product listings, page descriptions for search engines, image descriptions, video and course blurbs - Holding one tone across a whole site, because the brand answers you gave at the start are what later copy is generated from Okay at: - Images. Image Studio generates pictures and strips backgrounds, but it draws on the same credit pool as the AI-search checker, so it is not where you would do volume work - Blog drafting. It produces posts, outlines and ideas, but a piece with an actual argument in it still needs you to write the argument - Checking whether ChatGPT and Gemini mention your business — the tool works, but every run spends credits, and Squarespace limits it to English-language, version 7.1 sites, so on a 7.0 site it is not an option at all Avoid: - Treating the trial as a free tier — nothing you build is publicly reachable until you subscribe - Putting anyone's personal details into a prompt (the message you type) — Squarespace's AI Terms prohibit it, they do not merely discourage it - Medical, legal or financial advice, and anything that misleads a reader about where the content came from — both are barred by those same terms - Assuming an AI-designed site can move somewhere else later. Squarespace's export is a WordPress-format .xml file, and its own help centre lists style settings, custom CSS, store pages, product and video blocks and all but one blog page as things that do not come with you Starter: When you have a business but no website → then let the questionnaire build the whole thing first Answer Squarespace's questions about what you do and how you want to come across, and Blueprint AI returns a finished site — layout, copy and images already placed — that you edit from there. Squarespace publishes its own numbers on this feature: it says Blueprint AI has "helped 10% more people publish their site and reduced their need to search for stock images by 18% on the first day." Those are the vendor's figures rather than an independent test, but they describe a builder optimised to get you past the blank page, not to hand you a canvas. If the first site comes back wrong, change the brand-personality answers rather than editing twenty blocks by hand — they are upstream of most of what you would be fixing. Starter: When you would rather start in a chat window than a website builder → then use Squarespace's GPT inside ChatGPT Squarespace publishes a Squarespace GPT in ChatGPT's store. It interviews you about your business, tone and goals, shows a preview of the site it would build, and only then sends you to Squarespace to trial or publish. It is the one route here that lets you see what Squarespace's builder would produce for your specific business before you have created a Squarespace account or started the 14-day clock. Given that the clock is the real constraint on this platform, starting it after you have seen a preview rather than before is worth the detour. Starter: When a page you wrote reads badly → then rewrite it in place instead of starting over Squarespace's help centre says to select the text, click the three-circles icon in the toolbar, and ask for it shorter, longer or simpler. It rewrites in the block you are standing in. This is the part that is not metered: rewriting is the one AI action on this platform you can repeat as often as you like without watching a balance drop. Starter: When your shop page is empty → then photograph one product and let it write the listing The AI Product Composer takes a photo or a few basic details and drafts a description, a suggested price and copy aimed at search engines. A matching tool drafts discounts. Selling is switched on at every Squarespace tier — its help centre says all four plans allow unlimited products, with the entry plan differing by a 2% transaction fee rather than by a locked feature. So the shop exists whether you planned one or not, and this is the tool that stops it sitting empty. Treat the suggested price as something to argue with. Squarespace does not say what the suggestion is based on, so it is a starting number, not a valuation. Starter: When you want to know whether AI search mentions you → then run the visibility checker, sparingly AI Visibility runs prompts through ChatGPT and Gemini and shows whether your site appears, which prompts surface it, and which competitors get cited instead of you. Check first that it can run for you at all: Squarespace's help centre says the tool is "currently only available in English, on version 7.1 sites". It is on every website plan, but only the credit allowance changes by plan — a 7.0 site cannot use it whatever you pay. It is the only AI feature here that spends your credits on questions rather than on making something — Squarespace's pricing page counts them as "prompt runs". That budget makes this a periodic check, not a dashboard you leave running. Pitfalls: - The generated copy is a draft that looks finished. There is no blank space left to remind you which sections you never rewrote, so the placeholder-ish sentences ship to a live site looking as settled as the ones you wrote. Squarespace says it plainly in its own help centre — do not publish text "that you wouldn't have written yourself." Read every page once before you go public. - Credits disappear into images, not words. If you are planning to lean on generated pictures, work out roughly how many you need before you start, rather than discovering the ceiling halfway through a gallery. - The trial clock, not the plan, is the real deadline. An AI builder will happily fill a week you do not actually have. - Plan names in the documentation may not match your account. A feature documented under one tier can be sitting under a differently named tier in front of you, so check what your own billing page calls your plan before concluding a feature is missing. Where to next: - More AI built into business software: https://www.launchpadhq.io/discover?category=ai-plugins-business-software - Marketing and SEO tools: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai - Image generators that go further: https://www.launchpadhq.io/discover?category=image-generation-editing ### Getting Started with Stability AI Canonical: https://www.launchpadhq.io/platform/stability-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/stability-ai.json Last reviewed: 2026-10-23 Tool: Stability AI (https://www.launchpadhq.io/platform/stability-ai): starting price Free, from US$0.01 per credit, pay as you go; free tier yes; data as of 2026-10-25 The company behind Stable Diffusion and Stable Audio — use its models through a music web app, a pay-per-image API, or free downloads you run yourself. Stability AI makes generative models for images, audio and 3D. There are three ways in: the Stable Audio web app for making music in a browser; the Developer Platform, where your code pays a few cents per image; or downloading the open models and running them on your own computer with a tool like ComfyUI. What it is: - Stability AI is best known for Stable Diffusion, the family of image models anyone can download and run. Today it offers image models (Stable Diffusion and its "Stable Image" services, from the fast and cheap Core to the flagship Ultra), the Stable Audio family for music and sound effects, and 3D models that turn an image into a 3D object. - Unlike most AI companies, it publishes many of its models as open weights — files you can download from its Hugging Face page and run on your own hardware. Its model pages recommend ComfyUI or the diffusers code library for this. Running locally means no per-image cost and nothing leaves your computer, but you need a capable graphics card. - For people who don't want to install anything, there's the Stable Audio web app at stableaudio.com — Stability's one remaining consumer app, which it described as an early beta in August 2026 — and the Developer Platform, an API (a way for software to call the models) that charges per generation. Its consumer image app, Stable Assistant, no longer appears on Stability's site, and its old address now redirects elsewhere. Before you start: - API credits are pay-as-you-go. Stability's pricing page sells credits individually and gives new accounts a small free allowance; its getting-started guide describes it for accounts created with Google sign-in. An image from Stable Image Ultra costs more than twice as much as one from Stable Image Core. Its terms say credits are non-refundable and expire a year after purchase. - Stable Audio has its own plans. The web app uses a separate credit system from the API, with a free plan and paid monthly plans listed at stableaudio.com/pricing. Its credits don't carry over between months. - Running models yourself takes real hardware. Stability's own figure for running its largest Stable Diffusion model is about 19 GB of graphics memory (VRAM), or about 11 GB with an optimised version on recent NVIDIA cards. Smaller models need less, but plan on a dedicated graphics card. - Check the licence if you'll earn money. If you distribute the models, or a product or service built on them, you must also include the licence and display "Powered by Stability AI". Getting set up safely (Use stability.ai, platform.stability.ai, stableaudio.com and the official stabilityai page on Hugging Face. Downloads there are gated: you click "Agree" to the licence before you can get the files, which is also your reminder to read it.): - Protect your API key. The Developer Platform gives you a secret key that spends your credits. Keep it in your code's environment settings, never in a shared document or public code. - Decide about training before you upload. Stability's privacy policy says it may use your inputs and outputs to train its models, and that you can opt out in Account Settings. Do that first if you'll upload your own images or audio. - Confirm it works cheaply. Generate one image with Stable Image Core, the cheaper option, from the free allowance before trying the more expensive services. Security: - If you use Stability's own services, what you upload and generate is processed on its servers, and by default may be used to train its models until you opt out. Running the open models locally avoids that entirely — nothing is sent anywhere. Security: Rules worth knowing:: - Stability's Acceptable Use Policy bans sexually explicit content, content that harms others, attempts to get around its safeguards, and passing AI output off as human-made. - Its terms set a minimum age of 18 for most services; the audio services allow 13 and up with a parent's permission. - Stability's API FAQ says outputs can be used commercially as long as they don't infringe anyone's copyright, and its licence says you own your outputs as far as the law allows. Best at: - Open image and audio models you can run privately on your own computer, with no per-use fees. - Cheap, high-volume image generation from code through the API. - Making music and sound effects in the browser with Stable Audio. Okay at: - Being a beginner's first stop for images. Without its old consumer image app, the easy routes are the API or a local tool like ComfyUI, both a step up from a chat app. - Video. Stability's API no longer offers video; its video models are now download-only. Avoid: - Uploading private material before opting out of training. Starter: Make a track in the browser Open stableaudio.com, sign up for the free plan, and describe a short piece of music — a genre, a mood and a few instruments. Then refine it by describing what to change. It's the only no-install, no-code way into Stability's models, and audio is where its newest model family is. Starter: Price-test two image services In the Developer Platform, spend your free credits generating the same prompt with Stable Image Core and the pricier Stable Image Ultra, and compare. Costs vary a lot between Stability's services. Seeing the quality gap yourself tells you which one your project can afford to use at scale. Starter: Try an edit, not a new image Use the API's Remove Background or Search and Replace tool on a photo you own. These edit tools change one thing about an existing image without you drawing a mask — often more useful day to day than generating from scratch. Starter: Run a model locally If you have a capable graphics card, accept the licence for one of the Stable Diffusion models on Stability's Hugging Face page, download it, and load it in ComfyUI using the example workflow on the model card. This is what makes Stability different: once it's running, every image is free and private. Start with a smaller model if your graphics memory is limited. Pitfalls: - Two separate credit systems. API credits and Stable Audio credits aren't interchangeable. - Expensive extras. Stability's pricing lists its Creative Upscaler at many times the credit cost of generating the image. - Forgetting the licence conditions. Commercial use of the open models needs registration with Stability even under the revenue threshold, and distributing the models or a product built on them adds the licence copy and the "Powered by Stability AI" notice. - Out-of-date guides. Many older tutorials point to Stable Assistant or Stability's retired video API; check against Stability's own docs. Where to next: - More local & open-source AI: https://www.launchpadhq.io/discover?category=local-open-source-ai - Image generation & editing: https://www.launchpadhq.io/discover?category=image-generation-editing - Music generation: https://www.launchpadhq.io/discover?category=music-generation ### Getting Started with Suno Canonical: https://www.launchpadhq.io/platform/suno/guide Data: https://www.launchpadhq.io/data/v1/guides/suno.json Last reviewed: 2026-07-13 Tool: Suno (https://www.launchpadhq.io/platform/suno): starting price Free, from $10/mo; free tier yes; data as of not dated Describe a song in words and Suno generates a full track — vocals, lyrics, and instruments — in about a minute. You describe the song you want — a genre, a mood, a subject — the way you'd type a message to a chatbot, and what comes back is a complete track with vocals that you play in the browser. You listen, change the description, and generate again until one lands. What it is: - Suno is an AI music generator. You type a description (a style, a mood, a topic — and optionally your own lyrics) and it produces a complete song with vocals and instrumentation. It's aimed at anyone who wants original music without playing an instrument: content creators needing a soundtrack, hobbyists, people making a birthday song. No musical training required. Before you start: - The free plan gives you a bucket of credits that refills daily (enough for a handful of songs a day); credits don't roll over. Paid plans (Pro, and Premier above it) add far more credits, faster generation, and — crucially — commercial rights. - The rights situation is the thing to understand before you make anything you care about (below). In short: free-tier songs are for personal use only, and that can't be undone later by upgrading. - Expect to generate several versions before one lands — that's normal, and it's what the credits are for. Security: Free vs paid changes what you can do with the song: - Songs you make on the free plan are for personal, non-commercial use only — you can't sell them, put them on Spotify, or use them in monetized videos. - Upgrading later does not retroactively unlock the songs you already made on the free plan. If you might want to use a track commercially, make it on a paid plan from the start. - Paid plans grant commercial-use rights for songs made while your subscription is active — and those rights persist for already-generated tracks even if you later cancel. Security: - Even on a paid plan, read the fine print on ownership: Suno's current terms grant you a license to use and monetize the song but stop short of calling you the owner, and note that copyright may not vest in AI-generated audio at all. For casual use this doesn't matter; for anything you're building a business on, it does — and this is a legal question, not one an AI tool can answer for you. Security: - Anything you submit — your prompts, and especially any audio or lyrics you upload — may be used to help train and improve Suno's models. Don't feed it confidential lyrics, an unreleased melody, or a client's material you don't have the rights to. Best at: - Generating a full, original song with vocals from a text description - Background music and jingles for videos, podcasts, and social posts - Fast iteration — trying ten directions in the time it'd take to write one Okay at: - Matching a very specific reference track (it gets the vibe, rarely the exact thing) - Precise lyrical or structural control (better than it was, still imperfect) - Long-form or multi-section compositions Avoid: - Commercial releases made on the free tier (personal use only) - Cloning a real artist's voice or copying a specific copyrighted song - Uploading confidential or client-owned material as input Starter: A simple first song Genre + mood + instruments + vocal + energy is the recipe for a usable result on the first try. Swap the genre and vocal to hear how much they change the whole song. Prompt: An upbeat acoustic folk song about a road trip with friends, warm male vocals, claps and a catchy chorus, feel-good and bright. Starter: Bring your own lyrics Shows the biggest lever most people miss — you can write the words and let Suno handle the music. Keep the lyrics short at first; a couple of lines is enough to hear the treatment. Prompt: Lo-fi chill beat, soft female vocals, mellow and late-night. Use these lyrics: [paste two or three lines of your own here] Starter: Nail a specific use Most real uses are short and functional; specifying length, "no vocals," and the purpose gets you something you can actually drop into a project. Change the instrument and tempo to match the mood of your content. Prompt: A 30-second cheerful ukulele instrumental for a cooking video intro, no vocals, light and bouncy, loops cleanly. Pitfalls: - Free credits reset daily and don't stockpile. You can't save up a week of credits for one big session. - The free/paid commercial line is permanent per song. A track made free stays non-commercial forever, even after you upgrade. - "Commercial rights" is not the same as "copyright ownership." Fine for casual use; get real advice before building a business on AI-generated tracks. Where to next: - Music Generation: https://www.launchpadhq.io/discover?category=music-generation - Video Creation & Editing: https://www.launchpadhq.io/discover?category=video-creation-editing ### Getting Started with Surfer Canonical: https://www.launchpadhq.io/platform/surfer-seo/guide Data: https://www.launchpadhq.io/data/v1/guides/surfer-seo.json Last reviewed: 2026-10-14 Tool: Surfer SEO (https://www.launchpadhq.io/platform/surfer-seo): starting price $49/mo billed yearly; free tier no; data as of 2026-10-20 The tool that tells you what the pages already ranking cover — and what your draft is missing. You type the search phrase you want to be found for, and Surfer reads the pages already winning on Google for it. It opens a writing document with a checklist down one side — words to include, how long to go, how many headings — and a score that moves up or down as you write or paste text. What it is: - Surfer is a writing tool for people trying to get found in search: it compares what you are writing against the pages already ranking for your phrase and shows what they cover that you do not. - Since being bought by Positive Group — announced in October 2025 — it describes itself as an "AI visibility platform" rather than an SEO tool. Its newer half watches whether ChatGPT, Gemini and Perplexity mention your brand when someone asks them about your field; you supply the prompts (the messages a person would type into those assistants) you want checked. - It will also draft the whole article for you from a phrase and some instructions. Whether you write it yourself or let it draft, the same scoring panel sits beside you. - One cosmetic note: the parent company's rename shows up in browser tabs as "Positive Surfer." The product is still called Surfer and still lives at surferseo.com. Before you start: - There is no free plan. Surfer's help centre says new accounts get a 7-day trial of its Pro plan, but you enter billing details at checkout to start it, and on day 7 it converts to whatever plan you picked and charges the card. It does not lapse quietly. Paid plans start at Discovery and step up through larger tiers, with a custom-priced Enterprise tier above that; current prices are on its Launchpad page. - Two pieces of Surfer cost nothing and want no subscription. Keyword Surfer, a browser extension Surfer's own page calls "100% free. Forever.", shows search volumes inside your normal Google results; the Chrome Web Store lists it as last updated January 2026 with around 600,000 users. The AI Humanizer rewrites a short passage with no account, and a longer one once you register without a plan — English only, as the vendor states it today. - The trial runs once: Surfer's help centre says it is for new accounts only, and a card already registered with Surfer makes you ineligible, so you cannot restart it under a new email with the same card. - Annual billing saves up to about 17%, but Surfer's Regulations say a cancelled prepaid annual subscription is not refundable, with only a proportional recalculation available on request. Monthly costs more per month and risks less. Security: - Surfer is run by Surfer sp. z o.o., a company in Wrocław, Poland, so its customer privacy policy — last updated 23 January 2026 — is written to European privacy law. Most of what you put in is marketing copy about your own business, which makes this lower-stakes than a general chatbot. Security: Worth knowing before you paste your site into it: - That policy names roughly 23 outside services that touch customer data, split between European and American processing. - It does not say either way whether your documents are used to train AI models. If that matters to you, ask support rather than assuming the answer you would prefer. - Customer data is kept for the length of your contract plus around 30 days, which Surfer attributes to backup integrity. - Surfer's help centre warns that after you cancel, the account expires and stored data goes with it — export before your last paid day, to plain text, HTML or Markdown. Best at: - Telling you what the pages currently ranking for your phrase cover that your draft does not - Giving a long article a shape — how many headings, roughly how long, which subtopics — before you write a word - Getting a competent first draft out of a single phrase when the blank page is the problem Okay at: - Prose quality. The score rewards coverage, not writing. You still edit. - Small sites. A lot of what separates the plans is tracked pages and tracked prompts, which you may not have enough of yet to use. Avoid: - Reading the Content Score as a forecast. Surfer's own Regulations state that its analyses "are not meant to serve as clues or recommendations," and that acting on them is voluntary and at your own risk. The score says you resemble the pages that rank, which is not the same as beating them. - The article generator in Arabic, Hebrew or Maltese. Those are the only three languages Surfer's help centre lists as unsupported for AI article generation, out of roughly 29 it does support — while describing the analysis side as working in all languages. Check yours is on the list before you plan work around it. - Treating the AI Humanizer as permission. Surfer markets it as helping text "pass most of AI detection tools" — a claim about detectors, not about whether your client, publisher or institution allows AI-written text in the first place. Starter: Look at Surfer's data before you look at Surfer Add the Keyword Surfer extension to Chrome or another Chromium browser, then search Google for the phrase you would want to be found for. Search volumes and related phrases appear inside the results page you were already reading. The guidelines Surfer builds inside its paid editor rest on this same search-volume layer, and this is the one place you can inspect it without a login. Tested against Semrush, whose keyword numbers live inside a dashboard you have to go to: here you judge the data first, in a page you already trust. Starter: Overrule the average before it drafts anything This goes in the "context & instructions" box on the Content Editor's setup screen, before "Generate content." Surfer builds its guidelines from the pages currently ranking, so left alone it drafts toward your competitors' average; its own documentation says this box takes free-text instructions plus up to five source URLs, and it is the only point before generation where you can contradict that average. Tested against ChatGPT, which has not read the ten pages ranking for your phrase: there is no average for your instructions to push against. Replace the bracketed business with your own, and use your two strongest existing pages as the source URLs. Prompt: Write for the owner of a small independent [bike repair shop] who has never optimized a page before. Keep sentences under 25 words. Use British spelling. Do not claim we are cheapest — we compete on turnaround time. Treat these as the source of truth about us: [your about page URL], [your services page URL]. Starter: Fix one paragraph without leaving the score You select a paragraph in a Surfer document and click "Ask Surfy." Surfy exists only on selected text, and Surfer's docs say it runs on its own separate pool of credits (Surfer's documentation calls them tokens) rather than the allowance a full article generation spends, so small fixes are metered apart from big ones. Tested against ChatGPT: the same rewrite there means pasting back to find out whether the score moved. Prompt: Rewrite this paragraph to use [term from the guidelines panel] naturally, keep it to three sentences, and cut anything that reads like marketing. Starter: Ask the question your customer would actually ask an AI Add this as a tracked prompt in the AI Tracker; Surfer's docs say tracked prompts refresh daily. Surfer reports two different things about that answer — whether your brand was named, and whether your page was among the sources the model read — and its documentation flags that these often disagree. Tested against a conventional rank tracker: a Google result either lists your page or it does not; an AI answer can cite you without naming you, or name you without citing you. The entry-level plan currently tracks zero prompts, so this one needs Standard or above. Prompt: What's the best way to [the thing your business does] in [your city]? Starter: Spend a document on a page you already published Open a new document for a phrase one of your existing pages should be winning, paste that page in, and read the gap list without writing anything new. Surfer counts this against the same monthly "documents" allowance as a fresh article, so auditing an old page costs what creating a new one does; check your best pages first, while the allowance is full. Tested against Google Search Console, free and already in most people's hands: it tells you a page lost position, not which subtopics the pages that overtook you cover. Pitfalls: - The pages Surfer compared you to can move, so a score is a reading from one day rather than a standing. - Writing to the term list mechanically shows: the panel gives a suggested count range for each term, and hitting the top of every range is how Surfer-written pages come to read like Surfer-written pages. Treat it as subjects to cover, not a quota. - Unused limits do not roll over. Surfer's help centre says monthly limits reset rather than accumulate, so a quiet month is money spent, not money banked. - Surfer's own AI page calls its generated articles "ready-to-rank." They are drafted from competitor pages, not from your business, so every factual claim about you in that draft is a guess until you check it. - The interface itself is available in English, Polish and French only, which is separate from which languages it can analyse or write in. - Surfer's plan names are still settling after the acquisition: its own August 2026 changelog refers to tiers the live pricing page no longer lists. Every price and limit here is the published position on the day it was checked. Where to next: - Other sales & marketing AI tools: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai - Text & chat assistants: https://www.launchpadhq.io/discover?category=text-conversational-ai - Workflow automation: https://www.launchpadhq.io/discover?category=workflow-automation ### Getting Started with Synthesia Canonical: https://www.launchpadhq.io/platform/synthesia/guide Data: https://www.launchpadhq.io/data/v1/guides/synthesia.json Last reviewed: 2026-09-07 Tool: Synthesia (https://www.launchpadhq.io/platform/synthesia): starting price Free, from $29/mo billed monthly; free tier yes; data as of 2026-10-25 Type a script, pick a presenter, get a finished video — no camera, no crew, no being on screen yourself. You write what you want said, choose an avatar (a synthetic presenter — a person-shaped video generated from text, not filmed), and Synthesia renders it speaking your words, lips matched to the audio. What it is: - A text-to-video tool built on synthetic presenters: you bring the script, it brings the face, the voice and the visuals. Beginners pick it over filming for unglamorous reasons — no need to be presentable, no editing to learn, and you re-render rather than re-shoot when a policy changes. - It aims squarely at workplace video: onboarding, compliance training, internal explainers, product walkthroughs. Its other strength is localisation — one script, many languages. - Versus HeyGen, also listed here: heavy overlap, both good. Synthesia leans corporate — templates, brand controls, team workspaces, translation for rolling one video across a company. HeyGen leans creator and social. Before you start: - There is a genuinely free plan, not a time-limited trial: the Basic tier is currently free, takes no card, and doesn't expire. - Free is small but real: a small monthly allowance of video minutes, a limited set of stock presenters, and a Synthesia watermark (a mark identifying output as AI-made) on what you export. - The mechanism to learn is a monthly allowance that resets and doesn't roll over. Length is what spends it, and anything unused at period end is gone. If you want the watermark gone or more video each month, you need a paid self-serve plan; prices depend on the billing period, and current plans are on the tool page. - First step: sign up free, open a template rather than a blank project, and paste 60 seconds of your own words over the sample script. Security: - Want an avatar of a specific real person, including yourself? Consent here isn't a checkbox, it's a recorded step: Synthesia requires that person to submit their own footage plus a separate consent video recorded live, reading an on-screen passcode aloud, and checks both show the same person. You cannot upload someone else's video and make an avatar of them. The friction is deliberate. Security: What Synthesia refuses, and what you should disclose: - Impersonating a real person or organisation in a way likely to mislead. Stock avatars are licensed actors — implying one endorses your product, or works at your company, is prohibited. - News, political, election and other polarising content using stock avatars, without written permission. - Stripping the watermark to pass a synthetic video off as filmed is a violation, not a workaround. - Adult content, harassment, hate speech, fraud, and medical or legal advice presented as authoritative — videos are moderated and do get rejected. - Tell your viewers. No law may force you to, but one line saying the presenter is AI prevents the far worse moment where they work it out themselves. Security: - Your scripts and uploads sit on Synthesia's servers so it can re-render them, under standard account terms rather than a negotiated contract. Keep unreleased financials, customer names and anything under a confidentiality agreement out of the script box. Best at: - Training and compliance video that gets rewritten whenever a policy changes - Turning a document or written procedure into a watchable explainer - Producing one script as video in many languages - Short product walkthroughs and how-tos at volume Okay at: - Marketing and social video — fine, but reads generic unless the script is good - Talking-head segments inside a larger edited video Avoid: - Anything needing trust that a specific human is speaking — an apology, a crisis message, a condolence, a hard announcement to staff. A synthetic presenter reads as evasion. - Personal-brand work that depends on you being present: podcasts, vlogs, founder stories - Emotional performance and comedy timing. Avatars deliver information well and feeling badly. - Video of anyone who hasn't consented on camera. Hard line, not a preference. Starter: Turn one policy into a 60-second explainer Synthesia's editor is built around the script box rather than a timeline, so paste-and-generate is the whole workflow — and a template means framing, background and captions are handled. Prompt: Here is how expense claims work from this month. Submit within thirty days, and attach a photo of the receipt — a bank statement is not enough. Anything over two hundred pounds needs manager approval. Starter: Rebuild the same video in a second language Its 1-Click Translation currently spans 160+ languages and translates on-screen text as well as the spoken script, so an existing video becomes a second one in seconds, not a second shoot. Prompt: Welcome to the team. Over the next two weeks you will meet your manager, set up your accounts, and complete three short training modules. Starter: Write a script that survives being spoken aloud There is no fixing a weak script in the edit here — no performance to salvage, no second take, no cutaway. The text is the ceiling on the video. Prompt: Three things change on Monday. The old login page stops working, so use the new one linked below. Passwords now expire every ninety days instead of thirty. And update the mobile app, or it signs you out. Pitfalls: - Writing for the page, not the mouth. Long clauses and nested parentheses read fine and sound terrible. Read yours aloud first: whatever you stumble over, the avatar will too. - Burning the allowance on drafts. It resets without rolling over, and every render spends from it, including ones you delete. Finish the script first. Where to next: - More AI video tools: https://www.launchpadhq.io/discover?category=video-creation-editing - AI voice and speech tools: https://www.launchpadhq.io/discover?category=voice-speech ### Getting Started with Tabnine Canonical: https://www.launchpadhq.io/platform/tabnine/guide Data: https://www.launchpadhq.io/data/v1/guides/tabnine.json Last reviewed: 2026-09-08 Tool: Tabnine (https://www.launchpadhq.io/platform/tabnine): starting price Free, from $12/user/mo; free tier yes; data as of 2026-10-20 The AI coding assistant built around never touching your code — now sold only to companies. You work in your normal code editor with Tabnine switched on. As you type, grey suggested text appears ahead of your cursor and you press Tab to accept it or keep typing to ignore it. A side panel lets you select some code, ask a question about it in plain English, and read the answer without leaving the editor. What it is: - Tabnine is an AI coding assistant that lives inside an IDE (the program a developer writes code in) — it finishes the line you are typing and answers questions about code in a side panel, not a browser tab. - Its selling point is what it does not do with your code, backed by models Tabnine trained itself on open-source code (code that is public, so anyone can inspect it) under licences that allow reuse. - As of September 2026 it sells to companies only — two quoted plans, no individual sign-up. Tricentis, a software-testing company, bought Tabnine in July 2026 and plans to fold its context technology into its own platform; releases are still shipping. Before you start: - There is no free tier, no trial, and no way to buy a single seat. Both plans are priced per user and billed annually, and both buttons say "Get a quote" rather than taking a card. You get Tabnine because an employer bought it. - If your workplace already has it, ask whoever runs developer tooling for a sign-in or a private Tabnine server address — that address, not the download, is what unlocks it. It installs into Visual Studio Code, JetBrains editors, Visual Studio or Eclipse. - If you are learning to code on your own, the blocker is not your skill level; Tabnine does not sell to individuals. Our AI coding tools category lists assistants with free tiers you can install today. Security: - Tabnine's whole pitch is data handling, and its wording is unusually specific for this market: "Your code is never stored," and requests are "only ephemerally processed to provide coding suggestions and are then immediately discarded." Security: Where that promise stops: - It is written around "our proprietary models." Switch to Claude, GPT or Gemini — other companies' models (the AI "brain" that does the actual thinking) — and Tabnine's own docs say the protection "may be different." - It is a promise to your employer, not to you: the plans include "advanced governance and analytics," so your usage is visible to whoever administers the account. Security: - What the money really buys is where it runs — a company's own servers, or "fully air-gapped," meaning machines with no internet connection at all. Best at: - Finishing the line or the whole function you are part-way through typing - Explaining a block of code you have highlighted, without copying it out into a browser - Working somewhere that cannot legally send code to a public AI service — banks, defence, healthcare - Generating tests and documentation for code that already exists Okay at: - Anything that is not about code — it is aimed at your editor, not at being a general chatbot - Raw capability from its own models, which trade some of it for the licensing guarantee Avoid: - Leaning on the privacy guarantee with an outside model selected — the docs say protection there "may be different" - Building a personal workflow on it: there is no individual plan, so the tool leaves when the job does - Assuming the licence checking is on — it ships as a preview, and blocking non-permissive code is a toggle someone has to switch Starter: Find out whether you already have it Ask whoever runs developer tooling for a Tabnine sign-in or a private server address. The current extension is "Tabnine for Enterprise" and connects to "your dedicated Tabnine server," so an administrator's address is what unlocks it — with GitHub Copilot you just buy a subscription. Starter: Check which model is answering you Open Tabnine's model selector and see whether you are on a Tabnine model or on Claude, GPT or Gemini. Tabnine publishes one data promise for its own models and a weaker one for the rest, so the selector changes what protects you — Copilot's terms cover every model it serves. Starter: Make it write the test, not the feature Tricentis bought Tabnine to feed software-testing tools, so test generation is the part of this product with a stated roadmap behind it — for Copilot it is one feature among many. Prompt: Write tests for the function I've selected. Cover the empty-input case and one failure case, using the test framework this project already uses. Starter: Ask where a suggestion came from Tabnine checks chat output "against the publicly visible code on GitHub" and names the project and its licence, where Copilot's public-code filter only suppresses a match without saying what it matched. Prompt: Where did this code come from? Show me any public project it matches, and that project's licence. Pitfalls: - Comparison sites still list a free Tabnine plan. They are out of date — the vendor's own pricing page shows two paid tiers and no free or trial option. - The old "(Legacy) Tabnine" extension is still in the Visual Studio Code marketplace and still installs, but it says Tabnine "does not onboard new users to this plugin" — installing it gets you nothing. - Suggestions are accepted with the Tab key, which is also how you indent. Early on you will accept code you did not want; keep typing to dismiss one rather than reaching for undo. Where to next: - AI coding tools you can start free: https://www.launchpadhq.io/discover?category=ai-coding-development - Run a model on your own machine: https://www.launchpadhq.io/discover?category=local-open-source-ai - Chatbots that also help with code: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Tactiq Canonical: https://www.launchpadhq.io/platform/tactiq/guide Data: https://www.launchpadhq.io/data/v1/guides/tactiq.json Last reviewed: 2026-10-14 Tool: Tactiq (https://www.launchpadhq.io/platform/tactiq): starting price Free, from $8/user/mo billed annually; free tier yes; data as of 2026-10-25 Writes down what's said in your calls from inside your browser — nothing joins the meeting. You install an add-on into your browser and join a call as usual. A panel opens beside the meeting and fills with the words as people speak. You can tag a line while it's happening. When the call ends the transcript is waiting in your account, where you can search it, share it, or ask it questions. What it is: - Tactiq is a meeting notetaker for Google Meet, Zoom and Microsoft Teams. It runs as a browser extension — a small add-on that lives inside Chrome or Edge — so the transcript builds up in a panel next to the call while you're in it, and lands in your Tactiq account when the call ends, alongside summaries and action items you can generate from it. - What makes it different from Otter and tl;dv, both also on this site: those send a bot — a robot participant that joins the call and records it, visible in the attendee list. Tactiq doesn't. It transcribes from your own browser, and Tactiq's docs put it plainly: "nothing joins the meeting as a separate participant." If your workplace blocks notetaker bots, or you'd rather not put one in front of a client, that is the whole reason to pick this one. - The flip side of the same design: Tactiq only captures meetings you are actually sitting in, in a browser. There's no calendar bot covering the meeting you skipped, and it can't attend on your behalf. Before you start: - Free forever, no card. You install the extension from the Chrome Web Store or Edge Add-ons, sign in, and join a meeting — there's nothing to configure on a Google Meet call. Start there rather than on Zoom, which needs a setting you may not control. - The free plan currently gives you a monthly cap on transcripts and a smaller one on AI credits (Tactiq's unit of spend — each AI action on a meeting costs one). The shape here is unusual and worth knowing before you plan around it: the Pro plan, priced per user, lifts transcripts to unlimited but only doubles the monthly AI credits. Unlimited AI credits don't arrive until the Team plan, which costs more per user. So transcription is cheap to scale and AI is not. - Zoom and Teams each have a prerequisite; Google Meet has none. For Zoom you must join in Chrome or Edge, and Zoom's own automated captions have to be switched on in the host's account — if you're not the host, that's their setting, not yours. For Teams, use the browser or the browser-installed Teams app; the desktop Teams app isn't supported. There's also a Mac-only app, still labelled beta, for transcribing Zoom's desktop app. - Pick the language being spoken before the meeting starts. How many languages you can choose depends on the platform you're on — currently 60-plus on Google Meet, around 40 on Teams and around 30 on Zoom. Getting set up safely (Install from tactiq.io/download — the vendor's own download page, which links to the Chrome Web Store and Edge Add-ons listings.): - Install only from that page or from the store listing it links to. Search results for popular extensions attract copycats, and an extension that sits inside your meetings is a bad thing to get wrong. - The extension asks to read the page you're on so it can see the meeting. If you also install the Mac app for Zoom desktop, it asks for two more permissions — microphone and system audio — because that version captures sound directly rather than reading the call from a browser tab. - Sign in with the same account you use for your meetings, so transcripts attach to the right calendar entries. Security: - The thing worth understanding about Tactiq is *where the work happens*. The extension transcribes from inside your browser tab, which is why nothing shows up in the participant list — and why, for AI features, your transcript text is sent on to Microsoft's enterprise version of OpenAI's models. Tactiq states that provider will not use the data to train its models and that content logging is switched off, so transcripts aren't retained there for review. Screenshots are the one thing Tactiq captures visually, and only when you press the button yourself. Security: What the other people in the call actually see: - Tactiq announces itself in the meeting chat. It posts a message in your name saying your Tactiq extension is transcribing. On the free plan this is always on and cannot be turned off — only Pro and above can disable it. Treat that as a feature: it does the disclosure for you. - The chat message is not the same thing as consent. Tactiq's Terms of Use — last updated June 2019, and still the version published today — put the legal duty on you: "It may be an offense in your jurisdiction to record other individuals without their prior written consent," and it is your responsibility to follow the law and to tell participants. The privacy policy sitting behind those terms was refreshed in May 2026, so where the two seem to disagree, the terms are the binding document. Say it out loud at the start; don't rely on a line of chat nobody read. - Sensitive calls deserve a decision, not a default. Anything covering someone's health, pay, or an HR or legal matter is worth asking about first, whatever the law where you sit. Best at: - Getting a written record of a call in places where an outside notetaker would be blocked, rejected, or simply awkward to explain - Meetings spread across Google Meet, Zoom and Teams landing in one searchable place instead of three - Catching the exact wording of something while it's being said — you can tag or comment on a line in the live panel Okay at: - Calls that switch between languages. Tactiq runs one transcription language at a time and doesn't follow a conversation that changes language mid-way; you have to switch it by hand. - Calls that aren't in English. Tactiq doesn't say anywhere public whether its summaries and in-meeting answers work as well in every language it transcribes, so it's worth reading the output yourself on a non-English call before you rely on it. Avoid: - Meetings you don't personally attend — if that is what you need, tl;dv or Otter on this site are built the other way round. - Treating your Tactiq account as your archive of record. The Terms state that on cancellation "all of your content will be immediately deleted from the Service" and "can not be recovered." Export anything you'd miss. Starter: Transcribe one Google Meet you're already in Install the extension, join any Google Meet, and watch the panel fill in on the right. Nothing else to switch on. Google Meet is the only one of the three where Tactiq's own docs say you do not need the platform's captions turned on — so your first run can't fail on a setting the host controls, the way a Zoom attempt can. If the panel stays empty for the first minute, the usual fix is updating the extension rather than anything in the meeting. Starter: Set the spoken language before anyone talks Open the Tactiq panel at the start of the call and set the transcription language to the one people are actually speaking. Tactiq will not reprocess a finished transcript in another language — a language change applies only to speech captured after the switch, so a call transcribed under the wrong setting is a call you have to run again. Starter: Tag the line you'll need later, while it's on screen When someone says the thing that matters — a decision, a number, a deadline — add a tag, label or comment to that line in the live panel, or take a screenshot if it's on a shared slide. Tactiq builds the transcript in front of you during the call rather than after it, so the moment you can mark something is the moment it's said. A recorded-then-processed notetaker gives you that only once the meeting is over. Starter: Ask a question during the meeting instead of after it This goes in the Ask Tactiq AI tab inside the meeting panel, which Tactiq's help centre says is available on free and paid plans alike — so you can use the twenty seconds before the wrap-up to check nothing got dropped, rather than discovering it a day later. Swap in the question you actually need: "What did we decide about the budget?" works the same way. If the AI comes back unavailable rather than simply used up, check your credit balance before assuming the feature isn't on your plan. Prompt: Summarise what we have agreed so far, and list anything that is still open with who needs to resolve it. Starter: Turn the transcript into the follow-up you owe someone Pasted into the AI chat on a saved transcript, this uses one of the same monthly AI credits as everything else — so it's worth spending on the meeting that actually generates work for other people, not on every call. Prompt: Write a short follow-up email to the people on this call. Open with the decision we reached, then list each action with the person responsible and the date it is due. Keep it under 150 words. Pitfalls: - Check the panel is showing words in the first minute. Tactiq's own troubleshooting guide says it is "not possible to retrieve transcripts from past meetings" — if the capture didn't happen, there's nothing to go back for. Thirty seconds of glancing at the panel is the whole insurance policy. - Leave the meeting tab alone once it's running. Tactiq doesn't document whether closing the tab, refreshing the page or letting your computer sleep interrupts a transcript in progress — and since the capture runs inside your browser, it is safest to assume it does. - Uploads spend the same allowance. You can upload a recording, audio file or existing transcript (up to 2GB) to get Tactiq's AI onto a meeting it didn't attend — but each upload counts as one of your monthly transcripts, the same as a live call. - The transcript is yours, not the meeting's. Because the capture happens in your browser, a colleague who wants their own copy appears to need their own Tactiq running in the same call — sharing yours looks like the intended route, but it's worth confirming once you're signed in. Where to next: - Meetings & Notes: https://www.launchpadhq.io/discover?category=meetings-notes - Voice & Speech: https://www.launchpadhq.io/discover?category=voice-speech - Workflow & Automation: https://www.launchpadhq.io/discover?category=workflow-automation ### Getting Started with TextGen (oobabooga's text-generation-webui) Canonical: https://www.launchpadhq.io/platform/text-generation-webui-oobabooga/guide Data: https://www.launchpadhq.io/data/v1/guides/text-generation-webui-oobabooga.json Last reviewed: 2026-09-13 Tool: Text Generation WebUI (oobabooga) (https://www.launchpadhq.io/platform/text-generation-webui-oobabooga): starting price Free; free tier yes; data as of not dated Run AI models on your own computer with the controls left visible — free, private, and built for people who want to tinker. You download a ready-to-run build, unzip it, and double-click it — a window opens on your own computer. You drop a model file (the AI "brain" that does the actual thinking) into its models folder, pick it from a menu, and type into a chat box. Answers come back locally, and you can adjust the settings and retry. What it is: - TextGen — the project most people still know as oobabooga's text-generation-webui — runs AI models on your own machine and shows you the machinery while it does. Where Ollama and LM Studio pick an engine for you and hide the rest, TextGen currently lets you choose among five different engines for the same model, switch between them without restarting, and change dozens of generation settings from a dedicated tab. That control is the entire reason to choose it. Before you start: - Hardware decides how this feels to use. Unusually, the project publishes no minimum spec — it publishes a memory calculator instead, which reads a model file and estimates the RAM (your computer's short-term memory — what runs out when too much is open) or graphics-card memory it will need. A graphics card makes it much faster, but ready-to-run CPU-only builds exist for Windows, Linux and macOS, so one isn't strictly required. - This is a control panel, not an appliance. You choose the model file, and usually the engine and the generation settings too — nothing is chosen for you, and nothing is hidden to keep the first run simple. If what you want is a local model working in five minutes with no decisions, start with Ollama and come back here once you have a specific setting you want to change. - You will not need to pay. It's free, the code is public, there's no account and no paid tier — the only real cost is disk space, and the longer installation currently wants around 10 GB before you have downloaded a single model. - Ready-to-run builds cover Windows, Linux and macOS, Apple Silicon included. One asymmetry worth knowing up front: the macOS ready-to-run build currently ships only the llama.cpp engine, so the engine-switching this app is known for is mostly a Windows and Linux story unless you do the longer installation. - The project renamed itself. It is TextGen now, and nearly everything written about it still says text-generation-webui or just "oobabooga." Same project, same maintainer — but it means much of what you find describes an older layout. Getting set up safely (There is exactly one official source: the releases page of the project's own repository, **github.com/oobabooga/textgen**. Job one is not trusting anything else. Because the project renamed, search results are thick with the old name — download mirrors, "installer" pages, forks and repackagers, none of them the project — and you are installing software that will have full run of your machine. Links to the old `github.com/oobabooga/text-generation-webui` address still land in the right place. Anything that isn't one of those two does not.): - Decide which installation you actually need. The ready-to-run build is a download-unzip-double-click affair that runs GGUF files (the single-file model format) through one engine and nothing else. The longer installation adds the other engines, fine-tuning, image generation and extensions. - Know where the models pile up. Model files run to several gigabytes each, they live in a user_data/models folder inside the app, nothing clears them out, and a handful of them will take up far more room than the app itself. - Confirm it worked. Loading one small model and getting a reply back in the chat box is your checkpoint — do that before you change a single setting, so you know what a working baseline looks like. Security: - The default here is already the private one, and the app does nothing behind your back to change it: it runs on your own machine, makes no telemetry or update calls of its own, and the interface it serves is reachable only from the computer it is running on. Nothing you type is sent anywhere. What can undo that is never a hidden setting — this is an app built to hand you switches, and three of them decide who else can reach your machine and whose code gets to run on it. Security: Three switches worth understanding before you flip them:: - Sharing the interface. One option makes the app reachable from other machines on your network; another puts it on a public web address through an outside tunnelling service. The project prints its own warning when you use either without a password — that you may be exposing the interface to the entire internet with no access control. A chat interface that can load models and read local files is not a thing to leave open. - Untrusted model files. A model is a file you downloaded from a stranger. Most are inert data, but some ship Python code beside them, and there is a setting — off by default — that allows that code to run. The project's own wording is blunt: switching it on lets models execute arbitrary code on your machine. Leave it off unless you know exactly whose model it is and why it needs it. - Community extensions. None load by default, which is the good news. But an extension is a Python file that runs inside the app with the same access to your files that you have, and the community directory is a list of links to other people's repositories rather than a reviewed catalogue. Installing one is running someone's program, not adding a feature. Best at: - Handing you the controls — model, engine, generation settings and context size all changeable from the interface rather than a config file - Running one model file through different engines to find out which is actually faster on your hardware - Free, offline and unlimited use - Being one window for several jobs — chat, free-form text, document and image input, image generation and LoRA fine-tuning Okay at: - First impressions — it opens onto a dense interface, because showing you everything is the point - Answer quality, which is a property of the model file you chose rather than of this app - Telling you what to pick — it documents thoroughly what every control does and takes no position on which you should use Avoid: - Using it as your first taste of local AI. Every choice it hands you is a choice you then have to make, and made wrong they produce a slow or mediocre result with nothing explaining why. - Expecting a hosted fallback when your hardware runs out. There is no cloud version of this to switch to — unlike Ollama, which sells one. Starter: Your first model: one small GGUF file from Hugging Face Downloads a single small model file from Hugging Face and drops it into the app's user_data/models folder. The interface detects it on its own — there is no separate import step — and it appears in the model menu ready to load. Ollama pulls from a library it curates, so the shortlist is made for you before you arrive. TextGen has no library: the ready-to-run build reads GGUF files you go and fetch yourself, which makes picking the file — and its size — your job from the first minute. Start small, so a wrong choice costs you ten minutes rather than an evening. The compression level is in the filename. A Q4 file is roughly half the size of a Q8 of the same model, and a little worse at everything. Starter: Run the same model through a different engine Selects a different engine in the Model tab and reloads. The project currently lists llama.cpp, ik_llama.cpp, Transformers, ExLlamaV3 and TensorRT-LLM. In LM Studio or Ollama "which engine" is not a question you can ask; here it is a dropdown, and on the same file the answer changes measurably. Starter: Change one generation setting at a time in the Parameters tab Opens the dedicated Parameters tab, where the settings that shape an answer — temperature, top-p, min-p, repetition penalties and a long list besides — are exposed as controls, including the order in which they get applied. LM Studio gives you a few of these behind a panel. TextGen gives you the full list plus a sampler-priority control that reorders how they are applied, which most local apps have no interface for at all. Change one, regenerate the same message, compare. That loop is what the tab exists for, and it is what the extra friction of this app buys you. Temperature first. Low makes it repetitive and literal; high makes it wander. Every other control is easier to judge once you have felt that one. Starter: Try the Notebook tab instead of the chat box This is a separate tab with no chat turns in it at all, where the model simply continues whatever text is on the page. Handing it an unfinished sentence rather than a question shows you what the thing underneath is actually doing — continuing text — which is precisely what the chat format papers over. Ollama's app is a chat window and has nowhere to do this. Paste in a half-written paragraph of your own, let it carry on, then edit its continuation and run it again from there. Prompt: The three things nobody tells you about learning to sail are Starter: Generate an image without leaving the app Loads an image model in the app's dedicated image tab and makes pictures on the same machine, with a gallery that keeps what you made along with the settings that made it. With Ollama or LM Studio, an image means a second application and a second set of downloads. TextGen currently ships an image tab with its own memory-saving options, which matters on a modest machine: a text model and an image model competing for the same graphics memory is the usual reason one of them refuses to load. If the image model won't start, unload the text model first. There is an idle-timeout setting that will do it for you after a few quiet minutes. Pitfalls: - Downloading a model bigger than your machine can hold. The catalogue here is the whole of Hugging Face with no size gate anywhere. - Trusting a well-ranked blog post. Prefer the project's own wiki, however well the blog post ranks. - Pasting in a command that contains a sharing flag. Guides written for cloud notebooks include them because they have to; on your own computer the same flag turns a private app into a reachable one. - Extension dependencies quietly breaking the app. Extensions bring conflicting versions of the app's own requirements with them. If things stop working right after you added one, suspect that first. Where to next: - Local & Open-Source AI: https://www.launchpadhq.io/discover?category=local-open-source-ai - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with the Anthropic API Canonical: https://www.launchpadhq.io/platform/anthropic-api/guide Data: https://www.launchpadhq.io/data/v1/guides/anthropic-api.json Last reviewed: 2026-10-14 Tool: Anthropic API (https://www.launchpadhq.io/platform/anthropic-api): starting price $0.10 input / $0.50 output per million tokens; free tier no; data as of 2026-10-26 The same models as the chat assistant, billed per token — and under a contract with the opposite privacy default. Your own code sends text to Anthropic's models and gets a response back, instead of you typing into a chat window. You create an account on the developer console, generate a key, and your program sends requests with that key attached. You read what comes back, change the request, and send it again. What it is: - A pay-per-use interface to the same model family behind the Claude chat assistant. - Aimed at people building something — a tool, a script, a product feature — rather than at people who want to chat. - Billed by usage with no subscription floor, which means near-zero cost while you experiment and real cost at volume. - Governed by commercial terms, which have a different and stricter privacy default than the consumer chat product. Before you start: - You need to write or run code. There is no interface here — the console has a testing workbench, but the product is the endpoint. - Pricing is per million tokens and differs sharply by model: the largest models cost many times what the smallest do, and the text a model writes back costs more than the text you send it. Anthropic's pricing page lists the current rate for each model, including older ones. - Batch processing halves the cost in both directions for work that does not need an immediate answer. If your job can wait, this is the single largest saving available. - Set a spending limit in the console before your first real run. Runaway loops are the most common way beginners get an unpleasant bill. Getting set up safely (Anthropic developer documentation): - Create an account on the developer console — this is separate from any chat subscription you may already pay for, and one does not include the other. - Generate an API key and treat it like a password. Anyone holding it can spend your money. - Store the key in an environment variable, never inside your code. Keys committed to a public repository are found and abused within minutes. - Set an organisation spend limit before writing anything that runs in a loop. - Start with the cheapest model that could plausibly work. Moving up is easy; discovering you overpaid for a month is not. Security: - This is the same company as the Claude chat assistant and the default is reversed. API traffic falls under commercial terms, which means it is not used to train models by default, and inputs and outputs are deleted from backend systems within about thirty days. The consumer chat product trains on personal-plan conversations unless you opt out; the API does not train on them at all. Security: Practical implications: - Routing work through the API rather than a personal chat account changes the contract you are operating under, not just the interface. - Submitting feedback is the material exception to the no-training baseline — feedback you send can be retained and used. - Zero-retention arrangements exist for qualifying organisations, but they are arranged through sales rather than switched on in a dashboard. - Content flagged for policy violations is retained longer regardless of your configuration. Security: - The transferable lesson: which door you walk through determines which contract you are under. Two interfaces to the same underlying model can have opposite data defaults, and neither one advertises this at the moment you paste something sensitive. Best at: - Building the same operation into a repeatable pipeline - Processing many items in bulk, especially where batch pricing applies - Embedding assistant behaviour inside a product you are building - Work where a stricter data contract is a requirement rather than a preference Okay at: - One-off tasks, where the chat interface is faster and cheaper - Cost prediction, which is straightforward per call and surprising in aggregate Avoid: - Casual use — a subscription is cheaper for conversation - Any project without a spending cap configured - Shipping a key in client-side code, ever Starter: Make one call from the console before writing any code Uses the console's testing workbench to send one message to a model and read the reply, with no program involved. Confirms your key, your account, and your billing work, without a debugging session in the way. Send something deliberately trivial. You are testing the plumbing, not the model. Starter: Price your actual job before you build it Runs one representative item, then multiplies by your real volume. Per-call costs look negligible and per-month costs frequently do not. Do this arithmetic while it is still cheap to change your mind. Price it against the cheapest model too. The difference is often larger than the quality gap. Starter: Move a bulk job to batch processing Submits a set of requests together for processing later at half the per-token rate, instead of sending each one and waiting for its answer. Half price in both directions is the biggest lever available, and it costs nothing but patience. If you find yourself waiting on results interactively, the job was not a batch job. Starter: Put the key in an environment variable on day one Stores the key in your system's settings, where your program reads it at run time, so it never appears in the code itself. Every leaked-key story starts with "I was going to fix that later". Add the environment file to your ignore list in the same commit, not the next one. Pitfalls: - A chat subscription does not include API credit. They are separate products with separate billing. Paying for one buys you nothing on the other. - The promotional rate expires. The discounted mid-tier price ends on 2026-08-31 and reverts to the standard rate. If your budget assumes the lower number, it breaks that day. - Loops are how people get surprised bills. A retry that does not stop, or a batch larger than you thought, runs up cost quickly. Set the spend limit first, not after. - Keys in public repositories are found within minutes. Automated scanners watch for them continuously. Environment variables are not a best practice here, they are the minimum. - The most capable model is rarely the right first choice. Start cheap and move up only when a specific failure justifies it. The price gap between tiers is far larger than the quality gap for most tasks. Where to next: - Other developer APIs: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - Run models locally instead: https://www.launchpadhq.io/discover?category=local-open-source-ai ### Getting Started with the OpenAI API Canonical: https://www.launchpadhq.io/platform/openai-api/guide Data: https://www.launchpadhq.io/data/v1/guides/openai-api.json Last reviewed: 2026-07-13 Tool: OpenAI API (https://www.launchpadhq.io/platform/openai-api): starting price $2.50/$10 per 1M tokens); free tier yes; data as of 2026-10-26 The developer doorway to OpenAI's models — build your own apps on the same AI behind ChatGPT, billed by usage. You create an account, generate an API key (a secret credential that identifies your account), and your code sends a request to OpenAI's servers with that key attached. The model's response comes back to your program, which does whatever you wrote it to do with it. You adjust the request and send it again. What it is: - The OpenAI API is how developers build their own software on OpenAI's models (text, images, speech-to-text, text-to-speech, embeddings) — the same underlying models as ChatGPT, but called from your own code instead of a chat window. It's for people writing applications: a chatbot, a summarizer, a coding tool. It is not the ChatGPT app, and a ChatGPT subscription does not include API access — they're separate products with separate billing. Before you start: - There's no real free tier for building. New accounts may get a small one-time trial credit after identity verification; after that it's pay-as-you-go, billed per token — for the text going in and the text coming out, at a per-model rate, with output usually costing more than input. (There's a fringe "free daily tokens in exchange for sharing your traffic for training" program — skip it for anything real; see security.) - Set a hard spend limit in the billing dashboard before your first project. Agent-style or high-volume usage can run up a bill fast; a monthly cap is your safety net. - You'll write (or paste) a little code to make your first call — this page assumes you're comfortable with that. If you're not coding yet, start with the ChatGPT app instead. Security: Your API key is a password — treat it like one: - Anyone with your key can spend your money. Never commit it to a public repo, paste it into a webpage, or share it in a screenshot. - Store it in an environment variable or a secrets manager, not in your code. - If a key leaks, revoke it immediately in the dashboard and issue a new one. Security: - The important data point most beginners get wrong: standard API traffic is NOT used to train OpenAI's models by default (unlike the consumer free ChatGPT). The one exception is the opt-in "free tokens for data-sharing" program — that route does use your inputs and outputs for training, so don't use it for anything sensitive or proprietary. Security: - You are responsible for what your app produces and for the data you send. Don't send other people's personal or confidential information through the API without the right to do so, and add your own checks on model output before showing it to users. Best at: - Building custom apps on OpenAI's models from your own code - Access to a broad toolkit in one place — text, images, speech, embeddings - High-volume or automated tasks where a chat window won't scale - Fine-grained cost control by routing simple work to cheaper models Okay at: - Predictable flat cost (it's usage-based, so bills vary with traffic) - Beginners who don't code (the ChatGPT app is the better entry point) - A ready-made interface (you build the UI; the API is just the engine) Avoid: - Exposing your API key anywhere public - Running with no spend limit set - The 'free tokens' data-sharing route for anything sensitive Starter: Create a key and set a budget Sign up, generate an API key, and set a monthly hard spend limit in the billing dashboard before you write any code. The budget cap is the single thing that prevents a surprise bill; do it first. Also set a lower "soft" alert so you get an email before you hit the ceiling. Starter: Store the key safely Put your key in an environment variable (not in your code), so it never ends up in a repo or a screenshot. A leaked key is someone spending your money; this is the habit that prevents it. Using GitHub? Add your env file to .gitignore immediately. Starter: Pick a cheap model to start For learning and testing, choose one of OpenAI's smaller, low-cost models rather than the flagship — it's plenty for a first project and keeps costs near zero. Most simple tasks don't need the most expensive model; starting cheap builds good cost instincts. Move up to a bigger model only when a task actually needs it. Starter: Your first call A single round-trip shows the whole loop — key, request, tokens billed, response — with almost no cost. Swap the summarize task for something from a project you have in mind. Prompt: Send a request that asks the model to summarize a short paragraph of text into one sentence, and print the response. Pitfalls: - The API and ChatGPT are separate products. Your ChatGPT Plus subscription does not include API usage; the API is billed on its own. - No spend limit = a real risk. Agent loops and high volume can rack up charges quickly; set a hard cap first. - "Free" here has a catch. The only ongoing free route trades your data for model training — fine for throwaway tests, wrong for anything sensitive. Where to next: - AI APIs & Developer Services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - AI Coding & Development: https://www.launchpadhq.io/discover?category=ai-coding-development ### Getting Started with the Zillow Zestimate Canonical: https://www.launchpadhq.io/platform/zillow-ai-zestimate/guide Data: https://www.launchpadhq.io/data/v1/guides/zillow-ai-zestimate.json Last reviewed: 2026-09-12 Tool: Zillow AI / Zestimate (https://www.launchpadhq.io/platform/zillow-ai-zestimate): starting price Free, from See pricing; free tier yes; data as of not dated The free home-value number on Zillow listings — and how far off it usually is. You type an address into Zillow's search box, open that home's page, and scroll to the Zestimate — a dollar figure with an estimated sales range beneath it. There is nothing to install and nothing to set up. You read the number, read the range, and decide how much of it to believe. What it is: - The Zestimate is Zillow's guess at what a home is worth. It sits on the pages of well over 100 million US homes, and Zillow says it refreshes most of them multiple times a week, whether or not the home is for sale. - There is no dashboard and no settings: it is a number someone else's computer already worked out about your house, and the whole skill is reading it properly. - Zillow calls the current version the Neural Zestimate. It is a model (the AI "brain" that does the actual thinking) trained on public county records, prior sales, market trends and feeds from real-estate listing services. That is also the shape of its blind spot — it has never been inside the house. - How wrong is it? For a home actually listed for sale, Zillow currently reports a median error of around 2% — *median* meaning the middle case, so half of all homes land further off. For a home that is not for sale — Zillow's word is "off-market," the category your own house sits in — Zillow said in April 2026 that typical error has come down to about 7%. Before you start: - Free, and you don't need an account to look. Type an address into the search box at zillow.com and the Zestimate is on the home's page. There's no paid version and no upgrade to hit. - It's a US number. Zillow counts its coverage in US homes, so a house in another country won't have a Zestimate — though Zillow doesn't state a geographic limit outright anywhere we could find. Zillow does say plainly that estimates are for residential property only: large multi-family buildings, vacant land and commercial properties aren't eligible. - Some homes simply don't have one. Zillow withholds the figure where it doesn't have, in its words, "sufficient reliable data to generate an estimate that meets our accuracy standards" — incomplete or inconsistent property records, or an area with few recent sales. If you find a blank where the number should be, nothing is broken. - If the home is yours, you can claim it — confirm to Zillow that you're the owner, which opens an owner dashboard where you can correct the facts the estimate is built on: bedrooms, bathrooms, finished square footage, lot size, year built. Claiming is free, and expect to need a Zillow sign-in for it even though Zillow's claim page doesn't say so outright. Security: - The usual chatbot warning doesn't apply, because you aren't typing anything sensitive in. The flow runs the other way: Zillow has already published your home's bed and bath count, square footage, sale history and an estimated value on a page anyone can open, assembled from public county records, without asking you. Security: What that means in practice: - Looking up a home — yours, a neighbour's, anyone's — is invisible to whoever owns it. - Claiming your home is the one place you hand Zillow something. In return you get an owner dashboard, which Zillow describes as tools to "track your home's value, explore pricing options for a potential sale, and receive the Home Report email." That's a selling-oriented product and a recurring email. Go in expecting that, not a privacy control. - If what you want is your home *off* Zillow rather than corrected, that's a different question — and not one Zillow's claim page answers. Security: - None of this makes looking at a Zestimate risky. It's worth knowing because the reflex this page invites — "let me go fix my number" — is the step that hands a company a verified owner and an address. Best at: - Telling you roughly what price neighbourhood a home is in, free, in about ten seconds - Direction over time — whether an estimate is drifting up or down is steadier than the exact figure on any one day - Homes that are ordinary for their street, where Zillow has plenty of recent nearby sales to learn from - Being the number you walk in with, so a conversation with an agent starts somewhere Okay at: - Your own off-market house — the model works from records that may be years stale - Rentals — there's a separate Rent Zestimate, built from rental listings and public data, which Zillow presents as a starting point rather than an appraisal or a guarantee Avoid: - Anything an appraisal (a licensed professional's on-site valuation, the kind a lender orders) is for. Zillow's own Zestimate page says it flatly: "It is not an appraisal and can't be used in place of an appraisal." A lender will order its own regardless. - Treating it as advice you can act on. Zillow's Terms of Use, effective 28 October 2025, state that the services are provided "AS IS," "WITH ALL FAULTS" and are "NOT INTENDED TO provide you with any financial, real estate, or related advice of any kind." The number and that disclaimer come from the same company. - Setting an asking price by it, or arguing an offer down with it. Zillow's own valuation guide says the estimate "should be used as a starting point, but it shouldn't be the only data you use in determining a home's value." Starter: Look up the house you actually live in Type your own address into Zillow's search box and open the home's page. Read the Zestimate, then the estimated sales range under it, then the home facts beside it: bedrooms, bathrooms, finished square footage, year built. You are the only person here who knows whether those are right. Zillow's stated inputs for an off-market home are public county records, prior sales and market trends — nothing from inside the house. So the only correction available to you is fixing the facts underneath the number, which Zillow says feeds straight back into the estimate. If a fact is wrong, claiming the home is what lets you change it. Zillow says the update is instant, but that not every edit moves the value — a few square feet won't; an extra room should. That guidance comes from a 2015 Zillow help article and nothing newer contradicts it. Starter: Compare a listed home to an unlisted one Pull up a home that's currently for sale and see how close its Zestimate sits to the asking price. Then pull up a neighbouring home that isn't for sale and look at that estimate instead — the second is the one telling you what the model can do unaided. Zillow lists the asking price among the inputs it feeds the model for a home on the market. That's why the two error rates above diverge so sharply, and why the off-market figure is the one that applies to your house. Zillow breaks its accuracy statistics out by region as well as by listed/unlisted, so your own state's figures are worth more to you than the national one. Starter: Read the range before you read the number Under the headline figure Zillow shows an estimated sales range — a low and a high, the span of prices the home could reasonably sell for. Read the width of that band first: a wide band is the model saying it isn't sure. Zillow ties the width of that band to how much data it has for the area, which makes it the one place on the page where the model states its confidence rather than its answer. Then look the same home up on Redfin's estimate — where two independently built models diverge, that gap measures how little either one knows about this house. Starter: Ask Zillow's own AI about the number — if you have it Zillow has been rolling out what it calls AI mode: a conversational layer where you type a question in plain English instead of setting filters. Zillow's March 2026 announcement lists "explain and contextualize the Zestimate" among its jobs, and gives "How has this home's Zestimate changed over time?" as a sample question. The catch is availability — as of Zillow's own July 2026 update it was still in beta, open to a limited group of users with access "expanding over time," so there's a real chance you go looking and find no such box. Separately, Zillow's ordinary search bar does accept plain-English descriptions like "Austin TX homes with backyard" — though Zillow's own pages disagree about where: a September 2024 announcement says the iOS and Android apps with the website to follow, while a Zillow help page updated in April 2026 says the iOS app only. This is the only Zillow surface where the Zestimate becomes something you can put a question to, and its own sample questions are about the estimate's history rather than about finding listings. If you can't find it, nothing is lost — the other three things on this page work from the plain listing page. Pitfalls: - Believing the precision. The Zestimate is printed to the dollar — a formatting choice, not a claim. The range underneath it is the truer statement of what Zillow knows. - Assuming a renovation is in there. The off-market estimate is built from public records, so a new kitchen registers only if a permit or a reassessment put it on file — or if you told Zillow yourself. - Editing your home facts too late. Zillow's guidance is that home-fact changes need to be made *before* the home is listed for sale on Zillow. - Reading a missing Zestimate as a verdict on the house. Where the number is absent, that's a statement about Zillow's records, not about the property. Where to next: - Finance & Real Estate AI: https://www.launchpadhq.io/discover?category=finance-real-estate-ai - Research & Academic Tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with tl;dv Canonical: https://www.launchpadhq.io/platform/tl-dv/guide Data: https://www.launchpadhq.io/data/v1/guides/tl-dv.json Last reviewed: 2026-09-07 Tool: tl;dv (https://www.launchpadhq.io/platform/tl-dv): starting price Free, from $18/seat/mo billed annually; free tier yes; data as of 2026-10-25 Sends a bot to your video calls so you can stop typing notes and actually listen. You connect your calendar and your video-call app, and tl;dv sends a small bot that joins the meeting and records it. When the call ends you get a recording, a searchable transcript, and AI-written notes you can skim in a minute instead of rewatching an hour. What it is: - An AI meeting notetaker for Zoom, Google Meet and Microsoft Teams: it records the call, writes up what was said, and lets you search every meeting you have recorded — or ask its AI a question across a whole set of them. For people who sit in a lot of calls and lose the details. - Why this over Otter, also on this site: Otter is strongest on straight transcription, tl;dv leans on the *video* — clip a 60-second moment and send it, though free plans allow only a handful — and its free tier has no monthly minutes budget. - Why this over Zoom's or Teams' own notetaker: those cover only their own platform. tl;dv sits across all three, so meetings from different apps land in one library. Before you start: - Free to start, no card: sign up, connect Google Calendar or Outlook, and choose which meetings get recorded. The calendar connection is the one to think about. - The free tier's shape matters more than its numbers. Recording and transcription are generous; what is rationed is the *AI*: your first handful of meetings get full AI notes, then automatic notes cover only the opening minutes, and cross-meeting questions have a small allowance. Recordings age out — archived after days, deleted after months. - Record a couple of calls a week and want only the notes, and you won't pay. You hit the wall wanting full AI notes on *every* meeting, recordings kept long-term, or them pushed into a CRM, the system a sales team keeps its customer records in. Paid plans are priced per person and are cheaper billed annually; current prices are on its Launchpad page. - First step: connect your calendar, set auto-record to internal only, and record one with your own team. Don't point a bot at a customer on day one. Security: - Recording other people is a legal act, not a settings toggle. Consent rules vary by where each person sits: some places need only one person on the call to agree (that can be you), others need *everyone* to agree first. Mixed country or US state on the call? Assume the strictest rule applies. Practically: say out loud at the start that you're recording and why, let people object, and don't record anything sensitive — HR issues, medical or legal matters, someone's pay — without written agreement. None of this is legal advice; if the stakes are real, ask someone qualified. Security: What connecting your calendar actually hands over: - Every event on your calendar — titles, times and the email addresses of everyone invited, including meetings tl;dv never records. - Enough to auto-record meetings you don't even attend, if that setting is on — plus write access to edit your invites, if you turn on the consent flow. - Attendee emails can be matched to contacts in a connected CRM, quietly turning your calendar into sales data. Security: Where your meetings live afterwards: - tl;dv states it holds SOC 2 Type II — an audit of how a company handles customer data — encrypts stored data, keeps its data centres in Europe, and does not train AI models on customer data. - Deleting a meeting in the app is described as a permanent wipe, not a hide. - Sharing rules can automatically make recordings visible to your team or whole organisation based on things like the meeting title. Check those rules before you record anything you'd rather keep to yourself — a badly worded rule is how a 1:1 ends up org-wide. - Retention on the free plan is a few months, and free recordings generally can't be downloaded. Get a call you care about out while you can. Best at: - Recurring internal meetings where nobody wants to be the notetaker - Sales calls where you need to remember exactly what was promised - User research and interviews, where the exact words matter Okay at: - Very large webinars — transcripts get long and speaker labels get muddled - Being your only system of record; it captures what was said, not what was decided Avoid: - Any call where you haven't told people you're recording - Anything covered by medical, legal or financial confidentiality - Genuinely private one-on-ones, especially with org-wide sharing rules on Starter: Record one internal meeting first Sets the auto-record preference to internal meetings only, then lets the bot join one call with your own team. tl;dv's auto-record preference splits meetings by email domain, so "internal" means everyone on the invite shares your company domain — a safe sandbox where you learn what the bot looks like, how long processing takes and how good the notes are. Starter: Turn on consent collection before your first external call Enables consent collection in tl;dv's settings and limits it to external meetings. tl;dv builds this in: it can rewrite the meeting link in your calendar invite so consent is asked before anyone joins, and you can scope it to external meetings. Test it with a friend first. A declined consent can't be undone for that meeting. Starter: Build a meeting template Creates a notes template for one meeting type and assigns it, so that type's write-ups follow your structure rather than the default. tl;dv saves templates and applies them per meeting type, so your standup and your sales calls get written up differently — decisions, action items, objections rather than a generic summary. Most beginners skip it. Pitfalls: - Auto-record on "all meetings" is a firehose you'll regret — a library of five-minute check-ins, and a bot showing up somewhere awkward. Start narrow. - "Internal only" is decided by email domain, not by vibes. A contractor, an agency or a co-founder on a personal Gmail counts as external, so your "safe" setting can still put a bot in front of an outsider. - Sharing rules are set once and forgotten. A rule written to share "all sales calls" will happily share the one a customer assumed was private. Re-read them after a month. Where to next: - More meeting and note-taking tools: https://www.launchpadhq.io/discover?category=meetings-notes - Connect your notes to your other tools: https://www.launchpadhq.io/discover?category=workflow-automation ### Getting Started with Together AI Canonical: https://www.launchpadhq.io/platform/together-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/together-ai.json Last reviewed: 2026-10-14 Tool: Together AI (https://www.launchpadhq.io/platform/together-ai): starting price Free, from $0.04/M input tokens; free tier yes; data as of 2026-10-14 Two hundred-plus open AI models — text, images, video, speech — on one prepaid account. There is no free trial: it starts with a small prepaid top-up on a card. You sign in, put a small amount of money on the account, then open Together's playground in a browser, pick a model (the AI "brain" that does the actual thinking) from a long list, and type into a box. An answer comes back and the cost comes off your balance. Swap the model, ask again. What it is: - Together AI is infrastructure, not an app. It runs openly published models — made by other labs, not by Together — on its own machines and rents access through an API (a way for programs to talk to each other without a person clicking). Chatbots built on DeepSeek, Qwen or Llama often have something like Together underneath. - What sets it apart from the other rent-a-model companies is range: one account and one balance reach 200-plus models across about ten kinds of work, so a video clip and a page of text bill to the same place. It will also train a model on your own examples, or rent you the machines themselves by the hour. There is a plain chat window too — Together Chat, hosted in the US and Canada — separate from the developer platform. Before you start: - There is no free trial, and that is the first thing to know. Together removed its free tier in a July 2025 billing change; reaching the platform now means putting a minimum amount of credit on a card up front, and at a zero balance your access is suspended until you top it up. If you want to try open models without paying, Groq's free tier or a model on your own computer are the honest starting points — come back once you know what you want to build. - For text, that minimum goes further than you would expect: serverless chat is billed per million tokens at low rates, and a few models currently cost nothing per use. Video is the exception, billed per clip at a few cents each. - Nothing to install — a browser covers everything here. Together Chat may be usable without the card: it launched free with a daily allowance, though Together no longer publishes its limits, so treat what you get there as a bonus rather than a plan. Security: - What you type runs on Together's servers. Its privacy policy is unusually direct: Together says it does not use data collected from you to train its models without your explicit opt-in, and that consent is revocable. A zero-retention option in the Privacy and Security settings goes further, stopping your text and the answers being stored at all. Security: Practical rules: - Together's terms explicitly prohibit sending financial, medical or sensitive personal data — a rule you agree to at signup, not general caution. - Make an API key (a password that identifies your app, and that spends your money) only when you have code to put it in. - Set a usage limit before switching on auto-reload, which buys more credit from your card whenever the balance runs low. Best at: - Breadth on one bill — text, image, video and speech models from many different labs, on a single account - Running the big open families — DeepSeek, Qwen, Llama, GLM — so the machine bill is not yours - Growing without switching vendor: pay-as-you-go, training on your own examples, or whole machines by the hour Okay at: - Being looked around cheaply — the minimum top-up is a small door charge, but it is a door charge - The hardest reasoning and coding, where the leading closed models are still ahead - Stability of any single model name; the catalogue turns over constantly Avoid: - Sending financial, medical or sensitive personal data — Together's own terms list it as prohibited use, not merely discouraged - Hardcoding a model name. The published retirement policy gives ordinary models two to three weeks' notice and preview models under 24 hours, and an upgrade redirects your requests to a new version after three days - Making it your first ever AI tool — with no free trial there is nowhere to look around before paying Starter: Spend the first dollar in the playground, not in code Picks a mid-sized chat model in the browser playground and asks it something you already know the answer to. Together takes the minimum top-up before it shows you anything, so the number going down is your own prepaid balance, not a promotional grant burning off. Watching one real request cost real money is the fastest way to build accurate cost instincts. Starter: Ask one question of three models from three different labs Sends the same question to a DeepSeek, a Qwen and a GLM model, changing only the dropdown. The catalogue passes 200 models across about ten kinds of work, so one balance buys comparisons that are not limited to text: answer the question, then generate a picture of the answer from the same account, billed per megapixel rather than by the word. Then make one short video and stop. Video bills per clip, a few cents each, and it is the only thing here that visibly moves a small starting balance. Starter: Try Together Chat before deciding you need the platform Opens Together Chat — the plain chat window, separate from the developer platform — and asks it the questions you came with; Together's own pages suggest it works without the card, though the limits aren't published. Same open models in an ordinary chat window, on machines in the US and Canada, and the one route here that does not begin with a card. If it does everything you came for, what you wanted was a chatbot rather than an AI cloud — and you have saved the top-up. Pitfalls: - The minimum top-up is a door charge, not a subscription. People arrive expecting a free tier that has not existed since July 2025, and find they cannot send a single request. - Reading 'free model' as 'free account'. A model or two in the catalogue currently costs nothing per use, but you still need the funded balance to reach it. - Assuming Together built these models. It hosts open models made by other labs; whether one is any good this month is a question about DeepSeek or Qwen, not about Together. Where to next: - More developer platforms like this one: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - Running open models on your own machine, for free: https://www.launchpadhq.io/discover?category=local-open-source-ai ### Getting Started with Topaz Labs Canonical: https://www.launchpadhq.io/platform/topaz-labs/guide Data: https://www.launchpadhq.io/data/v1/guides/topaz-labs.json Last reviewed: 2026-09-12 Tool: Topaz Labs (https://www.launchpadhq.io/platform/topaz-labs): starting price $199; free tier no; data as of not dated Sharpen, clean up and enlarge photos and video you already have — a set of separate paid apps, not one. You open a picture or a video file in the app, choose from a short list of AI models (the AI "brain" that does the actual thinking), and a preview pane shows you the before and after side by side. You nudge sliders until it looks right, then export the result as a new file. What it is: - Topaz Labs isn't one app. It's a company selling several separate programs for improving pictures and video you already have: Topaz Photo (grain, softness and damage in still photos), Topaz Gigapixel (making a small image much bigger without it going mushy) and Topaz Video (the same idea for footage). There are browser and iPhone versions too, plus Bloom, for tidying up AI-generated art. They're sold separately, with an all-apps bundle called Topaz Studio. - None of them invent a picture from nothing: you bring a file that already exists and the software rebuilds it at higher quality. That's why photographers and video editors run these alongside Photoshop or Premiere rather than instead of them — and what you're paying for over the enhance button in an editor you already own is the choice of which model does the work and how hard it pushes. Before you start: - Start with your computer, not your credit card. Topaz's published minimums for the desktop apps are 16 GB of RAM (your computer's short-term memory — what runs out when too much is open) and, on Windows, a dedicated graphics card with at least 6 GB of its own memory; the newer models that invent detail want 8 GB. Topaz Video doesn't run on Intel Macs at all, and each app wants roughly 30 to 70 GB of free disk space. - Everything Topaz sells today is a subscription, monthly or yearly, renewing automatically. The buy-it-once-and-own-it-forever licences the company was known for were discontinued in September 2025 — an existing one keeps working, but you can't buy one now. Individual apps and the all-apps bundle are priced by the year, and monthly billing costs noticeably more. Topaz discounts often and deeply, so check the price on the day you buy. - There's no free trial of the current apps, which is unusual for software this dependent on your hardware. Topaz's published refund policy gives a money-back window of about two days from purchase, and its support staff have described that as the deliberate substitute for a trial. Two more things to check before you click: an annual plan can't be cancelled mid-term without a fee, and some plans advertised at a monthly price are still a twelve-month commitment. Topaz does still host its discontinued older apps — Photo AI, Gigapixel AI, Video AI and others — and its documentation says these "function as trial software by default"; the docs don't say what that mode restricts, and posts on Topaz's own forum describe trial output carrying a visible stamp, so expect some limit of that kind. They are no longer developed or supported, so treat them as a look, not a substitute. - If your machine can't meet the specs, the browser version is the way around it. Topaz says it renders "regardless of hardware," with the work done on its servers instead of your graphics card — image plans there are a monthly subscription. Video in the browser is sold with credits (the platform's unit of spend — each thing you make costs some) rather than flat use. Getting set up safely (Download only from topazlabs.com/downloads. This matters more here than for most software: Topaz is expensive, popular and heavily discounted, which makes it a favourite name for fake "cracked" installers — and an installer you run has the run of your whole machine.): - You'll be asked to log in with a Topaz account the first time an app opens — the licence follows your account rather than the machine, which is what lets you move it to a new computer later. - Run your own real files through it the same day you install: one big photo, one genuinely bad photo, and if you bought the video app, a clip the length you actually work with — the money-back window is short, and a full export takes far longer than a preview suggests. Security: - Where your file gets processed decides everything else here. Run the desktop apps and the work happens on your own machine — your photos never leave it. For a wedding photographer, a medical image, or anything you're contractually not allowed to share, that is the reason to prefer installed software over a website. Security: What changes the moment you use the cloud: - Your files get uploaded. The browser version, the phone app and the cloud-rendering option inside the desktop apps all send your media to Topaz's servers, and its licence agreement, updated February 2026, has you grant Topaz a licence to store, copy and modify the file to the extent needed to do the job. - You keep ownership. The same agreement states that as between you and Topaz you retain all ownership rights in your media, and Topaz makes no copyright claim over your files and no claim to royalties. - Share links are public. If you hand someone a result using Topaz's sharing links, the agreement says that file stays hosted for public access until you ask for it to be removed. - One silence worth noticing: Topaz's published privacy policy doesn't address whether media you upload for cloud processing is used to train or improve its models. It doesn't say it is, and it doesn't say it isn't. Stay on the desktop apps and the question doesn't arise. Best at: - Enlarging a photo far past the size it was shot at — Topaz advertises up to around sixteen times the pixels - Rebuilding old or low-resolution video so it is watchable on a modern screen Okay at: - Faces. There's a dedicated face-recovery model, but Topaz's own documentation notes its strongest version won't run on cards with less than 8 GB of graphics memory — and a face is exactly where anything invented shows first - Repeatability. Topaz's docs say of its most creative model that "there is no guarantee that outputs will match every time" Avoid: - Anything where the picture is evidence. These tools don't recover detail that was recorded and lost — they generate plausible detail that was never there. For an insurance claim, a legal exhibit, a medical image or an archival scan that isn't an improvement, it's a fabrication, and Topaz's own documentation concedes the point when it tells you to use a non-generative model "if your artwork requires accuracy." - Commercial work on the wrong licence. The February 2026 agreement restricts the Personal licence: if you work for an organisation with annual revenue over one million US dollars, commercial use needs a Pro licence. Topaz's own definition of commercial use covers advertising, published or printed media, stock platforms, game and media assets, and paid distribution on video platforms. - Feeding the output to another AI. The same agreement forbids using Topaz's output, or the models behind it, to develop, train, validate or improve any other AI system. Starter: Make a small photo big enough to print Takes an image that looks fine on a phone screen and rebuilds it several times larger. Gigapixel makes you choose between a true-to-the-original enlargement and its Wonder models, which invent detail to fill the gap and are built for sources that were already poor. Topaz's documentation steers you towards the non-generative option whenever accuracy matters. Compare at 100% zoom, never fit-to-screen. Detail the software made up is invisible zoomed out and obvious at actual size. Starter: Fix a dark, grainy phone photo Clears the speckle you get shooting in low light and sharpens what's underneath. Topaz Photo opens with a feature called Autopilot switched on: it inspects the file and applies a set of fixes before you've touched anything. That is a head start and a trap, because its confident default is easy to mistake for the right answer. Switch Autopilot off once and rebuild the same edit by hand. Starter: Rescue an old family scan Lifts dust, scratches and surface damage off a scanned print, and softness lost in the original photograph or the scan. Topaz Photo carries a mode built specifically for scanned and damaged prints — it looks for the failure modes of paper and scanners, not of cameras. Scan at the highest resolution your scanner offers before you start. Starter: Make old video watchable on a big screen Rebuilds footage shot years ago at low resolution, frame by frame, at modern sizes. Topaz Video has you pick a model per job and renders a short preview before you commit — its quick-start guide points beginners at Proteus first. That preview matters more here than in the photo apps: a full export can run for hours on the machine that produced the preview in seconds. Preview a few seconds of the worst part of the clip, not the best. If the difficult shot holds up, the rest will. Starter: Steer a generative enlargement with a written description Gigapixel's Redefine model carries the only text box anywhere in Topaz's desktop apps. The documentation attaches this "Image description" field to that model and no other, says it "helps guide the AI for desired results," and advises a descriptive statement rather than a directive — so you describe what the picture is, not what you want changed. Watch the Creativity slider beside the box — it decides how far the software is allowed to stray from your photograph. Prompt: A close-up photograph of an elderly man's hands holding a worn leather book, warm indoor light, shallow depth of field, fine skin texture and visible leather grain. Pitfalls: - Confusing the current apps with the discontinued ones. Topaz Photo and Photo AI are different products; so are Gigapixel and Gigapixel AI, and Video and Video AI. Search results, YouTube tutorials and forum answers mix them up constantly, and the older ones have different features and no support. - Cloud credits vanish at the end of the month. On the plans that include them they reset monthly and don't roll over, and video consumes them in proportion to length, resolution, frame rate and which model you picked — so one long clip can clear a month's allowance in a single export. Where to next: - Image generation and editing: https://www.launchpadhq.io/discover?category=image-generation-editing - Video creation and editing: https://www.launchpadhq.io/discover?category=video-creation-editing - Gaming and creative AI: https://www.launchpadhq.io/discover?category=gaming-creative-ai ### Getting Started with Tray.ai Canonical: https://www.launchpadhq.io/platform/tray-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/tray-ai.json Last reviewed: 2026-09-17 Tool: Tray.ai (https://www.launchpadhq.io/platform/tray-ai): starting price Custom; free tier no; data as of not dated Enterprise-grade automation you get by asking a salesperson, not by signing up. Tray's own walkthrough describes a browser canvas. You pick a trigger (the "when" — the event that starts an automation), then add steps underneath it, each one an app you have been given permission to use. You run the whole chain by hand and read what each step received, before you let it run on its own. What it is: - Tray.ai is the automation platform a company buys, not the one a person signs up for. It connects the systems a business runs on — Salesforce, NetSuite, Workday, Slack, a data warehouse — and keeps data moving between them without anyone copying it across by hand. - There is no free plan and no credit-card signup. Pricing is not published: the three plans on the pricing page (Pro, Team and Enterprise) carry feature lists but no dollar amounts, and the buttons say "Get a demo" and "Talk to sales." A free trial exists, but you request it on a form and wait for someone at Tray to set it up. If you want to build an automation tonight, start with Zapier, Make or n8n instead — they are in the same category on this site and all three let you in on your own. - The company renamed itself from Tray.io to Tray.ai in July 2024 and rebuilt its story around AI. Alongside the classic automation builder it now sells Merlin for building AI agents (AI that takes actions on its own rather than only answering). The platform is actively maintained — its public release notes show updates through 9 September 2026. - Its public connector directory says "474+ pre-built connectors" while its homepage and pricing page say "700+ apps," so treat the exact number as approximate; either way the list covers the enterprise systems a small tool usually skips. Before you start: - You cannot buy this by yourself, and that is the single fact to plan around. There is no self-serve signup — the login page has no create-account link — so every route in runs through a demo request or a sales conversation. - The free trial runs about 14 days, according to Tray's own Academy page. You request it by filling in your name, work email, job title, company and employee count, and describing what you want to accomplish. The page calls it a "managed trial" and says someone from the team will reach out to get it set up — so the clock starts when they provision it, not when you decide you are ready. - Tray's Master Subscription Agreement, last updated 28 January 2025 and in force today, is built around a negotiated Order Form and subscription terms that renew automatically for successive one-year periods unless either side gives 30 days' notice. This is an annual commitment with a contract behind it, not a monthly charge you cancel in a settings screen. - If you are reading this to learn automation rather than to buy it for an employer, the concepts transfer completely. Triggers, steps, testing a run before switching it on — build those on a platform you can open right now, and come back to Tray when a company hands you a login. Our Workflow & Automation category has the self-serve options. - Tray's documentation and connector directory are public and need no account, so you can read exactly what a given integration does before you ever speak to anyone. Security: - On Tray you are usually not the account owner, and that changes the whole shape of the risk. An IT or platform team stands up your workspace, decides which connectors and stored connections you are allowed to touch — Tray's governance controls are per-workspace, per-connector, per-connection and per-tool — and reads the log afterwards. Every workflow run and agent action is recorded with who triggered it, what ran and what came back, and those logs can be streamed to the company's own monitoring system. The question to ask on day one is not "what should I connect?" but "what am I permitted to connect, and who reviews it?" Security: Three things worth knowing before you build anything real:: - Ask which environment you are in. Tray separates development, staging and production behind approval gates, so the version of your automation that touches live customer records is one somebody signed off. Find out early whether you have that sign-off or need to request it. - Masking happens in the workflow, not after it. Tray can strip or mask personal and health data as it flows through a workflow. Whether it is switched on for your workspace is an admin's decision, so ask rather than assume. - Know where the AI steps send your data. Tray's documentation says its built-in Merlin features run on foundation models hosted on AWS Bedrock from providers including Anthropic and Amazon, that the models do not retain customer data after processing, and that OpenAI powers Merlin Chat and Build as an opt-in sub-processor restricted from training on your data. Admins can disable Merlin features entirely. That is Tray's own account of it, worth confirming against your employer's policy. Best at: - Connecting the systems a company actually runs on — the CRM, the finance system, the HR system, the data warehouse — where the integration has to survive staff turnover and an audit. - Letting several teams build their own automations inside boundaries one central team sets, rather than everyone wiring up their own accounts. - Reshaping messy data mid-flow, without a separate step to do it in. Okay at: - A single person automating their own work. It will do it, but you are paying for governance machinery you have no use for. - Getting started quickly. Between the trial request, the sales call and the provisioning, the gap between deciding and building is measured in days. Avoid: - Learning automation from scratch on your own schedule. You do not choose when the trial starts, and its days tick down whether or not you have time that week. - Assuming the compliance features come with the plan. Tray's pricing page lists HIPAA, single sign-on, regional hosting, 30-day log retention and 180-day insights as add-ons. Hosting defaults to the US (AWS-West); EU (Ireland) and APAC (Sydney) are available, but as something you buy rather than something you select. - Anything you need to keep running after the trial. There is no free tier underneath to fall back to, so whatever you build stops when the trial does unless a purchase is already in motion. Starter: When a deal closes in Salesforce → then create the customer everywhere else it needs to exist One closed-won record sets off the whole chain: the customer account in the finance system, the onboarding project in Jira, the channel in Slack, the row in the warehouse. The account IDs and field mappings this needs live in a project config — one place every workflow in the project reads from — instead of being retyped into each step. That is what keeps the build standing the day someone renames a Salesforce field. Ask what happens on failure before you ask what happens on success — Tray's own first-workflow walkthrough adds error handling as a step of its own. Starter: When a support email arrives → then read it, look up the account, and route it The workflow pulls the message, fetches that customer's contract terms from the system of record, has an AI step summarise and classify it, and sends it to the right queue. This is where Tray's masking earns its keep: personal and health data can be stripped before it ever reaches the AI step — a control set above the person building the flow rather than inside it. Starter: Every night at a set time → then pull records, reshape them, and load them somewhere else A schedule fires, the workflow reads from a database, rearranges the fields into the shape the destination expects, and writes them across. Reshaping is the step that breaks most automations. Tray added inline expression queries in September 2026, so a transform can run directly in a step's input field rather than needing a separate transformer step of its own — fewer boxes, and fewer places for a mapping to rot. Run it by hand against one night's real data and read the execution log for every step before you put it on a schedule. Starter: When an AI agent needs to touch a real system → then it goes through a governed tool, not the system itself Instead of handing an agent direct credentials, the agent calls a tool Tray exposes on its behalf, with a rule about which tool, which credential and which data it may reach. The agent never holds the credential, and each tool it may call is permissioned separately. Note this sits in the Agent Development add-on rather than the base plans, so it is a purchase decision, not a switch you flip. Pitfalls: - The trial is scoped to the sentence you wrote on the form. You describe what you want to accomplish, and the workspace is set up around that. Write that field as the thing you genuinely most want to test, not a vague summary. - The absence of a published price is by design. Go into the sales conversation with the shape of your usage already written down — which systems, how many people building, how often things run — or the quote will be built from someone else's assumptions. - You will probably not be the person who owns the connections. In most Tray installations an admin creates and holds the credentials, so "connect my Gmail" is a request you file rather than a button you press. Find out who that person is on day one. - Log retention is shorter than you would guess. The Pro tier's insights window is 7 days and Team's is 30. If your automation fails quietly and nobody looks for three weeks, on the lower tiers the evidence may already be gone. - Don't judge it against Zapier on first impressions. Tray asks more of you up front — expression queries, environments, approvals — and pays it back on the kind of integration that has to survive an audit. On a two-app automation that trade is simply a bad deal. Where to next: - Workflow & Automation: https://www.launchpadhq.io/discover?category=workflow-automation - AI Plugins for Business Software: https://www.launchpadhq.io/discover?category=ai-plugins-business-software - Sales, Marketing & SEO: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai ### Getting Started with Trinka AI Canonical: https://www.launchpadhq.io/platform/trinka-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/trinka-ai.json Last reviewed: 2026-09-15 Tool: Trinka AI (https://www.launchpadhq.io/platform/trinka-ai): starting price Free, from $20/mo; free tier yes; data as of not dated A grammar checker built for research writing, where the specialist words your usual checker underlines are the ones that are right. You paste or type your text into Trinka's web editor and it underlines problems as you go. Click an underline and you get the suggested fix plus a short reason; accept it or ignore it. Keep editing and it keeps rechecking, then you copy the cleaned-up text back out or download the file. What it is: - Trinka is a grammar and style checker built specifically for academic and technical writing, from Crimson AI — the company behind the research-editing service Enago. - What separates it from a general checker is as much what it leaves alone as what it catches. It is tuned not to "fix" field-specific terminology, the passive voice of a methods section, or the hedged phrasing research writing uses deliberately — and it checks things a general tool never looks at: style-guide conventions, consistency across a whole document, and whether a manuscript carries the declarations a journal will ask for. - Checking works on English and Spanish text, and for English you choose US or UK spelling. The site's menus come in several other languages, but that is the interface, not the checking. - Around the editor sits a set of separate, mostly pay-per-use research tools — plagiarism and AI-detection reports, citation checking and formatting, a journal finder — which is what makes it feel larger than a grammar checker. Before you start: - Signing up is free and needs no card. The free Basic plan gives you a monthly allowance of language checking, a separate allowance for the paraphraser that Trinka's feature page describes as roughly the same size again but does not itemise in its plan table, and a handful of AI-writing requests. When you reach a limit Trinka stops rather than quietly giving you something weaker — its own wording is that you "cannot continue using Trinka", which suggests it halts rather than warns you first. The counter resets a month from the date you registered, not on the first of the month. - The free plan also drops a few credits (the platform's unit of spend — each report costs some) into your account every 30 days. Unused free credits expire at the end of each 30-day period. Credits you buy never expire. - Paying is worth it for specific things rather than in general. Premium — billed monthly or yearly, with prices shown by region, so check the figure on its pricing page before you buy — unlocks the deeper "Power" grammar mode, the journal style guides (APA, AMA, ACS, AGU, IEEE), the full set of paraphrase modes and the Microsoft Word add-in. It does not include plagiarism reports, which still come out of credits. The browser extension, the consistency check and the journal finder are already on the free plan. Security: - The plan you are on decides the answer here. Trinka's terms of service say text from free-plan users is used to train its AI, and that every paid plan is exempt — so the tier this page starts you on is the one tier without that protection. If you are working on an unpublished manuscript, that is the fact to weigh before you paste it in. Security: - The posture around that is stronger than most tools this size: AWS servers in the US, encryption in transit and at rest, SOC 2 (an audit of how a company handles customer data) Type 2 and ISO 27001:2022, and a public Trust Center that sets out what happens to your files. Security: Worth knowing before you paste a manuscript in: - Paid does not mean deleted straight away — Trinka's Trust Center says Premium users' processed documents are cleared from its servers every 90 days. - Deletion immediately after processing is a feature of the separate, much pricier Confidential Data plan, not of an ordinary paid account. - The commitments above live on the pricing page, the Trust Center and the terms of service (last updated 15 April 2024). The privacy policy itself still carries an effective date of 1 January 2021 and does not spell out the plan-by-plan training and deletion rules. Best at: - Manuscripts, theses and technical reports, where correct jargon is what every other checker underlines - Catching inconsistencies that accumulate across a long document — colour and color, ultra-violet and ultra violet, TNF-alpha and TNF-α - Checking a paper against a journal's style guide before you submit it - Writing formally in English as a second language, where each suggestion arrives with a short reason attached Okay at: - Everyday writing — email, blog posts, marketing copy. It is tuned for formal prose and will push casual writing toward stiffness. - Spanish, which is supported for checking, though the academic machinery around the editor appears to be built for English — if Spanish is your writing language it is worth confirming which of the extras apply to it. Avoid: - Treating the free plagiarism check as a plagiarism report. The free check gives you a similarity number; the report that shows which sources matched costs credits, and the version that also searches paid journals costs more. - Assuming "unlimited" paid checking is unlimited. Trinka's terms set a monthly fair-use ceiling on both language checking and plagiarism scans. - Reading the journal finder's result as a verdict on your paper. It matches your abstract against journal scope, which produces a shortlist — not an assessment of whether the work is strong enough for any journal on it. Starter: Paste a real paragraph of your own and read only the explanations Runs the free grammar check over something you actually wrote and shows a one-line reason beside each suggestion. Tested against Grammarly, the signal here is the underlines you do not get. Reading only the explanations is the fastest way to find out whether Trinka's idea of correct academic prose matches your field's. Use a methods section if you have one — that is where passive voice is densest and a general checker is noisiest. Starter: Set your English variant, then run the consistency check across a whole document Choose US or UK English, then run Consistency Check over the full document to surface spellings, hyphenations, numerals and symbols you have used both ways. A sentence-by-sentence checker, Grammarly included, has no opinion about a spelling you used two ways forty pages apart. Trinka runs this check over the whole document at once, which is the only way a manuscript-length inconsistency surfaces. Save it for last, after the writing is finished — it is a tidying pass, not a drafting one. Starter: Run your abstract through the Journal Finder Paste your abstract and get a ranked list of journals whose published scope matches it, filterable by publisher, region and open-access status, with journal metrics alongside. It flags predatory journals (ones that take a fee and publish almost anything) — the part of journal selection a first-time author cannot easily do by hand, and the part no grammar checker attempts. Filter for open access first if your funder requires it, before you look at anything else. Starter: Ask Trinka to tighten one sentence without touching your terminology The /Ask field — typed as /ask in the editor, or opened by selecting a sentence — is the one place in Trinka where you write an instruction instead of clicking a suggestion. Telling it to leave your terminology alone is the instruction worth learning, because it is what the checker around it is already doing for you. Free requests are few, so spend one on the sentence you have already rewritten four times yourself. Prompt: Rewrite this sentence to be more concise and more formal. Keep every technical term exactly as written, and do not change what it claims. Starter: Run a publication readiness check before you submit Import your manuscript file and run the technical checks — more than 25 of them, covering ethics approval and consent statements, conflict-of-interest and funding declarations, author names and affiliations, figure and table captions, clinical trial IDs, and references that are obsolete or retracted. A paper bounced before review for a missing declaration is not a writing problem, and no grammar checker looks for one. The report costs a single credit from the free monthly allowance. Run it before the final proofread, so there is still time to go and get the declarations you are missing. Pitfalls: - AI editing can trigger a disclosure requirement, and the line is not where most people assume. Elsevier's generative-AI policy for journals (June 2026) states that basic checks of grammar, spelling and punctuation need no declaration, but that substantive changes to sentence structure or to how the text is organised must be disclosed — which is precisely what the paraphraser and /Ask do. The ICMJE recommendations go further, asking authors to disclose AI-assisted technologies used in producing a submission and to describe writing assistance in the acknowledgements. Check your target journal's own policy; they are not identical. - The free plan runs the shallow mode. Basic gets Lite mode grammar checking, so trying the free tier tells you whether you like the tool, not what it does at its best. - Trinka's own pages disagree about the AI detector, in two different ways. Its feature page presents AI content detection as free, while the plan-comparison table meters it in credits per block of words. The same two pages also disagree on the included allowance for Premium Plus. A separate free web version exists, capped per session and per day. Trinka has not reconciled either point, so assume the plan table governs your account and verify both once you are signed in. - It will still argue with your field. Specialist notation, gene names and unit formats do get flagged. Build up the personal dictionary early rather than overriding the same term fifty times. Whether that dictionary carries across the web editor, the browser extension and the Word add-in is not stated publicly, so that is another thing to check once you are signed in. - There are no refunds. Trinka's terms state that refunds are not allowed on purchased subscriptions; you can cancel to stop the renewal, but the period you have paid for is spent. Where to next: - General AI assistants for drafting: https://www.launchpadhq.io/discover?category=text-conversational-ai - Other research and academic tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Working with PDFs and documents: https://www.launchpadhq.io/discover?category=document-pdf-processing ### Getting Started with Udio Canonical: https://www.launchpadhq.io/platform/udio/guide Data: https://www.launchpadhq.io/data/v1/guides/udio.json Last reviewed: 2026-09-07 Tool: Udio (https://www.launchpadhq.io/platform/udio): starting price Free, from $10/mo; free tier yes; data as of not dated Make surprisingly good music from a sentence — just know that, right now, the song stays on Udio. You describe a song — genre, mood, instruments, the feeling you're after — and Udio plays back audio it made from that. You can supply your own words or let it write them. It builds short pieces first; you extend the ones you like and redo the parts you don't. What it is: - A text-to-music generator: describe a song in plain English, get a finished-sounding recording with vocals, instruments and structure. No instrument, no studio, no music theory. - It's built around editing rather than rolling dice: once something is close, you extend it, regenerate a weak section, or restyle it, rather than starting over. - Before you compare it with Suno, settle one thing. Udio's draw is a detailed, textured production sound and finer control over reworking a track — but its current terms bar downloads and limit you to personal, non-commercial use. If you want a file you can keep, post or release, that isn't Udio right now, and check Suno's own terms before assuming it is either. Before you start: - Free to start, and the free tier is real: no card, a small daily credit allowance plus a separate monthly pool, and a daily cap on full-length (about 2 minutes 10 seconds) generations. Sign up at udio.com and put one sentence in the prompt box — the message you type. - How credits work. Credits are the platform's unit of spend, and Udio generates in pairs: a pair of 130-second songs costs twice what a pair of 32-second clips does. Monthly credits don't roll over; separately purchased packs don't expire. - You won't need to pay to try it. You'll hit the ceiling wanting longer songs, more per day, or the upload and voice-editing features, which are paid-only. Paid plans are monthly (cheaper prepaid annually), from an entry plan with a free trial up to a pricier top tier — prices and credit amounts change, so check the pricing page. Security: Rights and ownership — read this before you make anything you care about: - Udio's Terms of Service (last revised 12 November 2025) say Udio and its licensors own all right, title and interest in the output, and that any ownership that would vest in you is assigned to Udio. - The same terms say you may use output "solely for your personal and non-commercial purposes" and that you may not download copies of your output "for any purpose." - Downloading of audio, video and stems is currently disabled in the product — Udio confirmed this in its own help article about the Universal Music Group partnership. - They also ask you not to republish or distribute generated songs on streaming or user-generated-content platforms; YouTube, TikTok, Spotify, Instagram and SoundCloud are named. - Warning: parts of Udio's own help centre predate the settlement and say the opposite — that you own your songs and can use them commercially. Those pages are stale; the Terms of Service are the binding document. Security: - Udio is a social site as much as a tool: by default other people can remix, extend and restyle songs you make. You can change that per song under "Edit permissions," but only one at a time — there's no way to set it library-wide. Treat what you generate as potentially visible. Security: - Uploading your own audio is paid-only, and Udio requires that you own what you upload. Under the terms, uploading grants Udio a broad, worldwide, perpetual and irrevocable licence to use, store and modify it, including to train its AI — and what you type and what comes back can be used for training too. Best at: - A plain-English description turned into a structured, listenable song in under a minute - Genres and production styles you couldn't play or record yourself - Reworking a track you almost like: extending, regenerating a section, restyling - Auditioning different musical treatments of the same words — the style prompt and the lyric field are separate Okay at: - Songs built around lyrics you wrote — phrasing and pronunciation wander - Precise structure: exact bar counts, key, tempo. You steer, you don't dictate - Long-form pieces, which take repeated extends and burn credits fast Avoid: - Any commercial release or paid client work. The current terms grant ownership to Udio, permit personal non-commercial use only, and disable downloads. Don't build anything commercial here until you can confirm otherwise from Udio's own terms. - Prompting in the name of a living artist for their voice or style — the terms prohibit impersonation, it's the fastest route to a legal problem, and it's unfair to the artist - Anything needing a file — a podcast intro, a game soundtrack, video background music — because you can't currently take the audio off the platform - Work you'd be upset to lose — it lives on Udio's servers under terms that have already changed sharply once Starter: Make your first song from one sentence Udio responds to descriptive production language — instruments, texture, room, mood — rather than commands, so stacking short phrases is the native way to steer it, at a couple of credits a go. Change one word — warm to sparse — and regenerate. Prompt: dreamy indie folk, female vocal, fingerpicked acoustic guitar, soft brushed drums, warm and a little melancholy, late autumn evening Starter: Extend a clip into a full song Takes a 32-second clip you like and uses Extend to add the next section, repeating until it is a full-length track. Udio composes in pieces rather than one shot — the core of how it differs from tools that hand you a fixed track. Longer generations cost more credits and free accounts have a daily cap on full-length ones, so extend deliberately. Describe what happens next — a break, a key change, a stripped-back final chorus. Starter: Fix one section instead of rerolling the whole song Regenerates or restyles just the part you don't like, leaving the rest intact. This is where the credit system rewards you: patching a section costs far less than re-rolling a whole track. Beginners hit regenerate until something clicks; editing instead is what changes your results. Pitfalls: - Almost every guide online is out of date. The rules reversed after Udio settled with Universal Music Group in late 2025, but reviews and videos still describe output you own and can use commercially — and some of Udio's own help pages haven't been updated either. The Terms of Service govern. - People still arrive expecting an MP3. There was a brief window after the settlement when downloads reopened so users could save existing work, and it closed. Confirm any platform's export terms before spending a week on it. - Burning your allowance on long songs. Full-length generations cost roughly double a short one per track, and free accounts can only make a handful each day. Prototype at 32 seconds until the style prompt is right, then commit credits to length — unused ones are gone at the reset. Where to next: - More music generators: https://www.launchpadhq.io/discover?category=music-generation - Voice and speech tools: https://www.launchpadhq.io/discover?category=voice-speech ### Getting Started with Unriddle (now Anara) Canonical: https://www.launchpadhq.io/platform/unriddle/guide Data: https://www.launchpadhq.io/data/v1/guides/unriddle.json Last reviewed: 2026-09-09 Tool: Anara (https://www.launchpadhq.io/platform/unriddle): starting price Free, from $16/mo; free tier yes; data as of 2026-10-20 Read a stack of research papers by asking them questions — and click straight to the passage each answer came from. You upload your PDFs into a folder, then type a question in a panel beside them. The answer comes back with small numbered markers in it, each pointing at the source behind it. You go on asking, add more files, and write your own notes in the same window. What it is: - Unriddle is a research workspace for reading papers. The company renamed it Anara in 2025 — unriddle.ai now redirects to anara.com — so the name on the site will not match the name that brought you here. - The real alternative is attaching a PDF to a general chatbot, and the difference is that this keeps a library. Your papers stay uploaded between sessions, and a question can be aimed at a whole folder rather than one file. - Those citation numbers are the point of the tool. Anara's help centre says clicking one opens the source document and scrolls to the exact passage the answer used, so checking a claim is meant to take a click rather than a re-read. - There is also a writing editor in the same window, so a draft and the sources it rests on sit side by side instead of in two applications. Before you start: - The free plan needs no card, and its allowance refreshes daily rather than monthly — the help centre says free limits reset on a rolling daily basis. That makes the free tier genuinely usable for a reading week rather than a one-time trial. - One cap will decide whether this works for you: a single upload on the free plan is limited to 20MB or 120 pages. A journal article fits with room to spare. A thesis, a textbook chapter set, or a scanned course reader often will not, and the limit is per file rather than per month, so it does not go away by waiting. - Free connects to Zotero and Mendeley — the reference managers most students already keep their PDFs in — so an existing library can come across without re-uploading anything. The academic-database connectors start on the Plus plan, and cloud-storage connectors on the higher Pro plan. Most people reading this will not need either. - When you run out, the tool stops rather than getting quietly worse: new questions, AI edits and file imports pause, while reading, annotating and every answer you have already received keep working. Buying extra credits (the platform's unit of spend) is a paid-plan option only, and the help centre says support cannot raise a free limit. Security: - Anara's security page is unusually specific for a product this size: encryption in transit and at rest, SOC 2 Type II certification (an audit of how a company handles customer data), and a flat statement that it never uses your documents or chats to train AI, in any form, on any account rather than as a setting you have to find. Security: - Where that promise is written down matters more than the promise. It appears on the security page. The binding privacy policy's own explicit no-training sentence is scoped to a narrower case, and the policy separately lists AI providers among the companies your text is passed to — Anara says those providers keep nothing after answering. So the strong version is a vendor statement rather than a clause you could point a supervisor at, which is the right way to weigh it if the answer matters institutionally. Security: Where the stakes actually are: - Unpublished work — a draft manuscript, a thesis chapter, a paper someone sent you before publication. Published papers are already public; those are not, and they are what a research tool ends up full of. - If your work is funded, ethics-approved or otherwise regulated, check your institution's rules on cloud processing before the library gets large. Moving four hundred files later is a worse day than starting somewhere else. Best at: - Finding which of your forty PDFs actually contains the thing you half-remember reading - Pulling the same detail out of many papers at once — sample sizes, dates, effect sizes — into one table you can scan - Getting a plain-language read on a paper from outside your field - Keeping a library that persists, so next week's question does not start with re-uploading everything Okay at: - Questions about the open web — it can search, but a general chatbot handles anything not in your library better - Scanned pages and figures — text recognition and image understanding start on the Pro tier, so a photographed page can upload fine and then behave as though it is nearly blank - Writing prose. The editor is real and it cites as it goes, but this is a reading tool before it is a writing tool Avoid: - Patient-identifiable or clinical material on any plan below Enterprise. The security page says the company is compliant with US health-privacy law, but the additional agreement that actually covers patient records is offered to Enterprise customers only. - Routinely uploading paywalled PDFs from a university library subscription. The terms make you warrant you have the right to provide anything you upload, and a site licence to read a journal is not obviously a right to hand it to a third-party service. - Group work on the free plan. Folder collaborators are listed from Plus upward and the free tier's description does not mention them, so a shared folder is not something to build a group project on without checking first. Starter: Ask one paper what it cannot show Because Anara's help centre says a citation click lands you on the passage the answer used, the third part is checkable in a way a summary is not: a limitation you cannot land on in the source is a limitation to distrust. Add "and quote the sentence where they say it" to make the check faster still. Prompt: Read this paper and tell me three things in plain language: the question it set out to answer, what it actually found, and what the authors themselves say the study cannot show. Cite the passage for each of the three. Starter: Turn five papers into one table Anara can create and edit spreadsheets from data in your documents and run small calculations on what it pulls out — so the comparison ends as a file you can sort, not a paragraph you have to re-read. Ask it to add a column for study design, then sort by it — the odd one out is usually the interesting one. Prompt: Across every paper in this folder, build a table with one row per paper and these columns: authors and year, sample size, population studied, main outcome measured, headline result, and one line on the study's biggest limitation. Leave a cell blank rather than guessing, and cite each row. Starter: Bring in the library you already have Those connectors being on the free plan is what makes a first session useful with a hundred papers in it rather than one. Run it again in a month; the same question over a bigger library is a different answer. Prompt: Across my whole library, which papers disagree with each other about their main finding? For each disagreement, name the two papers, state each one's position in a sentence, and cite the passage where each says it. Starter: Draft a paragraph where the citations come with it Anara offers around a dozen citation formats, including the ones supervisors actually ask for. It also tells you, in its own help centre, to check that each citation represents what the source says — take that instruction literally. Set your citation style in preferences before you start rather than reformatting afterwards. Prompt: Using only the papers in this folder, draft one paragraph of about 150 words summarising what this literature agrees on and where it is unsettled. Attach a citation to every claim, and do not make a claim you cannot cite. Starter: Look for what you have not read yet Anara searches published literature from inside the same window as your own files, so a new paper lands in the library it will later be questioned in — the specialist database connectors, though, start on the paid tiers. Ask for the five most-cited and the three most recent separately; they are rarely the same papers. Prompt: Based on what is already in this folder, what important work am I missing on this topic? List up to eight papers I do not have, with one sentence each on why it matters and how it relates to what I already have. Say clearly which ones you could not verify. Pitfalls: - A summary can read as complete and still have skipped the limitations section. The retrieval decides the answer: a passage that was never fetched is simply not in it, and nothing marks the hole. A confident three-line summary of a paper's findings that never mentions the small sample, the single site, or the funding source is the most common way this tool misleads a careful person. - A citation being real is not the same as the citation supporting the sentence it is attached to. The format will be correct and the paper will exist. Whether that passage says what the claim says is the part only you can check. - The rename catches people out in a specific way: a great many tutorials, reviews and course guides still say Unriddle and describe an older interface. If a walkthrough you find never mentions Anara, treat its screenshots and its stated limits as out of date. - Your department's rule is the one that binds you, and nothing in the product signals where the line is. Anara publishes no position on acceptable academic use — no honest-use policy, no statement about coursework — and its terms contain no academic-integrity clause at all, while assigning you the rights to whatever it produces. Reading and searching with it rarely raises an eyebrow anywhere. Text its editor drafted, and citations it chose for you, are precisely what disclosure policies are written about, and those policies are usually set per course rather than centrally. Check the rule for the specific module before, not after. Where to next: - More PDF and document tools: https://www.launchpadhq.io/discover?category=document-pdf-processing - Research and academic tools: https://www.launchpadhq.io/discover?category=research-academic-tools - General AI chatbots: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Upstage Canonical: https://www.launchpadhq.io/platform/upstage-solar/guide Data: https://www.launchpadhq.io/data/v1/guides/upstage-solar.json Last reviewed: 2026-09-09 Tool: Upstage (Solar) (https://www.launchpadhq.io/platform/upstage-solar): starting price See pricing; free tier yes; data as of not dated Turns messy PDFs and scans into text that software can read — priced by the page, and built for people who write code. You sign in to Upstage's console in a browser and, by Upstage's own description, drop a PDF or scan into its playground and press run. What returns is the same document as structured text, with headings, tables and paragraphs marked up. Adjust an option, run it once more, then download what comes back. What it is: - Two products share one name here. Solar is Upstage's family of models — the AI "brains" that do the actual thinking — and it is not the part of Upstage that reads your documents. That work belongs to separate products called Document Parse, Document OCR and Information Extract, which take a PDF, a photographed form or a slide deck and hand back clean structured text. - Upstage is a Korean company, now one of Korea's larger AI firms, and it sells both halves the same way: as an API — a way for programs to talk to each other without a person clicking — charged per page for documents and per chunk of text for Solar. There is no consumer app and no monthly subscription. - It also publishes some of its models as files anyone can download and run on their own hardware, under a licence Upstage wrote itself. The document engines are not published that way; those you rent. - A layer called Upstage Studio sits on top, where you assemble the document steps into a workflow from a template rather than by writing code. Before you start: - Upstage is sold to developers, and you will get further with it if you have someone who can call an API. That is a fact about the product, not about you — if what you actually want is to ask questions of one PDF and read the answers in a browser, Adobe's Acrobat AI Assistant on this site does that, and does not involve an account with a spending balance. - Signup is self-serve — no sales call, no demo booking. Upstage's console sign-in page currently advertises free credit for new accounts; the pricing page doesn't repeat that offer and no expiry is published anywhere public, so treat it as current rather than guaranteed. Separately, the pricing page says every Studio agent includes ten free runs with no credit card required. - Beyond that you're buying credits — the platform's unit of spend, drawn down as you go. Document Parse is priced per page, and the enhanced setting costs about three times the standard one; plain text extraction is cheaper still, and pulling named fields out of a form costs more. The sign-up credit covers more ordinary pages than most people trying this out will get through. - The console, the documentation, the pricing page and the help centre are all in English and prices are quoted in US dollars. No regional restriction on signup is stated on any public page — though that is an absence of a rule rather than a published promise. Security: - Every document you send leaves your machine and is processed on Upstage's infrastructure, which its privacy policy says runs on US cloud providers. What happens to it next depends on which tier you are on — but check the date on the terms before you rely on any of this. The version that draws the line sharply does not bind anyone yet. Upstage's terms of service take a new version on 21 September 2026, and that version's Article 22 says that for paid services Upstage will not use your inputs or outputs for service improvement or model training without separate consent, while for free services the data may be used for service improvement and for AI research and development, training included; it also tells free-service users not to input personal data at all. What governs a reader today is the version before it, which Upstage dates 1 July to 20 September 2026, and we could not retrieve its text from Upstage's public pages to confirm whether it splits free from paid the same way. Assume the free tier may be trained on until you have read the version in force on the day you sign up. Security: Practical rules: - Work out which side of that line you are on before you upload anything sensitive. Genuinely free endpoints — the ones the pricing page marks free or beta — sit on the free side. Upstage's terms carve out "temporarily free promotional services" from the training clause, which is arguably where a sign-up credit lands, but the terms do not say so in those words and the ambiguity is yours to carry. - The Solar Chat demo is separate and stricter about saying so: the privacy policy states demo inputs are stored with personal information masked and may be used for AI model training. Do not treat it as a private scratchpad. - Article 22 also confirms the useful half — you keep ownership of what you put in and what comes out. Best at: - Turning documents software normally chokes on — scans, forms, slide decks, tables that run across a page break — into structured text with the layout preserved - Volume: Upstage's published figure is about 0.6 seconds a page, or a hundred pages in under a minute - Korean paperwork specifically — it reads .hwp and .hwpx, the Korean word-processor formats almost nothing outside Korea handles - Giving you an exit: the downloadable Solar Open models move the language half of the stack onto your own hardware Okay at: - Being explored without code — Studio and the playground exist and Upstage says they are for non-developers too, but everything around them is shaped for engineers - Everyday chatting — Solar Chat exists as a demo rather than a product, and this is not a chatbot company - Single documents — the account, the balance and the per-page pricing are built for thousands of pages, not one PDF Avoid: - Putting anything you'd mind being learned from through a free endpoint — the terms treat free-service data differently from paid (see Security) - Using what it returns to build or improve a competing document model — the terms prohibit that outright, including indirectly through a third party - Hardcoding a Solar version and walking away — Upstage has shipped Solar Pro 2, 3 and 4 inside roughly a year, and the older ones sit in the price table at different rates Starter: Run your worst scan through it first Document Parse ships against exactly these cases: Upstage's published fix list names 90 and 180 degree rotations, tilted scans, and tables that continue across a page break merged back into one. Prompt: Upload a crooked phone photo of a printed form or receipt with a stamp across it, then compare the table it returns against the original page. Starter: Parse the same page twice — standard, then enhanced The tripling from standard to enhanced is one Upstage publishes but never explains in words, so four cents of your own documents buys the answer cheaper than guessing wrong across ten thousand pages. Do it again on a plain page of body text. If the two outputs are identical there, you've learned the setting only matters for a subset of your documents. Prompt: Send a page thick with tables through the standard setting, then the same page through the enhanced setting, and put the two outputs side by side. Starter: Compare plain text extraction against full parsing Upstage prices these nearly seven times apart because one returns the words and the other returns the words plus where they sat. Most people over-buy here. Use a page whose meaning lives in its layout — an invoice, a lab result, a timetable. That is where the cheap option falls down. Prompt: Run one scanned page through Document OCR, then the same page through Document Parse, and read both outputs. Starter: Run a prebuilt agent from the Studio library The ten free runs per agent make this the only part of the platform you can judge before payment details exist anywhere. Studio bills as the sum of the steps an agent runs, so the same test shows what one real workflow costs per document. Note which steps the agent chained together. That list is the shape of the job you would otherwise be writing code to do. Prompt: Open Upstage Studio's agent library, choose a template matching your document type, and run it against ten of your own pages. Starter: Send the parsed text somewhere else and question it Document Parse's stated output formats are HTML and Markdown, chosen specifically so a language model can read them, and pasting the result somewhere neutral is the fastest test of that claim. Because the models and the parser are sold separately, it also tells you whether you need Upstage's models at all. Ask a question whose answer lives inside a table cell rather than a sentence. That is where a bad parse shows up. Prompt: Take the structured output from a parsed invoice, paste it into the chat AI you already use, and ask it for the total, the due date and the vendor's tax number. Pitfalls: - The prepaid balance is not a spending cap. Upstage's pricing page states that when your credits run out, usage is not interrupted — the charges go to your registered payment method instead. If you want a hard stop, the only reliable one is not registering a card. - A single setting can triple your bill. Standard and enhanced parsing live at the same endpoint, as do the two field-extraction tiers, and nothing warns you which one a script is using. - Every speed and accuracy number on Upstage's site is Upstage's own. The 0.6-seconds-a-page figure, the "5-10x faster than competitors" claim and the published table-recognition scores are vendor benchmarks, not independent tests. Your documents are the only benchmark that binds. - The terms were rewritten this quarter. The version effective 21 September 2026 names the one before it as running 1 July to 20 September 2026, so if you read Upstage's data terms earlier this year, you did not read these. - Two names, two documentation trees. The document engines are documented separately from Solar. Where to next: - Other tools that read documents for you: https://www.launchpadhq.io/discover?category=document-pdf-processing - More developer services like this one: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - Running open models on your own machine: https://www.launchpadhq.io/discover?category=local-open-source-ai ### Getting Started with v0 Canonical: https://www.launchpadhq.io/platform/v0/guide Data: https://www.launchpadhq.io/data/v1/guides/v0.json Last reviewed: 2026-09-09 Tool: v0 (https://www.launchpadhq.io/platform/v0): starting price Free, from $30/user/mo; free tier yes; data as of 2026-10-14 Describe an app in plain language; Vercel's v0 builds it, runs it, and puts it online. You type a description of the app or page you want. v0 works on it and shows a live, clickable version next to your message. You can click something in that preview and say what to change, or look at the code behind it. You keep replying until it's right, then publish it to a live web address. What it is: - v0 is Vercel's app builder: you describe a website or app in ordinary language and it writes the working code. - It began as a tool that produced single pieces of interface. Since August 2025 Vercel has positioned it as something that builds the whole thing — the pages people see and the machinery behind them. - What comes back is real code, not a picture of a design. You can send it to GitHub, where developers keep and track their code, or export it and carry on somewhere else. - Vercel's terms assign you its rights in that code, which is why v0 sits closer to the start of a real project than to a demo you throw away. Before you start: - v0 lives at v0.app; the older v0.dev address redirects there. Signing in uses a Vercel account. The free plan costs nothing, though no public page says whether a card is requested during sign-up — worth watching for at that step. - The free plan's shape is the thing to understand before you start: a small monthly allowance of credits (credits are the platform's unit of spend — each thing you make costs some) and a low daily cap on messages. A single small project usually takes more rounds of "no, change that" than one day allows, so plan on spreading work across days, or on paying. - How fast credits drain depends on which model (the AI "brain" doing the work) handled the message. v0 publishes four model tiers whose rates differ by more than twentyfold. v0's pricing docs say unused monthly credits roll over and expire after 65 days; when the balance reaches zero, generation stops. - Paid plans are priced per user per month; they raise the monthly credit allowance, add a small bonus of free credits on each day you log in, and open access to all models. v0's own FAQ also lists an older Premium plan it describes as being sunsetted — don't plan around it. - You don't need to know how to code to get something running. What you do need is patience for the moment something breaks and the fix isn't obvious — the code is real, and real code has real bugs. Security: - Vercel's AI Policy says user input, prompt content, queries and other interactions "may be used by Vercel... to improve our AI Products." Which way the default falls depends on your plan: the policy excludes Enterprise customers outright, and v0's pricing page lists "Training opt-out by default" as a Business-plan feature — which is a plain admission that the cheaper tiers are not opted out for you. The policy says you can change the setting yourself in Team Settings. Security: Check who can see a chat before you paste anything into it: - v0's sharing docs list four levels: private, everyone on your team, anyone with the link ("not be indexed by search engines"), and anyone on the web — which the docs say "may be indexed by search engines and appear in public galleries." - The docs don't say which level a new chat starts on, and that can't be confirmed without an account. Open the sharing control on your first chat and look, rather than assuming. - Inviting named individuals is described as a Plus, Business or Enterprise feature, so on the free plan sharing is a link, not a guest list. Security: - One trap specific to what v0 builds: anything you store in a setting whose name begins with NEXT_PUBLIC_ is sent to every visitor's browser. v0's security docs say it analyses your project and warns you about this, but the warning is a safety net rather than the rule — keep passwords and any key that spends money out of those. Best at: - Getting a working, clickable version of an idea in front of people fast — you send a link instead of describing what you mean. - The common furniture of the web: dashboards, landing pages, forms, sign-up flows, admin tables. - Starting from a picture. v0's docs describe generating interfaces from wireframes or mockups you upload, and its July 2026 changelog adds reading a Figma file directly. - Handing work to a developer later, rather than dead-ending in the chat. Okay at: - Sustained work on one project. It will keep going; the free plan's daily message cap is what limits the pace. - Exact visual design. It produces a competent, recognisable house style, and pulling it onto an exact brand takes real fighting. - The machinery behind the pages. The docs describe connecting data and services, but it is the part where someone who doesn't code is most likely to get stuck when it misbehaves. Avoid: - Assuming what you get is yours alone. Vercel's AI Product Terms warn that the output "may not be unique" and that other users may receive "the same or similar Output." - Putting generated code in front of real users unreviewed. Vercel's own policy says code output "should be paired with human review, where applicable, to ensure suitability for any commercial use." - Learning to code from it. It writes the code for you; watching that happen is not the same as being able to repair it. Starter: Get a link on day one You end up with a live address you can send to someone, not a file: v0 is made by the company that runs the hosting, so putting the result online is a control in the chat header rather than an export-and-upload errand — and v0's pricing page lists deploying to Vercel on the free plan. Swap in your own business, but keep the last sentence — it is how you find the publish control without hunting for it. Prompt: Build a one-page site for a dog-walking business called Trot. Include a short intro, three service cards with prices, a strip of photo placeholders, and a contact form that shows a thank-you message when it is submitted. Make it work properly on a phone. When it's done, tell me exactly what to click to put it online. Starter: Fix it by pointing, not describing v0's quickstart tells you to open Design mode and select elements in the preview to tweak their styles visually or in words, changing one thing without rebuilding the whole page — and v0's pricing page lists Design mode on the free plan, where spending one of the day's messages to nudge a margin is a bad trade. Name the section and the exact change. Vague instructions like "make it nicer" are what burns a message. Prompt: The pricing section feels cramped and the middle card should stand out. Increase the space between the cards, make the middle one slightly wider with a subtle border, and change the section heading to Simple pricing. Leave everything else exactly as it is. Starter: Make it tell you the plan first You get a short written plan to correct before any code is generated. Every generation draws on the same daily message cap and monthly credit balance described above, so with v0 the cheapest fix is the build that never happens — and a wrong build on an expensive model costs several right ones on a cheap one. Ask for five bullets instead of three for anything with more than one screen. Prompt: Before you build anything, tell me in three bullets what you are going to make and what you are assuming about it. Then stop and wait for me to say go. Starter: Hand it over without a rewrite This moves the code out of the chat into the place developers work, and v0's FAQ describes a bi-directional GitHub integration — changes made locally sync back into v0 — so the handoff is not a one-way export that strands the conversation that produced it. No GitHub account? Ask it to export the code instead, and to say where the download lands and what to do with it. Prompt: Connect this project to GitHub. Then summarise for me in plain language what files you created, what each one is responsible for, and which one I would open first if I wanted to change the wording on the homepage. Starter: Ask it to break its own work Vercel's launch post says v0 can search the web, read files and inspect a live site, so "go and look at the page you just made" is a described capability rather than the model recalling what it intended to build. Swap in a specific visitor: someone using a screen reader, someone on a bad connection, someone who has never bought anything online. Prompt: Act as a first-time visitor using a slow phone. Walk through this page and list every place it could confuse someone, break, or fail silently. Then fix only the top three problems and tell me plainly what you changed. Pitfalls: - The free plan's daily message cap counts your corrections, not your projects — so write longer, more complete messages here than you would to a chatbot. - Messages and credits are two separate caps: running out of the day's messages does not release your remaining credits, and having credits left does not buy you another message. - Publishing is not testing. Getting a production address takes seconds, which makes it easy to send someone a link to something you have looked at but never actually used. - Changes ripple. Because a message rebuilds rather than edits a single line, say what should stay exactly as it is, not only what should change. - v0's pricing page lists access to all models as a paid feature, so if the model choice looks thin on the free plan, that is the plan rather than a fault. Where to next: - Other AI coding tools: https://www.launchpadhq.io/discover?category=ai-coding-development - Chat assistants: https://www.launchpadhq.io/discover?category=text-conversational-ai - Developer APIs and services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services ### Getting Started with VEED Canonical: https://www.launchpadhq.io/platform/veed/guide Data: https://www.launchpadhq.io/data/v1/guides/veed.json Last reviewed: 2026-10-15 Tool: VEED (https://www.launchpadhq.io/platform/veed): starting price $25/mo monthly or $147/yr; free tier yes; data as of 2026-10-14 A video editor that runs in your web browser, where you can cut a talking video by deleting words from its transcript. You open VEED in your web browser, upload a video or recording, and work on it in an editor: your clips sit on a timeline (the strip showing your footage in order) and the tools sit in a menu down the left. Click a tool, watch the preview change, adjust it, then press Done and download the finished video. What it is: - VEED is an online video editor made by Veed Limited, a London company. Everything happens in a browser tab: you upload footage, trim it, add subtitles and music, and download an MP4 at the end, without installing editing software. - Its strength is talking-to-camera video — explainers, tutorials, social clips, recorded presentations. VEED turns your speech into a transcript, and several of its tools work from that transcript: delete a sentence from the text and the matching footage disappears from the video. - Since August 2026 the first screen you see after signing in is VEED AI, a box where you type a topic and VEED drafts a script and then a video from it. That is one way in, not the only one; most of this page is about editing footage you already have. Before you start: - You need a web browser and an account; VEED's help centre starts every tutorial with "go to veed.io and log in or create an account." After signing in you land on the VEED AI prompt box (where you type the message — your prompt) — to work on your own video, click Edit Video on the dashboard and upload a file. - The AI features that generate new material — VEED AI videos, AI Playground clips, subtitle translation — spend credits (the platform's unit of spend — each thing you make costs some). VEED shows the credit cost before you commit, doesn't charge for a generation that fails, and on monthly plans any credits you haven't used are gone at the end of the month. The everyday editing tools on this page — cutting, subtitling in your own language, removing silences — are not described by VEED as costing credits. - A few tools only understand certain languages. Magic Cut, which trims pauses and filler words automatically, supports English only, and editing by transcript works only in languages that put spaces between words — so not Chinese, Japanese or Thai, for example. Security: - VEED's privacy policy, last updated 6 January 2025, says that personal data in content uploaded on its free plan may be used for 'training and developing' its artificial intelligence models. You can object, but the policy says that if you do, you cannot keep using the free plan — you would have to upgrade or delete your account. The policy gives no settings switch for this; it directs you to contact VEED's data privacy manager using the details in the policy itself. Security: Also worth knowing before you upload:: - VEED Limited is a UK company, and its privacy policy says many of the outside companies it relies on are based outside the UK and Europe. It also says it does not let those companies use your personal data for their own purposes. - VEED's terms for its AI tools say you own the AI images you make, but give VEED the right to host them and to use them 'to market and improve our products and services.' - The same terms ban using VEED's AI presenters or text-to-speech voices to impersonate any person, and ban AI content meant to influence elections or campaigns. Filming yourself is fine; making a digital someone-else is not. Security: - None of this makes ordinary editing risky. Just treat footage of other people, and anything filmed at work, as something you are handing to a company that may learn from it. Best at: - Cleaning up a recording of someone talking — cutting stumbles, pauses and 'um's by editing the words rather than scrubbing through footage. - Adding subtitles quickly, then styling them and burning them into the video for social posts that play on mute. - Rescuing audio recorded in a noisy room, one clip at a time. - Editing on a computer that can't run desktop editing software, since everything runs in the browser. Okay at: - Fine, frame-by-frame editing of music videos or action footage. There is a timeline and it works, but the tools that make VEED quick all assume someone is speaking. - Generating video from scratch. VEED AI and AI Playground will do it, but the result is a draft you will usually still edit. Avoid: - Reshuffling a talk by dragging sentences around in the transcript — VEED's script editor can delete and restore words but not reorder them; you move footage on the timeline instead. - Using AI presenters or synthetic voices for anything political, religious or about other sensitive topics; VEED's AI-tool terms rule those out for avatars and text-to-speech. Starter: Type a topic, read the script before you spend anything VEED AI drafts the full script and shows it to you before any video is made — you can edit any line or regenerate the whole script — and the estimated credit cost sits on the Generate button. So the first attempt costs you nothing but reading: fix the words first, then pay for pictures. Paste a script you've already written instead of a topic. VEED's announcement also says a video with a voice but no on-screen character uses no credits, so try that version first if you're watching your balance. Prompt: A 45-second vertical video for Instagram explaining three ways to keep houseplants alive through winter, friendly tone, for complete beginners. Starter: Auto-subtitle a clip and fix the orange words Upload a clip of you talking, open the Subtitles tab on the left, pick the spoken language and click Auto-subtitle. Read through the result, correct anything it misheard, choose a style, then click Done and switch on Burn Subtitles so the words are part of the video. VEED colours the words it is least sure about in orange, so you check those instead of proofreading every line. Its help centre also warns that if you cut the video after subtitling, the timings drift and you need to regenerate them — so subtitle last, not first. Add Translation in the same panel to subtitle into another language — VEED says translation spends credits. Starter: Edit the video by editing its words With a clip on the timeline, click Video in the left panel and then Edit with Script. Highlight a sentence you fluffed and press Delete: the words go strikethrough and that footage is gone. Highlight it again and click Restore to bring it back. The transcript also shows pauses as entries like '(( 0.5 s ))' that you can delete one by one, or all at once with Delete Silences, plus a Filler Word Removal option — but it only works from subtitles VEED generated itself, not ones you typed or uploaded, which is another reason to run Auto-subtitle first. For a one-click version, select the clip and use Magic Cut under the Video menu. It removes pauses, filler words and mistakes for you, then shows the cuts for you to adjust — English only. Starter: Clean up the sound on a noisy clip Select a clip, open the AI Tools tab on the left and click Clean Audio. VEED strips hum, wind, traffic and echo, evens out the volume, and gives you a preview to compare. Clean Audio works on the clip you selected and nothing else — VEED's help centre says a project with several clips needs it run on each one separately. Miss one and that section will sound different from the rest. Starter: Generate one short clip in AI Playground In AI Playground you pick which video model to use, and VEED says each one gives a different look at a different credit cost, shown before you click Generate. Run the same description through two models and compare cost against result before settling on one. Choose the shape (vertical for phones, wide for YouTube) before generating — AI Playground asks for the aspect ratio as a separate step. Prompt: A slow push-in on a steaming cup of coffee on a wooden table beside a rainy window, soft morning light, calm and cosy. Pitfalls: - Changes after you download don't follow the file. Any edit you make once you've exported needs a fresh export — and VEED renders it again, which takes time on longer projects. - Expecting VEED to sharpen your footage. Its export settings let you pick a resolution, but VEED's help centre says it won't improve footage beyond its original quality. Blurry phone video exported at a higher setting is still blurry phone video. - Cancelling at the last minute. VEED's help centre says to turn off auto-renewal at least two business days before your renewal date, that you have to finish a short questionnaire for the cancellation to go through, and that cancelling does not refund what you've already paid. Look for the confirmation email; if none arrives, contact support. Where to next: - Video Creation & Editing: https://www.launchpadhq.io/discover?category=video-creation-editing - Voice & Speech: https://www.launchpadhq.io/discover?category=voice-speech - Text & Conversational AI (for writing your script first): https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Vertex AI (Now Gemini Enterprise Agent Platform) Canonical: https://www.launchpadhq.io/platform/google-vertex-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/google-vertex-ai.json Last reviewed: 2026-09-08 Tool: Google Vertex AI (https://www.launchpadhq.io/platform/google-vertex-ai): starting price See pricing; free tier yes; data as of not dated Google Cloud's platform for building with AI rather than chatting with it — renamed in 2026, and free to try for 90 days without a card. You sign in with your Google account, pick a model (the AI "brain" that does the actual thinking) from a list in your browser, type an instruction, and read the reply in the same window. You change a setting, try again, and when you like the result you copy code that calls that model from your own program. What it is: - Vertex AI is where Google Cloud lets you build with AI instead of only talking to it: a workspace in your browser, plus the plumbing to call the same models from your own software. Google renamed it in April 2026 to Gemini Enterprise Agent Platform — "agent" meaning AI that takes actions on its own rather than only answering — and both names are still in circulation. - It is built for people making software or wiring AI into a company's systems. If you mainly want to ask an AI questions, the Gemini app or Google AI Studio gets you there in a minute. Before you start: - The name changed and the internet has not caught up. Google announced on 22 April 2026 that Vertex AI's services now ship as Gemini Enterprise Agent Platform, and the old documentation carries a banner saying it is no longer updated. Guides you find elsewhere — and some of Google's own pages — still say Vertex AI, and usually still describe the right thing. - There is a free way in that most write-ups skip: express mode signs you up with nothing but a Google account, and Google's FAQ states that no payment information is required for 90 days, within set usage limits. A window, not an ongoing free plan. - The better-known offer wants a card: a Welcome credit across a 90-day trial, spendable on all of Google Cloud, so anything else you leave running draws down the same pot. After both windows close it is metered by the amount of text going in and coming back, with no permanent free allowance. Security: - The reason a company picks this over the free consumer tools is in the documentation: Google's data governance page says it "won't use your data to train ... models without your prior permission or instruction." Google's own free developer tier says the opposite — there, content is "used to improve our products." Security: Two things to know before you send anything real:: - An API key (a password that identifies your app, and that spends your money) is the one secret here. Anyone holding it can bill your project. - Letting the model check Google Search before it answers is the exception to that promise: Google keeps what you sent and what came back for 30 days, with no way to turn it off. Best at: - Reaching many companies' models from one account — Model Garden advertises 200+, from Google's own Gemini to Meta's Llama and Anthropic's Claude - Letting a model check Google Search before it answers, with a monthly allowance of checks at no charge - Keeping company data out of model training, stated in writing rather than implied Okay at: - Being somewhere to simply talk to Gemini — it works, but you set up a cloud project to get there - Guiding a newcomer — the documentation is written for engineers, and it is currently mid-rename Avoid: - Parking anything you want to keep in an express mode project — Google's own FAQ says that without billing, access ends after 90 days and "30 days after that, your data will be marked for deletion" - Trusting any one Google page on naming — the data governance page still says "Vertex AI", and a tutorial titled "Agent Studio in express mode" sits at a URL containing vertex-ai-studio - Assuming what you learned on Google AI Studio carries over — separate product, different data terms Starter: Sign up in express mode before you set up a billing account Signs in with an ordinary Google account and chooses express mode, which skips the billing setup for 90 days. Google's FAQ states that no payment information is required to try express mode for 90 days. Amazon Bedrock, the closest equivalent, has no such door — an AWS account with a payment method comes first. It also leaves the trial credit unspent until you need it. Write down the date. Express projects stop at 90 days; deletion follows 30 days later. Starter: Ask about something that happened this week, with Search checking behind it Turns on the setting that lets the model check Google Search, then asks it about an event from the past few days and reads the sources it returns. Google is selling the one asset its rivals cannot license: the pricing page currently includes a monthly allowance of such checks at no charge, then bills for the rest. Azure AI Foundry can check Bing, not Google Search. Ask again with the check switched off. The gap between the answers is what you pay for. Starter: Send one instruction to a Google model and to a non-Google one Opens Model Garden, picks a Gemini model and an Anthropic or Meta one, and sends each the same instruction from the same project. Model Garden puts Anthropic's and Meta's models under the same project, terms and invoice as Google's own, so the comparison costs no second signup and no second vendor relationship. Google AI Studio, the free door most beginners find first, serves Google's models only. Use something with a right answer — differences show up most clearly when one model is wrong. Pitfalls: - Enabling billing without setting a budget alert. Charges land on the whole project's Google Cloud bill, mixed in with storage and computing, not on a standalone AI invoice with a prepaid balance. - Expecting express mode to cover everything. Some services are capped to small usage limits there, and some need a billing account before you can deploy. Where to next: - AI APIs & Developer Services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - AI Coding & Development: https://www.launchpadhq.io/discover?category=ai-coding-development - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Viz.ai Canonical: https://www.launchpadhq.io/platform/viz-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/viz-ai.json Last reviewed: 2026-09-12 Tool: Viz.ai (https://www.launchpadhq.io/platform/viz-ai): starting price Custom; free tier no; data as of not dated Hospital software that watches scans as they are taken and alerts the specialist who can act — and that no individual can buy. There is no app you open. A hospital scanner sends an image to Viz.ai automatically, software checks it for a suspected finding, and if it sees one a phone buzzes in a specialist's pocket with a case to look at. The user is a clinician receiving an alert, not someone typing. What it is: - Viz.ai makes software hospitals install between their scanners and their specialists' phones. When a scan is taken it checks for a short list of urgent findings — a blocked major artery in the brain, bleeding in the brain, a clot in the lungs — and if it suspects one, alerts the doctor who would treat it, at the same time as the radiologist does the normal read. - It is here as one of the clearest examples of AI regulated as a medical device: each piece cleared by the US Food and Drug Administration for one stated purpose, and required to publish what that purpose is. Viz.ai's homepage calls the platform "driven by over 50 advanced, FDA-cleared algorithms" and says it is trusted by over 2,000 hospitals; a July 2026 announcement puts that at 2,000 hospitals and 230 million patients across the US and Europe — Viz.ai's figures about itself, published without a method. Before you start: - You are at neither end of this. Viz.ai's end user licence agreement, last updated March 2026, says the app is "made available to you as an authorized user under an existing agreement with a hospital or health care system" and that you must "only use the App and Services if you are a medical specialist with the requisite clinical expertise." That binding document is the answer to "can I try this": no, and not because of a waiting list. - There is no price to compare. Viz.ai publishes no pricing page — viz.ai/pricing does not exist — and offers no trial and no self-serve sign-up; the only routes in are a demo request form and three phone numbers reaching a sales team. Any confident "Viz.ai pricing" figure online comes from a third-party directory, not the company. - The app being free in the app stores is a red herring: it downloads for nothing and the store listing carries no access restriction at all, but where a store page and the licence quoted above disagree, the licence binds. - Clearances are granted country by country. The modules on the indications-for-use page are described as FDA-cleared for the US, and in June 2021 the company announced a CE Mark for its stroke software, permitting sale within the European Economic Area. Which modules are cleared where today is not published module by module, so treat any "it's available in X" claim — including one you read here — as something to check. - What you can do today is read: the indications-for-use page, the licence agreement, the privacy policy and the trust centre are public and un-gated. If you came for an AI tool you can open this afternoon, start with a general chatbot like ChatGPT. Security: - The sensitive information here is not yours: it belongs to the patients being scanned. Viz.ai's US privacy policy for subscription services, effective 20 August 2025, puts the medical data outside that policy — "Protected Health Information is governed by the HIPAA Business Associate Agreement between the Customer and Viz.ai and not by this Privacy Policy." The hospital–Viz.ai contract, which is not public, governs patient scans. Security: Three things in the vendor's own documents worth knowing: - The same policy says Viz.ai may commercialise the data it treats as non-medical: "We may also share and sell this Usage Data to interested third parties to help them improve their medical programs." What falls inside "Usage Data" is defined elsewhere in the policy. - Its trust centre lists a SOC 2 Type II audit (an audit of how a company handles customer data), a European privacy programme, and — announced in May 2026 — certification to ISO/IEC 42001, an international standard for managing AI. All are the company's statements about itself; the audit reports go to customers, not the public. - Nothing published says whether patient scans are used to improve Viz.ai's own software. The privacy policy routes that question to the hospital contract, which is not published — so it is genuinely unanswered from outside, not answered quietly in the vendor's favour. Best at: - Getting a suspected finding to the specialist who can act on it, in parallel with the normal radiology read rather than after it. - A short list of urgent findings: the indications-for-use page names blocked major arteries in the brain, bleeding in and around the brain, clots in the lungs, tears in the aorta, and some heart conditions. - Automated measurements the company says are consistent: for brain bleeds, volume, thickness and how far the brain has shifted. - One alerting layer across neurology, cardiology, vascular, trauma and lungs, not a tool per department. Okay at: - Breadth. Each module looks for one thing, so a scan normal for that finding produces no alert. It is a watchlist, not a second opinion on the scan. - Its newest layer. Viz Assist, launched 20 October 2025, is described by the company as "a suite of autonomous AI agents" (an agent is AI that takes actions on its own rather than only answering) drafting notes from recorded doctor–patient conversations and record data. - Telling you anything about itself in use. Everything past the marketing site sits behind a hospital deployment, so nothing here comes from using the product — only from documents the company publishes. Avoid: - Expecting it to answer a question. The indications describe an automatic, notification-only path. Nothing published describes a clinician asking it anything, and there is no box to type into. - Taking "detects" to mean "diagnoses". Viz LVO's own indication says it is "limited to analysis of imaging data and should not be used in-lieu of full patient evaluation or relied upon to make or confirm diagnosis" — the company's wording. Starter: Read the page Viz.ai is required to publish Opens viz.ai/indications-for-use, where Viz.ai lists module by module what each piece of its software is cleared to do. The wording is specific enough to check a claim against. Viz ANEURYSM "is limited to detecting aneurysms at least 4mm in diameter"; Viz LVO "is intended to analyze terminal ICA and MCA-M1 vessels" — two named blood vessels, not the whole brain. Those are Viz.ai's numbers, not the category's; RapidAI clears its own scope separately. Read one module's entry twice: once for what the software finds, once for what it is not for. Starter: Find the sentence where the vendor says it does not diagnose Opens the end user licence agreement at viz.ai/eula and finds the passage the company sets in capitals. Viz.ai puts it in capitals in a binding document — "THE APP AND THE SERVICES DO NOT DIAGNOSE CONDITIONS and are intended only as a supplement to your knowledge and professional judgment" — while its product pages lead with detection performance. Both are true, written for different readers under different rules. Starter: Separate the cleared parts from the new parts Compares the cleared imaging modules on the indications-for-use page against Viz Assist. Viz.ai's launch announcement for Viz Assist calls its imaging algorithms FDA-cleared and says nothing about a clearance for Viz Assist itself; the product page carries no regulatory statement either. An unstated status, not a denied one — two regulatory positions side by side under one brand. Starter: Work out what the phone is actually allowed to be Reads the mobile-app clause that recurs across several of Viz.ai's indications. It says images previewed on the phone "are compressed and are for informational purposes only and not intended for diagnostic use beyond notification," and that "notified clinicians are responsible for viewing non-compressed images on a diagnostic viewer." The phone's job: wake someone up, not become the screen they decide on. Viz HCM, which reads heart tracings rather than scans, is worded differently and aimed at cardiologists reviewing patients aged 18 and over. Starter: Test the vendor's own head-to-head claim Traces the comparison line on Viz.ai's Viz LVO page back to the publication it cites. The page states "Viz performed better than Rapid with a higher specificity and PPV" — two measures of how often a positive flag turns out to be right — and names its source (listed as Delora et al., 28 July 2023). The citation being there means the claim can be followed rather than taken on trust, though it is still the vendor's choice of result. Pitfalls: - Reading a suite name as a product. "Viz Neuro Suite" and "Viz Cardio Suite" are website groupings; clearances attach to the modules inside — Viz LVO, Viz ICH, Viz CTP, Viz SDH, Viz ANEURYSM and the rest — each worded separately with its own limits. - Mistaking the training app for the product. A separate free "Viz Training" app exists in the app stores, labelled "not a medical device" and "intended for informational purposes only," its listing showing a last update in February 2020. - Taking the homepage's "over 50 advanced, FDA-cleared algorithms" as covering everything on the site. It counts cleared algorithms; it does not say every feature or assistant carrying the Viz name is one, and the indications-for-use page names far fewer than fifty. Where to next: - Healthcare AI: https://www.launchpadhq.io/discover?category=healthcare-ai - Research & academic tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Chat assistants you can open today: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with vLLM Canonical: https://www.launchpadhq.io/platform/vllm/guide Data: https://www.launchpadhq.io/data/v1/guides/vllm.json Last reviewed: 2026-09-13 Tool: vLLM (https://www.launchpadhq.io/platform/vllm): starting price Free; free tier yes; data as of not dated The engine that serves one open AI model to a lot of people at once — free software that expects a serious graphics card and a Linux machine. On a Linux machine with a graphics card, you run one command naming the model (the AI "brain" that does the thinking) you want. It downloads, loads, and keeps running as a service. Nothing opens on screen — instead other programs, or a chat app you point at it, send questions and get answers back. What it is: - vLLM is the engine underneath a lot of AI products you have already used: the part that takes a question, runs it through a published AI model, and hands back an answer — for many people at the same time, on one machine, without falling over. It is built for the person putting a model in front of other users: a team standing up its own internal AI service, or an organisation that wants requests staying on hardware it controls. If you have heard of Ollama or LM Studio, those are the laptop-scale relatives of the same idea — one person, one conversation, your own computer. vLLM is that idea scaled up and pointed outwards. Before you start: - The graphics card decides whether any of this is possible. vLLM currently needs an NVIDIA card of compute capability 7.5 or newer — the docs name the T4, RTX 20-series, A100, L4, H100 and B200 as examples — or an AMD card on ROCm (MI200/MI300/MI350, Radeon RX 7900 and 9000 series, Ryzen AI MAX and AI 300), or an Intel Data Center or Arc GPU. There is no published minimum in gigabytes, because the real rule is that the model has to fit on the card with room to work; vLLM claims about 92% of the card's memory for itself on startup, so a run either fits or stops immediately. For scale, the project's own worked examples serve models of around half a billion to one-and-a-half billion parameters (a rough measure of a model's size — more is smarter but heavier), not the famous large ones. - It runs on Linux, and that is not a soft preference. The docs state plainly that vLLM does not support Windows natively; on Windows you run it inside WSL, Microsoft's built-in Linux layer. On a Mac there are two paths and both come with an asterisk: vLLM-Metal, which the docs describe as a community-maintained plugin, uses Apple's graphics hardware but needs its own separate set of model files; and a CPU-only Mac build that the docs call experimental and that you have to compile yourself. If you are on a MacBook and what you actually want is to try running an AI model privately on your own machine, Ollama is the tool built for that and will get you there this evening — come back to vLLM when the question is serving other people. - The software is free; the machine it needs usually is not. vLLM is free to download and use with no account, no card and no usage limits, and its code is public under a permissive licence that allows commercial use. What costs money is the hardware. Most people meet that by renting a GPU machine by the hour — the rate depends heavily on the card — which is billed for as long as the machine is switched on, whether or not anyone is asking it questions. If you already have a suitable card sitting in a desktop, your cost really is zero. - The model's licence is a separate question from vLLM's. vLLM being permissively licensed says nothing about what you may do with what comes out of it — that is governed by whichever model you load. Some are equally permissive: the model the quickstart uses, Qwen2.5-1.5B-Instruct, is Apache-2.0. Others are not: Meta's Llama models carry a custom Llama Community License and the download is gated, so you have to accept the agreement and share contact details before you can fetch the files at all. Check the model's page before you build anything on it. Getting set up safely (Install from the vLLM project's own installation guide at docs.vllm.ai, and nowhere else. It resolves to exactly two official things: the `vllm` package on PyPI, and the official container image `vllm/vllm-openai` on Docker Hub. One trap worth naming: those same official docs link a community-maintained Windows fork, in the same paragraph where they say Windows is not natively supported. Being linked from the official docs does not make a fork official.): - Check the card before you install anything, because this is the step that decides the whole evening. Confirm the GPU is on the supported list first; the project's installation guide covers the rest. - Know where the model files pile up. vLLM downloads models from Hugging Face on first run. These are large files and they accumulate silently; if you are on a rented machine, they also disappear with it and download again next time, which is worth planning around rather than discovering. - Confirm it worked before you build anything on top. Start a small model and ask the running server to list what it is hosting — getting the model's name back is how you know the install landed. Security: - Almost everything else in this corner of the catalog is something you run for yourself. vLLM is something you run for other people, and that changes the shape of the risk entirely. Nothing you send it leaves your hardware — the model sits on your machine, the questions go to your machine, there is no company on the other end — and that part is genuinely private. But a service built to be reachable by others ships with no password on the door: authentication is something you switch on, not something you switch off. That is the single most important thing to understand before you start it on a machine with a public address. Security: What that means in practice:: - There is no login by default. You get one by passing --api-key when you start the server — and vLLM's own security page is unusually blunt that this is not enough on its own: other addresses on the same server run exactly the same model and stay open with no key at all. The project's own instruction is not "set a key" but "put a proxy in front of it that only lets through the addresses you meant to expose." - Do not assume it is only listening to you. The quickstart says the server starts at localhost:8000, while the security page warns that vLLM and the libraries beneath it may open services on every network interface, sometimes outside vLLM's own control — so treat the address it is listening on as something to check and firewall, not something to assume. An inference server reachable from the open internet is not a theoretical problem: anyone who finds it can use your hardware, and can call the addresses that pause it or change what model it is running. - --trust-remote-code runs code the model's author wrote. Some models on Hugging Face need that flag before vLLM will load them. It does what it says — it executes their Python on your machine. Use it for models from sources you would trust with a software install, and not otherwise. - If you ever spread it across more than one machine, the traffic between them is unprotected. The docs state that all communication between nodes is insecure by default and must be placed on an isolated network — a documented design choice, not an oversight. Best at: - Serving one model to a lot of people at once — the memory handling and request batching that make that work are the whole reason the project exists - Getting the most answers per hour out of a graphics card you are already paying for - Standing in for a paid AI service in code you have already written, because it speaks the same request format the big providers use — often a change of address rather than a rewrite - Keeping prompts, documents and customer data on hardware you control, with no third party in the path Okay at: - Modest hardware. It does run on ordinary processors without a graphics card, and on Apple Silicon through a community plugin, but the docs describe those as basic and experimental respectively - One person having one conversation. Nothing breaks — but nothing about the design pays off either Avoid: - Judging it by a run on a MacBook — whatever you measure on the community Mac plugin, it is not the thing the benchmarks are describing - Reaching for it as your first attempt at running an AI model yourself — Ollama answers that question today, and this is still here afterwards - Chasing a model announced this week. Support often lands in the daily development builds before it lands in a release, and running a development build to get one model is a different commitment from installing a stable one Starter: Start with a model small enough that the card is not the question Serves one of the small models the project's own documentation demonstrates with — around half a billion to one-and-a-half billion parameters — so that your first run is testing your installation rather than your hardware. An oversized model here fails at launch instead of just running slowly. On Ollama an oversized model grinds along on a mix of graphics memory and ordinary memory, so "too big" feels like slowness; here it is a stopped server. Starter: Decide which single model this server is for Names one model at launch. That server hosts that model until you stop it; wanting a second model means a second server on a second port. The docs state it directly — the server hosts one model at a time — so on vLLM "which model" is a deployment decision you make once, not a per-request choice. Ollama's server loads whichever installed model each individual request names and unloads it again when idle. If you find yourself wanting to switch models often, that is a signal you want the laptop-scale tool, not this one. Starter: Send it many questions at once, because that is the actual product Fires a batch of requests simultaneously rather than one after another, and compares total answers per minute against sending them in sequence. The engine folds new requests into work already running on the card instead of queueing them behind it; its docs describe requests being set aside and recomputed when the working memory fills. Ollama has no equivalent; it is designed around one person at a keyboard. Starter: Rent the graphics card for an evening before you buy one Spins up a rented Linux machine with a supported card, installs from the official docs or runs the official container image, serves one model, and shuts the machine down again. The project publishes a ready-made container image because running it on hardware you do not sit in front of is the normal case, not the exotic one. Renting a machine to run Ollama would defeat Ollama's entire proposition, which is your own computer. Set yourself an alarm. Pitfalls: - Gated models fail with an unhelpful error. If you have not accepted Meta's licence agreement on Hugging Face, the serve command simply fails to fetch the model — and the reason is on a web page, not in the error. - The first start is slow, and it is not broken. vLLM compiles and prepares work for your specific card before it will answer anything, which makes the first boot noticeably longer than later ones. The project caches that work so repeat starts are faster. Waiting is the correct response; restarting it is not. Where to next: - Local & Open-Source AI: https://www.launchpadhq.io/discover?category=local-open-source-ai - AI APIs & Developer Services: https://www.launchpadhq.io/discover?category=ai-apis-developer-services - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Wispr Flow Canonical: https://www.launchpadhq.io/platform/wispr-flow/guide Data: https://www.launchpadhq.io/data/v1/guides/wispr-flow.json Last reviewed: 2026-10-23 Tool: Wispr Flow (https://www.launchpadhq.io/platform/wispr-flow): starting price Free, from $15/user/mo; free tier yes; data as of 2026-10-14 Hold a key, talk, let go — Wispr Flow types what you said, cleaned up, into whatever app your cursor is in. You install Flow, then hold a shortcut key and speak. When you let go, your audio goes to Wispr's servers, is turned into text, tidied up — filler words removed, punctuation added — and pasted wherever your cursor is: an email, a chat box, a document. It needs an internet connection; transcription happens in the cloud. What it is: - Wispr Flow is dictation that works across your whole computer or phone rather than inside one app. Its help centre describes it as working in any app with a text box. You hold a key, speak naturally, and release; the cleaned-up text appears where you were typing. - The cleanup is the point. Flow's Smart Formatting, on by default, removes filler words and fixes grammar, punctuation and capitalisation. It also understands spoken corrections — say "coffee at 2, actually 3" and Wispr's docs say you get "coffee at 3". Saying "first… second…" turns into a list. It won't fix a word it misheard, which is what the personal dictionary is for. - Beyond plain dictation there are snippets (say a short trigger phrase and Flow inserts saved text, such as your email signature), styles that make your writing more formal or casual depending on the kind of app, and Command Mode, which edits selected text by voice — for example "make this shorter". Wispr's docs say Command Mode needs a Pro plan or an active trial. - Flow runs on Mac (macOS 12 or later), Windows 10 and 11 (not Windows on ARM), iPhone (iOS 18.3 or later) and Android 13 or newer. iPad, Linux and Chromebooks are not supported. Wispr says it handles more than 100 languages, one language per dictation. Before you start: - The free plan is a weekly word allowance. Wispr's plans page lists a free plan with a weekly word cap on Mac and Windows. Phones get a smaller allowance, and the docs say individual accounts may differ. - Pro removes the limit. Pro is priced per user, costs less per month if you pay for the year, and Wispr's pricing page shows prices in your own currency; current prices are on its Launchpad page. Wispr's docs say new accounts get a 14-day Pro trial, and its pricing page offers students and educators with an education email a discount on Pro. - Check your setup is supported. Remote desktops, virtual machines and Windows on ARM laptops are not supported, and on a Mac you will need to grant both Microphone and Accessibility permissions so Flow can paste text into other apps. Getting set up safely (Download only from wisprflow.ai, or from the Apple App Store or Google Play. Wispr's help centre warns about fake Android apps impersonating Wispr Flow — on Android, install only the listing from Google Play.): - Expect some powerful permissions. On a Mac, Flow asks for Microphone and Accessibility; on Android it needs Accessibility (to insert text) and permission to display over other apps; on iPhone it works as a keyboard with "Allow Full Access" turned on. These are how it types into other apps, so only grant them to the genuine app. - Run the built-in practice. Setup includes a microphone test, your shortcut and your languages. Do the mic test with the microphone you will really use — Wispr's docs say Bluetooth and AirPods mics often perform worse than a built-in or wired one. - Confirm it works. Open any text box, hold the shortcut (Fn on a Mac, Ctrl+Win on Windows by default), say a sentence and let go. Text appearing where your cursor was means you're set up. Security: - Everything you dictate is sent to Wispr's servers to be transcribed — there is no on-device mode. That matters because Flow is designed to be used everywhere, including in emails and messages you would not normally share. Security: Defaults to review in Settings → Data and Privacy:: - Model training is on by default on Free and Pro. Wispr's security FAQ says the "Improve the model for everyone" setting can be turned off; it was previously called Privacy Mode. Even with it off, some metadata (such as which app you were in and word counts) is still sent. - Cloud storage is on by default. A separate setting stores your transcripts, audio and history on Wispr's servers. Turn it off if you don't want a history kept there. - Context awareness is on by default. To format text to suit the app, Flow reads the app name and website, nearby on-screen text and the text around your cursor. Wispr says it doesn't take screenshots and skips banking and financial apps and standard password fields — but custom or web-based password fields may still be read like normal text. On iPhone there is no switch for it. Security: - Wispr says it holds SOC 2 Type II and ISO 27001 certifications for dictation and offers a click-through HIPAA agreement in settings, but its own docs suggest confirming with support before using it for health information. Until you have, don't dictate patient, client or financial details. Best at: - Writing emails, messages and first drafts several times faster than typing, in any app. - Talking the way you normally do — ums, restarts and "actually" corrections — and still getting clean text. - Repeated text, via snippets, and names or jargon it learns through the dictionary. Okay at: - Mixing languages. It handles one language per dictation well; switching mid-sentence is less reliable, per Wispr's own limitations page. - Very long dictations. Sessions run up to 20 minutes, and a dropped connection can leave a long one incomplete. Avoid: - Dictating sensitive information before you have reviewed the privacy defaults. - Offline use — no internet, no dictation. Starter: Hold, don't tap In any text box, press and hold your shortcut (Fn on Mac, Ctrl+Win on Windows), say a two-sentence message, and release. For long thoughts, use hands-free mode instead: Fn+Space on Mac, Ctrl+Win+Space on Windows. Wispr's docs say a press shorter than half a second is treated as a tap and cancelled. Getting the hold right first saves the most common early frustration. Starter: Talk like a human Dictate "Let's meet at 2, actually 3, and first bring the slides, second bring the budget" and see what Flow types. This shows the two features that make Flow different from basic dictation: spoken corrections are resolved and spoken lists become real lists, so you don't have to talk like a robot. Starter: Teach it your words Open the Dictionary and add your own name, your company's name and two or three jargon words you use daily. Smart Formatting won't fix a word it misheard. The dictionary is how you stop the same names coming out wrong every time. Starter: Make one snippet Create a snippet with a short trigger phrase like "my signature" that inserts your full email sign-off. Snippets work on every plan and every device. One you use daily pays back the setup time within a week. Starter: Edit by voice On Pro or during your trial, turn on Command Mode under Settings → Experimental. Select a paragraph you wrote, hold Fn+Ctrl on Mac (Ctrl+Win+Alt on Windows) and say "make this shorter and friendlier". Command Mode turns Flow from a dictation tool into an editor. Try it during the trial so you know whether it's worth paying for. Pitfalls: - Text doesn't appear in some apps. Flow pastes through your clipboard; if an app refuses, use paste-last-transcript (Cmd+Ctrl+V on Mac, Shift+Alt+Z on Windows). - Permissions can reset after updates. Wispr's docs suggest re-checking Mac permissions if dictation stops working after an update. - Hitting the free weekly limit mid-week. Heavy use slows down past the allowance and then stops; the trial is the time to judge whether you need Pro. - Assuming the cleanup is proofreading. It fixes grammar and filler words, not misheard words — read before you send. Where to next: - More voice & speech tools: https://www.launchpadhq.io/discover?category=voice-speech - Meeting note-takers: https://www.launchpadhq.io/discover?category=meetings-notes - Chat assistants to polish what you dictate: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Wix AI Canonical: https://www.launchpadhq.io/platform/wix-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/wix-ai.json Last reviewed: 2026-09-09 Tool: Wix AI (https://www.launchpadhq.io/platform/wix-ai): starting price $17/mo; free tier yes; data as of not dated Describe your business, watch a website appear, then argue with it until it's yours. You type a description of your business into a box on Wix's site. A whole site appears while you watch — pages, layout, wording, placeholder pictures. From there you either drag things around by hand or type another request into a chat panel and watch the page change. You repeat that until it looks right, then publish. What it is: - Wix AI is a label, not a product — Wix sells nothing under that exact name. What it covers is a cluster of AI features inside the Wix website builder: a generator that builds a whole site from a paragraph you write, an assistant called Aria that redesigns pages and rewrites copy when you ask it to in a chat panel, a separate assistant called Astro that runs your business dashboard by chat, and image and visitor-chat tools scattered through the editor. - The comparison most people are actually making is Wix against Squarespace, and the difference is money. Squarespace has no free plan — every site starts on a 14-day trial, so the meter runs from the first day. Wix keeps a free tier that doesn't expire, so paying is an upgrade rather than the price of being online at all. - The other difference is control. Wix lets you drag any element anywhere on the page; Squarespace holds you to its layout structure. That's more freedom and more rope — which is why a bad Wix site looks worse than a bad Squarespace site, and a specific Wix site can look like nothing else. Before you start: - You can build and publish a Wix site without paying anything, and there is no card at signup. A free site lives at an address like yourname.wixsite.com/yoursite, carries a Wix ad banner across the top and a Wix icon in the browser tab, and comes with 500MB of storage and a 1GB traffic allowance. Paying is what buys your own domain name and removes that branding — from the cheapest Light plan upward, with the ability to take card payments on your site starting one tier up on Core. Wix says prices differ by country, so check what yours shows. - Most of the AI is free to use as much as you like: Wix states plainly that chatting with Aria costs nothing and does not consume AI credits (the platform's unit of spend — each thing you make costs some). Credits cover the heavier tools instead — the visitor chat, the marketing and phone agents, the Wixel design app — and a free account currently gets a small daily credit allowance, with a monthly cap on top. Premium plans include a monthly allowance that resets. - Wix has more than one editor, and which one you land in decides which AI you get. Aria lives in the newer Wix Harmony editor; the classic Wix Editor has AI tools behind an icon but no free-flowing chat, and there is a third, newer entry point called Wix Vibe with its own limits. Wix's docs don't say which one a brand-new signup gets today, so look at what's on your screen before following any tutorial. - If you meet Wix ADI in an old article, it's gone. Wix stopped supporting sites built with it on 10 November 2024; surviving ADI sites now open in the Wix Editor, and Wix names the AI Website Builder as its replacement. Security: - Wix's own generative AI policy says it develops, improves and trains its software tools — including its AI systems — on data collected since the company started, and it names user-generated content such as images and page layouts among those sources, under the licence you grant in the Terms of Use. That is Wix stating it in its own help centre, not an inference. Security: Three Wix-specific things to weigh:: - Your requests leave Wix. Wix publishes the list of outside AI companies behind its features — currently around fourteen, including Anthropic, OpenAI, Google, Meta, Microsoft Azure AI Services, ElevenLabs and Stability AI — and tells you to read their policies yourself before using its AI features. - Wix singles one out. On that same page it says users should not input their personal information when using Stability. Treat that as the floor rather than the ceiling: it's the only tool Wix names, not necessarily the only one worth caution. - Whatever the AI wrote, you published it. Wix's own risk page says AI-generated content is not guaranteed to be accurate, may resemble other companies' copyrighted material or trademarks, and may even contain other people's personal information — and that you remain responsible for the content on your site. Best at: - Getting from nothing to a live, presentable multi-page site in an afternoon, with your own wording and pictures already in place rather than filler. - Editing by conversation — asking for a different colour palette, a shorter About page or a whole new section and watching it happen, instead of hunting for the setting that does it. - Small local businesses whose site needs to take a booking, show a service list and carry a contact form more than it needs to be beautiful. Okay at: - Selling online. It works, but not on the cheapest paid tier — budget for Core before you build a shop. - Mobile layout. Wix says Aria only appears while you're editing the desktop view; switch to the mobile view and the chat panel goes away. Avoid: - Anything you might later want to move to another host. Wix's help centre says a Wix site 'needs to be hosted and operated on Wix's servers' and that its architecture doesn't support external hosting. You can point your own domain at it; that is the extent of the portability. - Committing to a template you're unsure about. Wix states you can't switch an existing site to a different template — the official route is to build a second site and copy your pages, media and plan across to it. - Starting in Harmony if you may want a different Wix editor later. Wix says a site created in the Harmony editor cannot be transferred to the Wix Editor or the Studio Editor afterwards, and existing Wix Editor sites cannot be migrated into Harmony. Starter: When you don't know what the site should say yet Type one messy paragraph about the business → Wix generates the pages, layout, colours and starter wording for you to react to. Reacting to a wrong draft beats staring at a blank page, and Wix makes the draft genuinely free: you can throw the whole site away, describe it differently and generate again as many times as you like — on a free account that costs nothing but your afternoon. Wix's own instructions invite you to edit its suggested description before generating, and to add 'preferred colors, style, or anything important to you'. Put your one non-negotiable in there rather than fixing it afterwards. Starter: When the site exists but the wording sounds like nobody Select a block of text → use Ask Aria on the action bar to shorten it, simplify the language, or change the tone. Rewriting is on the free side of Wix's meter, so you can rework the same paragraph twenty times without spending anything countable. Start with the About page. Wix names About pages among the things its AI drafts for you, and invented biography is the fastest way to lose a visitor who was otherwise ready to call you. Starter: When you need a picture and the stock photo is obviously a stock photo Open the Media panel → Photo Studio → Create AI Images → describe the shot, pick a style, save it to your site. It lands in your site's own media library rather than your downloads folder, so it's already where the page can use it. Know the meter first: Wix caps this at 1,000 generated images per user — a lifetime ceiling, not a monthly one — and says every generation counts against it even if you never save the image. Because throwaways still count, write the description you actually want before pressing generate, rather than fishing with one-word attempts. Starter: When visitors ask the same question at eleven at night Switch on Wix Smart Chat (formerly AI Site Chat) → it answers visitors from what's on your site while you're asleep. This is the one Wix AI feature that runs when you aren't there, and the one that spends the credit allowance described above rather than being free. Wix says that when the credits run out the chat drops to offline mode instead of continuing to answer — the safer failure of the two. Decide what happens at zero before you switch it on. Wix says you can configure an offline setting or hand conversations back to manual chat. Starter: When you're running the business rather than building the site Open Astro from your Wix dashboard → ask it about the site in plain English instead of navigating menus. Astro is the half of Wix's AI that isn't about design at all. Wix's launch announcement describes tracking site performance, analysing visitor behaviour and generating reports, drafting blog posts and marketing emails, and even managing collaborator roles and permissions — all by chat, in the dashboard, not the editor. Ask it a question before you ask it to change anything — 'how did the site do last month' — so you can see whether its answer matches what you already know. Pitfalls: - The free site announces itself. Perfectly fine for a hobby page — check how the ad banner actually looks on a phone before you send the link to a customer. - The mobile view is the part you build by hand. The generator works desktop-first and Aria doesn't follow you into the phone view — which is the version most of your visitors will see. - Free storage and traffic run out quietly. The allowance goes further on a text page than on a site full of AI-generated photos, which are full-size images. - The meters differ between surfaces. Wix Vibe gives a free site a small daily allowance of chat requests and a paid site a much larger one, with nothing rolling over to the next day — a different meter from the AI credits elsewhere in the product. Where to next: - More AI built into business software: https://www.launchpadhq.io/discover?category=ai-plugins-business-software - Marketing, sales and SEO tools: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai - Image generation and editing: https://www.launchpadhq.io/discover?category=image-generation-editing ### Getting Started with Wondershare PDFelement Canonical: https://www.launchpadhq.io/platform/wondershare-pdfelement-ai/guide Data: https://www.launchpadhq.io/data/v1/guides/wondershare-pdfelement-ai.json Last reviewed: 2026-09-09 Tool: Wondershare PDFelement AI (https://www.launchpadhq.io/platform/wondershare-pdfelement-ai): starting price $79.99/year; free tier yes; data as of not dated A full desktop PDF editor with an AI panel bolted on — the answer and the file stay in the same window. You open a PDF in a desktop window and work on it the way you would a document. When you want the AI, you click the button in the corner and a panel opens beside the page. You type a question about the file, read the answer, and keep asking. What it is: - PDFelement is a desktop PDF editor first and an AI tool second. Its main job is the unglamorous work: editing text inside a PDF, turning a scan into text you can actually select and search, filling and signing forms, converting to Word, combining files, blacking out details you don't want sent onward. - The AI features — asking a document questions, summarising it, translating it, flagging clauses in a contract — are a layer added on top of that. - That layering is the whole reason to choose it. A tool built only to answer questions about PDFs can tell you what page a clause is on; it cannot then edit that clause, redact the name next to it, and hand you back a signed file. Here the reading and the fixing happen in one window. - The trade is that you are buying a large piece of software to get at a side panel. Before you start: - It is software you install, on Windows or Mac, with matching apps for iPhone, iPad and Android and Linux builds on the same download page. It is not demanding — Wondershare's own install guide still lists a floor of around 512 MB of memory and 500 MB of disk space, which dates the page more than it describes the requirement. Any computer bought in the last decade will run it. - If you only want to try the AI, there is a browser version — PDFelement Online — that needs nothing installed. Its free tier currently allows a small daily number of uses of the PDF tools and a handful of AI uses in total, which is enough to find out whether the thing is useful to you before anything lands on your hard drive. - The desktop download doesn't expire, but it is a trial, not a free tier, and the limits are specific. Wondershare's trial-limits page currently states that converting a PDF to another format does half the pages — and at most three pages if the file is longer than ten; text recognition on scans stops at three pages; combining stops at three files; batch data extraction stops at two files; and compression and XFA form filling are switched off entirely. The AI panel gets around ten uses. - There are two shapes of paying for it. Wondershare currently sells a yearly subscription and a one-time perpetual licence, which costs more up front; both are priced per platform, with cross-platform costing more. Note that "single-platform" means one of Windows or Mac — not both, and not your phone. Discounts run on the store pages more or less permanently, so a crossed-out price is not a limited event. - AI use is metered separately from the licence. Plans include a limited number of AI uses; past that it is a separate add-on, sold monthly or yearly. Wondershare counts these as credits (the platform's unit of spend — each thing you make costs some), charged per use rather than per page, so a long document and a short one cost the same. Getting set up safely (Download from Wondershare's own download centre — pdf.wondershare.com/download.html — and nowhere else. This is worth more attention than it sounds: a search for "PDFelement download" returns repack and "cracked Pro" sites alongside the real one, and this is a program you are about to point at every document you own.): - Decide which platform you're on before you download rather than after. The download page carries Windows, Mac, Linux, iOS and Android builds plus older versions 6 through 11, and the licence you buy later is tied to a platform — grabbing the Mac build when you meant to buy the Windows licence is a wasted purchase, not a fixable mistake. - Skip the install entirely if trying the AI is all you want. The browser version at the same site runs the AI features with nothing to install and nothing to uninstall afterwards. - Know what registering actually does. The app opens and works without an account; a Wondershare ID is what activates a licence you've bought, and it appears to be what your AI allowance is counted against — expect to be signed in and online for the AI panel even though the editing is local. - Confirm it works before money is involved. Open one real PDF, edit a line, and run the AI panel once. That's your "installed correctly" checkpoint — and it also tells you whether your document is real text or a picture of text, which decides whether the AI can read it at all. Security: - Because PDFelement is a program on your own computer, it's easy to assume everything stays there. The editing does. The AI doesn't. Wondershare's own General Terms of Use are unusually plain about this: for features such as Summarize PDF, Grammar Check, Translate PDF, Generate PPT, AI-Written Detect and PDF to Markdown, "all these features would upload your PDFs to your personal Document Cloud first." The file goes to a server before you get an answer. Security: What that actually means for you:: - Editing offline is genuinely offline — changing text, reordering pages, filling a form and signing are local work on a local file. - The AI panel is a different decision: it needs a connection to work at all. - Wondershare's privacy policy states that it does not use this data to train its own general AI models. The same policy describes AI work being routed to outside model providers under contract — but the providers it names by name are from its video product, not PDFelement, so exactly whose servers handle a PDF isn't something the public pages settle. - Wondershare describes the desktop AI assistant as powered by ChatGPT, which is consistent with the upload clause: the thinking is happening somewhere else. Security: - None of that makes it a bad choice — it's a disclosed mechanism rather than a hidden one. The practical line is that the confidential contract is the one you edit locally and don't run the AI over, which, unlike with a browser-only tool, is an option you actually have here. Best at: - Doing the reading and the fixing in one place - Scanned documents: running text recognition on a scan and then asking questions about what it found - Contracts and forms — reviewing, filling, signing and redacting are all first-class parts of the editor rather than add-ons - Ordinary PDF chores that have nothing to do with AI — the editor half of it is a full one Okay at: - Being an ask-your-PDF tool on its own — it does that competently, but the metered AI allowance means it isn't the thing you reach for fifty times a day - First impressions. It's a dense, ribbon-heavy editor, and the AI is one panel inside a large program rather than the front door Avoid: - Buying it as an AI tool. The included allowance is small and volume is sold separately (see "Before you start") - Assuming one cheap licence covers your laptop and your phone — single-platform means one of them (see "Before you start") - Running the AI over documents you're not permitted to send to a third party. The editing is local; the AI upload is in the terms (see Security), and "it's desktop software" is not a defence Starter: Ask a scan what it says A scan is a picture of text, and the AI can't read a picture — so you run PDFelement's text recognition on it first, in the same window, and then ask. If the answers come back empty or nonsensical, that's your signal the text recognition step hasn't run — not that the document is unclear. Prompt: Summarise this document in plain language, then list every date or deadline it contains and tell me which page each one is on. Starter: Find out what you're agreeing to When the answer points at page 7, you can strike the clause, sign the document and export it without the file leaving the window. Wondershare's V13 release adds a dedicated contract-review feature on top of this; treat that as its description of the tool rather than something we've watched work. Follow up with "now list the three questions I should ask before signing" — the second pass is usually more useful than the first. Prompt: Read this agreement and list, in plain language, every obligation it puts on me, every fee it can charge me, and how either side ends it. Flag anything unusual and say what page it's on. Starter: Catch the personal details before you send it PDFelement can black out what it finds so the text is genuinely gone from the file rather than covered by a rectangle — which is the failure mode that leaks documents. Run this on anything you're about to attach to an email. It catches the details in headers, footers and form fields that you stop seeing after the third read. Prompt: List every person's name, home address, phone number, email address, date of birth and account number that appears in this document, with the page each one is on. Starter: Read something that isn't in your language An unfamiliar foreign-language form is exactly the kind of document people paste without thinking, and translation is one of the features that uploads to the cloud first (see Security). Ask for both at once — "give me the translation, then a three-line summary of what it's asking me to do" — and you'll usually know whether the full translation was even worth reading. Prompt: Translate this page into English. Keep the headings, the numbering and the order of the list exactly as they are, and tell me if any part of the original was unclear. Pitfalls: - The trial lets you do the work and then stamps it. You can edit, comment and sign in the free version — and when you export, Wondershare's own trial-limits page says the file comes out carrying a watermark, a visible mark printed across the page. You find this out after the work, not before it, which is what makes it a trap rather than a limit. - PDF editors are a standard vehicle for bundled junk, because they are software people install in a hurry — which is why the download source named above is worth being fussy about. - "Perpetual" means perpetual for this version. Wondershare's store pages describe the one-time licence as perpetual access to PDFelement 13 with free minor updates, and note that major upgrades may cost extra. It's a genuine one-time purchase; it just isn't a licence to every future version. - Asking a scanned document questions before it has been through text recognition is the single most common reason PDF AI features seem broken. Where to next: - Document & PDF tools: https://www.launchpadhq.io/discover?category=document-pdf-processing - Research & academic tools: https://www.launchpadhq.io/discover?category=research-academic-tools - Text & Conversational AI: https://www.launchpadhq.io/discover?category=text-conversational-ai ### Getting Started with Workato Canonical: https://www.launchpadhq.io/platform/workato/guide Data: https://www.launchpadhq.io/data/v1/guides/workato.json Last reviewed: 2026-09-17 Tool: Workato (https://www.launchpadhq.io/platform/workato): starting price Custom; free tier no; data as of not dated Enterprise automation you can finally sign up for on your own — with a free tier licensed for practice, not for real work. Workato calls each automation a recipe. You build one in a browser builder: pick a trigger (the "when" — the event that starts an automation), then add actions (the "then" — what happens once the trigger fires), choosing each from a list of apps rather than writing code. Each finished recipe then runs on its own. What it is: - Workato is an integration platform built for companies. It connects the large systems a business runs on — Salesforce, NetSuite, Workday, SAP, Slack, databases — so records, files and messages move between them without anyone copying and pasting. Its own connector directory currently lists more than 1,200 pre-built connections, and you can browse that directory publicly, before you sign up for anything. - It is shaped around a central team rather than one person. A Workato workspace holds up to 100 collaborators, separate environments for testing and live use, and audit logs of who changed what. Even working alone you are using a console designed for an organisation, which is why it feels heavier than a consumer automation tool — that weight is the product, not a flaw in it. - For most of its life you could not buy it without talking to a salesperson. That has changed: there is now a free self-serve account and a paid self-serve plan you can put a card on. The oddity is that Workato's own pricing page still shows no numbers and offers you a demo — the actual self-serve figures live in the documentation site instead. - Since 2026 the company has pushed hard toward AI agents: Agent Studio for building them, packaged "Genies" for jobs like IT support, and AIRO, released generally in July 2026. None of that is needed for a first recipe, and the governance story — who is allowed to let an agent do what — is what Workato actually sells there. Before you start: - Signing up is genuinely self-serve now. Workato's docs list email and password, Google, GitHub or Office 365 as ways in, with no sales contact; its free-signup page mentions signing in with a work email, so whether a personal address is accepted is something you will find out at the form rather than in advance. The free plan comes with a one-time grant of credits (the platform's unit of spend — each thing that runs costs some). One-time, not monthly: nothing refills it, and Workato does not publish what happens once it is gone. - The free plan is licensed for practice, not for real work. Workato's Self-Service Addendum, dated 18 June 2026, limits the non-paid edition to "internal trial, experimental and evaluation purposes" and has you agree not to use it "for any production purposes." Learn on it as much as you like. Anything your job depends on belongs on a paid plan. - Paying for yourself means a monthly self-serve plan, card on file, whose price climbs steeply with its credit allowance, per Workato's self-service pricing docs. Even the entry tier is steep beside Zapier or Make, which both have free tiers you are allowed to run real work on. If what you want is to learn automation rather than to connect a company's systems, start with Zapier or Make on this site and come back here when a NetSuite or a Workday is the thing standing in your way. - You cannot work out in advance how far that grant goes. Workato's docs say credits are consumed by tasks, by requests other programs make to your automations, by rows of data processed and by AI actions — but the conversion rate is not published, and the docs point you to your contract for it. You find out by running things and watching the meter. - Free accounts get no promised help. The same addendum says Workato "is not obligated to provide any technical support" on the non-paid edition, though it may offer some. The documentation site and the community forum are the realistic places to get unstuck. Security: - Workato sells its security controls by tier, and a free self-serve account sits at the bottom of that ladder. Its documentation states plainly that self-service users cannot choose the region where Workato stores and processes their automation data; those accounts run in US data centres, while paying enterprise customers pick from nine, including the EU, UK, Japan, Singapore and Australia. If the data you would route through a recipe has to stay in a particular country, a free account is the wrong place to route it. Security: Three things worth knowing before you connect anything:: - Connect test accounts, not your employer's live ones. A workspace you opened to experiment in is not where your company's real Salesforce login should sit, and Workato holds that login continuously once you grant it. - The certifications cover the platform, not your recipe. Workato publishes SOC 1, 2 and 3, ISO 27001, 27701 and 42001, HIPAA, PCI-DSS and others, and encrypts data in transit and at rest. None of that has an opinion on whether the automation you built sends the right data to the right place. - Serious deployments do not keep credentials in Workato at all. It connects to AWS Secrets Manager, Azure Key Vault and HashiCorp Vault so passwords are fetched rather than stored. You will not set that up on a free account, but it tells you what the platform expects of a real one. Best at: - Connecting the big business systems that lighter automation tools cover thinly or not at all, where the depth of the pre-built connector is the entire point. - Automations a team owns rather than a person — the shared workspace, the separate test and live environments and the audit log are all there from the first recipe. - Long processes with branching, loops and error handling, rather than a two-step "when this, then that." - Putting guardrails around AI agents — deciding centrally which systems an agent may touch and logging what it did. Okay at: - Being anybody's first automation ever. It will work, but you are learning an enterprise console to do something a simpler tool does in three clicks. - Being cheap to leave idle. On the paid self-serve plans, credits expire at the end of each billing month and do not roll over, so a quiet month is money gone. Avoid: - Running anything your work depends on from the free plan — its own addendum licenses that plan for evaluation only. - Any automation whose data must stay outside the United States while you are on a self-serve account, since choosing a region is an enterprise feature. - Reading the published self-serve prices as what Workato costs. Those are the self-serve tiers. The four contract editions — Standard, Business, Enterprise and Workato One — carry no published figures at all, and that is the version most companies actually buy. Starter: Watch a credit get spent When a new row is added to a Google Sheet → then post a message into Slack. The Slack message is not the point; watching the grant tick down for the first time is, on a platform that publishes no conversion rate between credits and runs. Zapier tells you a task is a task and gives you a fixed monthly allowance of them, so the only way to calibrate here is to spend a little deliberately and look at the meter. Fire the trigger by hand the first few times — a schedule that runs every five minutes is the fastest way to find out what your grant was worth. Starter: Start from somebody else's recipe When you need a common connection → then clone a published recipe from Workato's community library and change it, instead of building from an empty canvas. Workato's docs describe cloning community recipes straight into your own workspace. On a plan whose credits never refill, that means fewer failed runs spent learning what a working one looks like — a different calculation from Zapier's templates, where a botched attempt costs a task out of an allowance that resets. Starter: Connect one big system, not a small one When a record changes in Salesforce, NetSuite or Workday → then update a sheet or notify a channel. Because the connector directory is public, you can check whether your particular NetSuite or Workday is covered before you make an account. Point Workato at Gmail instead and you have learned nothing it does better than Zapier. Workato says AIRO can draft an automation from a plain-English description of the goal, but whether that is switched on for a free self-serve account is not stated publicly, so treat it as something to look for in the builder rather than something to count on. Starter: Break a recipe on purpose, then read the run record When a recipe fails → then open its job history to see which step stopped and exactly what data it was holding. This record has a shelf life. Workato publishes job-history retention of 30 days on its Standard and Business editions and 90 on Enterprise, and states no figure for self-serve accounts — so assume what you can look back at is limited, and read a run while it is fresh rather than reconstructing it later. Pitfalls: - Build with a trigger you fire by hand before you build one that fires on a schedule. Every tutorial run, every failed test and every accidental loop comes out of the same one-time grant, and a recipe left on a five-minute schedule overnight is the classic way to wake up to nothing left. - "Recipe" here means one automation, not a set of instructions for an AI. Workato has used the word since long before AI tools borrowed it, and the docs, the community library and the support forum all assume that meaning. Searching for "Workato recipes" gets you automations. - If you look at the marketing site and conclude there is no published price, you have found the sales funnel rather than the answer. The figures are on the docs site. - Features you read about in the docs may not exist in your account. Choosing a data region, extended concurrency, longer job retention and raised quotas are all attached to paid editions. Workato's documentation describes the whole platform, not your plan, and it does not always say which is which. - A paid self-serve subscription renews itself. Workato's addendum says subscriptions renew automatically at its then-current pricing and that fees are non-refundable; you can cancel at any time, but it takes effect at the end of the billing period rather than immediately. Where to next: - Workflow & Automation: https://www.launchpadhq.io/discover?category=workflow-automation - AI Plugins for Business Software: https://www.launchpadhq.io/discover?category=ai-plugins-business-software ### Getting Started with Writesonic Canonical: https://www.launchpadhq.io/platform/writesonic-chatsonic/guide Data: https://www.launchpadhq.io/data/v1/guides/writesonic-chatsonic.json Last reviewed: 2026-09-17 Tool: Writesonic / Chatsonic (https://www.launchpadhq.io/platform/writesonic-chatsonic): starting price $79/month; free tier no; data as of 2026-10-25 Find out whether ChatGPT and Gemini mention your business — and what to do when they don't. You add your brand name, your competitors, and a list of questions your customers might actually ask. Every day the platform puts those questions to the big AI assistants and records the answers. You come back to a dashboard showing whether you were mentioned, which sites were quoted instead, and how that has moved since last week. What it is: - Writesonic started life as an AI writing tool and a lot of the internet still describes it that way. It isn't one anymore: the company has rebuilt around a single question — when someone asks ChatGPT, Gemini, Claude or Perplexity for a recommendation in your line of work, does your name come up? - It currently says it checks around ten AI assistants — ChatGPT, Claude, Gemini, Google AI Overviews, Perplexity, Grok, DeepSeek, Microsoft Copilot and Meta AI among them — by working through each one's normal web interface the way a customer would, rather than through a back-door connection. That is the company's own claim, and it is why the numbers move day to day. - Alongside the tracking, the same subscription includes tools to act on what you find: an AI article writer, site audits that crawl your pages and flag what is holding them back, and content scoring. - Chatsonic, the chatbot this product was once best known for, still lives in the left-hand menu of the account. It is no longer what Writesonic sells, and you should not buy this expecting a ChatGPT competitor. Before you start: - This one costs money on day one. There is no permanently free plan. The signup page offers a 7-day free trial with no credit card, and after that the cheapest plan — currently Starter, which costs less per month if you pay for a year up front — is the floor. - You need a website you control before this is worth anything: roughly half of what you are buying is diagnosis of your own pages — audits, content scores, fixes. With no site to point it at, you are paying for a dashboard that tells you a competitor is winning and gives you nowhere to put the answer. - The number that actually runs out is prompts (the messages you type — here, the customer questions you ask the platform to track). Plans meter these monthly, currently in the region of 100 on Basic and 200 on Growth, with separate monthly caps on articles and site audits. Article count is what people look at; tracked prompts is what constrains what you can learn. - Come with a list of the questions a real customer would type before choosing someone like you — "best accountant for freelancers in Leeds," not "accounting." Those literal sentences are the raw material. Security: - Writesonic's privacy policy, last updated 13 May 2026, is unusually direct on the point people worry about most: it states the company does not use customer data to train or fine-tune any general-purpose, foundation or large language model, though it reserves the right to use de-identified data to improve the service. Security: Where your text actually goes: - The policy names the outside companies that process content on Writesonic’s behalf: OpenAI; Anthropic; Microsoft Azure OpenAI Service; Microsoft Azure, for Anthropic models; Stability AI; OpenRouter; Google Cloud Platform; AWS, for custom models. - Writesonic says each handles data differently and that some are configured so they retain nothing. - Practical reading: your drafts pass through several third parties, under Writesonic’s contracts rather than yours. Security: What you are handing over when you connect things: - The write-and-publish side asks to link to accounts that matter: Google Search Console, WordPress and SEO tools such as Ahrefs are the ones the vendor names. - A Search Console link exposes your real search performance data; a WordPress link can put text on your live site. - Connect them one at a time, and only once you have decided you are keeping the subscription. Security: - On retention, the policy says data from deleted accounts is held up to 30 days and then permanently deleted, while billing records are kept seven years and support tickets three. Cancelling is not the same as erasing. Best at: - Answering a question no other tool on this site answers: whether AI assistants currently name your business, and who they name instead - Showing change over time — the daily re-run turns a curiosity into a trend you can act on - Naming the specific pages and sites the AI assistants quoted, so you can see who is actually being read - Comparing you against competitors you name yourself, rather than against whoever the tool guesses Okay at: - Writing the articles. The drafts are serviceable, but this is no longer where the company's effort goes and a dedicated writing tool does it better - Small or very local businesses. The method depends on AI assistants having an opinion about your category at all, and for a narrow niche the honest answer may be that nobody is asking Avoid: - Judging the product on the 7-day trial. What you are buying is a trend line built from daily runs, and a week gives you a dashboard and almost no signal. Budget a month - Buying it for Chatsonic. The chatbot is still in the app but absent from the pricing comparison table, the homepage, and about six months of changelog entries — what an unsupported feature looks like before anyone announces it - Treating the visibility score as comparable to anyone else's. It is Writesonic's own measurement of its own prompt list — a direction of travel, not an absolute - Relying on the generated text without reading it. The terms of service dated 12 May 2026 put this in writing — outputs "may be inaccurate, incomplete, biased, offensive, infringing," and the customer is "solely responsible for evaluating Outputs" before publishing them Starter: Track the question that actually decides a sale Add this sentence to the tracked-prompt list, not a chat box: it is not sent off for an answer, it goes onto a list re-run daily, and the output is a record of whether your name appeared. Tested against ChatGPT, which answers once, now, and Jasper, which helps you write copy about being the best option: neither keeps asking on a schedule and keeps score. Add a version with a location or price qualifier — answers diverge sharply between "best CRM" and "cheapest CRM for two people." Prompt: What is the best [your category] for [your specific customer]? Give me three options and say why. Starter: Name your competitors before you look at anything else In setup, enter the rivals you actually lose deals to; the reports then show share of voice and citations for you and them side by side. The comparison is the product; a visibility score with nothing to compare it to is a number without a unit. Tested against Jasper, whose brand voice settings shape how your copy sounds rather than who cites you, and ChatGPT, which can tell you who your competitors probably are but will not track them daily against a fixed prompt list. Starter: Run the two free tools before you pay anything Writesonic publishes an AI Crawl Checker and an llms.txt Generator as free tools — llms.txt being a small file you put on your site to tell AI crawlers what it contains and how to read it. Run both against your domain first. These are the only parts of the platform that need no account, and they answer the cheapest version of the question — whether AI crawlers can read your site at all, because if they cannot, no amount of visibility tracking fixes that first. Tested against ChatGPT: it can explain what an llms.txt file is; it cannot generate one against your live site or tell you whether your server is blocking crawlers. Starter: Run one site audit and read only the citation findings Point a site audit at your domain. It crawls up to the page cap on your plan and returns a list of issues; on the first pass read only what it says about being quotable. The audit's page cap is one of the things the plan tiers actually meter, so this is a resource you spend, not a free scan. Tested against Jasper, which does not audit sites at all, and ChatGPT, which cannot crawl your site on a schedule or hold the previous crawl to compare against. Starter: Ask Chatsonic the one thing it is better placed to answer than a plain chatbot Chatsonic's documented reason to exist inside this product is those connections, so the only prompts worth typing here rather than into ChatGPT are ones needing your real data. It is documented as reachable from the left-hand menu, but whether it is in the tier you bought is something you may need to check once you are signed in. If the connections are not available to you, this card is the one to skip — it is the only one that depends on them. Prompt: Using my connected Search Console data, which of my pages already get impressions for questions people would ask an AI assistant? List them with the query. Pitfalls: - Writesonic's own pages disagree about what it costs. The live pricing page currently shows Starter with Basic and Growth above it, while the developer docs' subscription page lists no Starter at all and gives Basic a different price. Read the pricing page on the day you buy; third-party review figures, including this year's, are not reliable. - Its marketing pages are stale in a way that will mislead you: writesonic.com/chat still advertises Chatsonic with a free tier and a cheap Individual plan, and still names models from 2024, while the live plan table has neither. The pricing table is what you will be charged. - Diary the refund window. The terms dated 12 May 2026 give self-serve customers a full refund within 7 days of purchase by email request, and fees are non-refundable after that except where the law requires otherwise; enterprise and agency subscriptions are non-refundable from the start. That window and the 7-day trial are not the same thing. - Everything you see is scoped to the prompts you wrote. If you guess the questions badly, the platform will faithfully report that you are invisible for questions nobody asks. Revisit the list monthly; it is the input, not a setting. - You own what it generates, but the terms make that ownership conditional — on third-party rights in underlying material and on the terms of the model providers behind the scenes. Where to next: - Marketing and SEO tools: https://www.launchpadhq.io/discover?category=sales-marketing-seo-ai - AI chatbots, if a chatbot is what you wanted: https://www.launchpadhq.io/discover?category=text-conversational-ai - Research tools: https://www.launchpadhq.io/discover?category=research-academic-tools ### Getting Started with Zapier Canonical: https://www.launchpadhq.io/platform/zapier/guide Data: https://www.launchpadhq.io/data/v1/guides/zapier.json Last reviewed: 2026-06-25 Tool: Zapier (https://www.launchpadhq.io/platform/zapier): starting price Free, from $19.99/mo billed annually; free tier yes; data as of 2026-10-25 Connect the apps you already use and let them talk to each other — no code. You build little automations called Zaps. Each one has a trigger (the "when" — something that happens in one app) and an action (the "then" — what should happen in another). You set them up by clicking through menus, not by writing code. What it is: - Zapier is the duct tape of the internet. It links apps that don't naturally work together — Gmail, Slack, Google Sheets, your calendar, thousands more — so when something happens in one, something automatically happens in another. - It's the go-to first step for anyone who keeps repeating the same copy-paste between two apps. - It's also leaned into AI lately: a built-in Copilot can draft an automation from a plain-English description of what you want, and Zapier now offers AI chatbots and "agents" (free versions included) — though you don't need any of that for your first automation. Before you start: - Free to start, no card. You connect your existing accounts — your real Gmail, your real Slack — so Zapier acts on your behalf inside them. Keep that in mind from the very first click. First step: sign up at zapier.com, connect one app you already use, and try a starter automation below. - Will you need to pay? Free is fine for a couple of simple automations. The free plan gives you a small monthly allowance of "tasks" (one task = one automated action that runs) and limits you to two-step Zaps (one trigger plus one action — it's called "two steps" because the trigger counts as the other step). A busy automation can burn through the free allowance fast, sometimes in a week. You'd move to a paid plan for multi-step flows or more volume. - Run out of tasks and your automations pause rather than vanish — they're held and re-run after the monthly reset. Security: - This is different from a chatbot: the risk isn't what you type, it's what you connect. Zapier holds ongoing login access to every app you link, which makes your Zapier account a master key — if someone got into it, they could reach everything you've connected to it. Security: Three habits keep that safe:: - Lock down the Zapier account itself — a strong, unique password plus two-factor authentication. This matters more here than for most tools you use. - Connect only the apps you actually need, and disconnect any you've stopped using. - Be thoughtful about sensitive apps — data in a Zap passes through Zapier's servers, so think twice before routing banking, health, or confidential information through one. Best at: - Connecting two apps that don't talk to each other — "new form submission → add a spreadsheet row → ping me on Slack." - Killing repetitive copy-paste between tools you already use. - Letting non-technical people automate real work without writing any code. Okay at: - Complicated multi-branch logic — possible on paid plans, but fiddly; Make and n8n handle complex flows more cleanly. - Instant reactions — the free tier checks for changes on a schedule rather than instantly, so "real-time" can really mean several minutes. Avoid: - High-volume automation — the per-task pricing climbs fast. - Using it as a database or for heavy data crunching — it moves data between apps, it doesn't store or process it. - Anything that genuinely can't fail for a second while you're on the free tier. Starter: Never lose a form lead When someone submits your Google Form or Typeform → then add a row to a Google Sheet. A two-step classic with an instantly visible payoff — and the cleanest possible first Zap to build. New to building Zaps? Describe it to Zapier's Copilot in plain English — "when someone submits my Google Form, add a row to my sheet" — and it'll draft the Zap for you to review before you switch it on. Starter: Turn starred emails into a to-do When you star an email in Gmail → then create a task in Todoist or Trello. Teaches an important idea: your own action can be the trigger, not just an event coming from outside. Starter: Get a heads-up when something matters When a new row appears in a specific Google Sheet → then Slack or email yourself. Notifications are the most satisfying first payoff — you watch the automation do its job in real time. Pitfalls: - You'll burn free tasks faster than you think. Before you switch a Zap on, ask: how often will this realistically fire? - Always test before switching on. A misconfigured Zap can fire a hundred times in a row — spamming you and torching your task budget. Use the test step every single time. - The free two-step limit bites mid-build. Know that ceiling before you plan a clever four-step flow you can't actually run yet. - Set-it-and-forget-it cuts both ways. A Zap you built months ago is still quietly running. Keep a mental list of what you've got switched on. Where to next: - Workflow & Automation: https://www.launchpadhq.io/discover?category=workflow-automation - Meetings & Notes: https://www.launchpadhq.io/discover?category=meetings-notes ### Getting Started with Zoom's Built-in AI Canonical: https://www.launchpadhq.io/platform/zoom-ai-companion/guide Data: https://www.launchpadhq.io/data/v1/guides/zoom-ai-companion.json Last reviewed: 2026-09-09 Tool: Zoom AI Companion (https://www.launchpadhq.io/platform/zoom-ai-companion): starting price See pricing; free tier yes; data as of not dated The meeting summaries Zoom used to call AI Companion — already in the app, if your admin let you have it. You're in a Zoom call. An AI icon sits in the meeting toolbar; the host clicks it to set the AI going, and everyone's icon changes colour while it runs. Mid-call you can type it a question without interrupting anyone. When the call ends, a write-up lands in your email and in the meeting chat. What it is: - Zoom's AI is the note-taking and summarising built into the Zoom app you already have. It writes up your meetings, answers questions while a call is still running, and drafts chat and email replies. Until recently all of it was branded AI Companion; Zoom began removing that name from Zoom Workplace in June 2026, so the settings are now labelled by what they do — meeting summary, chat compose — and searching your account settings for "AI Companion" may turn up nothing at all. - Why this rather than a separate notetaker: there is no second vendor — nothing joins your call as a guest, no other company ends up holding your recordings, and nothing extra to buy if your employer already pays for Zoom. - What is not included: ZoomMate is a separate paid product that searches across your files and other apps and carries out multi-step tasks. It replaces the old Custom AI Companion add-on and is priced per user each month. Everything else on this page comes with the plan you already have. Before you start: - The first question isn't what it costs — it's whether your admin has switched it on. These are account settings, not personal ones. Whoever runs your organisation's Zoom account enables them for everyone or for particular groups, and can lock the choice so an individual can't override it. If the icon isn't in your toolbar, there is no setting of your own that will summon it. - Assuming it is on, the core features come with the plan rather than as an extra charge. Zoom's own comparison page says AI is included at no extra cost with paid Zoom Workplace plans, the cheapest being an annual Pro subscription. Zoom's support articles list Pro, Business and Enterprise as the plans that qualify for meeting summary, and don't list the free Basic plan. - The free plan isn't empty, but it is metered. Zoom's pricing page currently shows a small monthly allowance on free Basic — only a handful of meeting summaries, with similarly small limits on the other features. Zoom's help centre and its pricing page don't currently say the same thing about free access, so check what your own account actually offers before planning around it. - First step: host one internal meeting, start the AI yourself, and read the write-up against your own memory of the call. Don't point it at a customer on day one. Security: - Zoom's own terms put this on you, by name. Section 7 of the Terms of Service says the host is responsible for compliance with all laws governing the monitoring or recording of conversations. Whether one person's agreement is enough or everyone's is required depends on where each person is physically sitting, and a call spanning several places is governed by whichever of those rules is strictest. The specifically Zoom-shaped trap is the gap between recording and summarising: Zoom documents a consent pop-up for recording, where a participant clicks OK or leaves the meeting, but its article on meeting summary describes something much quieter — the AI icon changing colour to show the feature is active. An icon is not a disclosure. Say it out loud when you start it, and keep pay, health, discipline and legal matters out of the transcript entirely. None of this is legal advice. Security: What Zoom says it does with what it hears: - Zoom's Terms of Service — the binding document, not a blog post — states that Zoom does not use your audio, video, chat, screen sharing, attachments or other communications-like content to train Zoom's or third-party AI models (the AI "brain" that does the actual thinking). That wording was added in August 2023 after an earlier version of the terms was widely read as granting Zoom exactly those rights; the reversal was reported at the time and the commitment has been repeated on Zoom's own pages since. - Some answers are produced by outside model providers — Zoom names Anthropic, OpenAI and Perplexity — unless the account is configured to use only models Zoom runs itself. That is an account-level choice made by whoever administers the account, not by you. - Zoom offers a choice of which region the processing happens in, which is the kind of control that only exists on a business plan and only if someone has actually set it. Security: Where the write-up ends up: - The write-up is addressed to the meeting, not to you — so a correction you make is a correction everyone sees. - Zoom's documentation says people who joined without signing in to a Zoom account don't receive it automatically. The person most likely to be misquoted can be the one who can't read the write-up. - Whether the transcript behind the summary is retained for other AI features to use is one account setting; whether you, as host, can see and delete that transcript is a second, separate one. You may end up with a summary you can read and a transcript you can't reach. - Cloud recordings are described as retained for as long as the account is active, unless someone has set them to auto-delete. Best at: - Recurring internal calls, where the write-up only has to be good enough to jog four people's memories. - Joining a call late and needing to know what was already decided without stopping the meeting to ask. - Organisations already standardised on Zoom, where the notes need no purchase order. Okay at: - Meetings that aren't on Zoom. The AI can be sent into Google Meet and Microsoft Teams calls, but it takes more unlocking than anything else here. - Being your record of what was decided. It's reliable about what was said, which isn't the same thing. - Long webinars and calls with heavy cross-talk, where speaker labels get muddled. Avoid: - Any meeting you're not hosting. Starting and stopping is a host and co-host control, so on someone else's call the most you can do is ask them. - Planning around ZoomMate outside North America right now — Zoom launched it there on 1 June 2026 and says EMEA and APAC are expected later, so you may read the marketing before you can buy the product. - Assuming a summary means you also have a recording. Zoom keeps meeting summary and cloud recording as separate settings, so a call can end with a tidy write-up and no video to check it against. Starter: When you're hosting an internal call → then start the AI before anyone speaks The host clicks the AI icon at the top of the meeting, before anyone has started talking. tl;dv and the other notetakers on this site send a bot that appears in the participant list and has to be admitted; Zoom's runs inside the meeting client itself, so there's no extra attendee for anyone to query and no waiting room to clear. Ask a colleague what appeared on their screen. That is the only honest way to learn what your participants are actually told. Starter: When you join twenty minutes late → then ask it what you missed Type a question into the AI panel while the meeting is still running instead of interrupting to ask. Otter is built around the transcript you read after the call ends. This one is documented as answering during the meeting, which is what makes it usable at the moment you actually need it — though only if the host has already started it. Ask it something you already know the answer to. Two minutes of that tells you more about its accuracy than a week of reading its summaries. Starter: When the meeting isn't on Zoom → then send Zoom's AI to it With calendar sync set up, Zoom's AI can be invited into a Google Meet or Microsoft Teams meeting to transcribe and summarise it. This one carries an admin setting of its own on top of calendar sync — so it's the feature most likely to be missing on a locked-down company account where everything else works fine. tl;dv covers all three platforms out of the box with nothing to unlock. Zoom describes this as inviting its AI to the meeting, so expect it to show up as a participant there — the no-extra-attendee advantage is a Zoom-calls-only benefit. Starter: When the icon isn't there on a work account → then ask for the setting by its name Sends your admin to a specific switch in the admin portal rather than to a general question about AI. Zoom didn't build one master toggle. Meeting summary, in-meeting questions and third-party meetings are each their own separately named account setting, so "please turn on AI" gets you a shrug and naming the one you want gets you the feature. If it comes back locked, that's a decision rather than a fault. A personal Zoom account on a paid plan is the way round it, at your own expense. Pitfalls: - A summary nobody re-read is not a record. It arrives fluent, fast and before anyone has had time to think about the meeting, which is exactly why it gets forwarded as if it were minutes. - Auto-start runs without you. Zoom's settings allow a summary to begin automatically, and its own article notes this can happen without the host being present — so a call that starts before you join can already be being written up by the time you arrive. - It hears the two minutes before the meeting starts. Small talk, a comment about a colleague, someone mentioning why they were off last week — it's all speech and it can all land in the transcript. Start it when the meeting starts, not when you open the room. - Nothing tells you when the free allowance runs out except the absence of a summary. On a metered plan the button can still be there when the month's allowance isn't, so check that the write-up actually arrived rather than assuming it did. Where to next: - Notetakers that work across Zoom, Meet and Teams alike: https://www.launchpadhq.io/discover?category=meetings-notes - More AI hidden inside software you already pay for: https://www.launchpadhq.io/discover?category=ai-plugins-business-software - Push your meeting notes into your other tools: https://www.launchpadhq.io/discover?category=workflow-automation ## Workflows Canonical: https://www.launchpadhq.io/workflows ### Blog post generator Canonical: https://www.launchpadhq.io/workflows/blog-post-generator Data: https://www.launchpadhq.io/data/v1/workflows/blog-post-generator.json Research, outline, draft, and polish a publish-ready article using complementary AI tools. #### Optimal: Cited research, long-form draft, then a grammar pass Time: About 45–90 minutes for a 1,000-word post Result: A cited outline plus a publish-ready draft with a grammar pass - Perplexity AI (https://www.launchpadhq.io/platform/perplexity-ai): starting price $20/month; free tier yes; data as of 2026-10-26. Best for research with cited sources. Ideal when you need accurate, web-grounded facts for your post. - ChatGPT (https://www.launchpadhq.io/platform/chatgpt): starting price Free, from $8/month; free tier yes; data as of 2026-10-24. Best for drafting and rewriting in a conversational flow. Easy for beginners to iterate on tone and structure. - Grammarly (https://www.launchpadhq.io/platform/grammarly): starting price Free, from $12; free tier yes; data as of 2026-10-26. Best for polishing. Catches mistakes and improves clarity so your final draft reads professional. 1. Research with citations (Perplexity AI): Ask for sources on your topic. Save the URLs you will actually cite. Do not copy prose from the answer. Prompt: Research this topic for a beginner audience: [TOPIC]. List 6–8 current facts with source URLs. Flag anything that looks opinion, marketing, or older than 18 months. I need a blog post, not a recap of your answer. 2. Outline and draft (ChatGPT): Paste your notes (not a dump of search results). Ask for an outline first, then a draft in your voice. Edit facts against the sources from step 1. Prompt: Using only these notes and URLs: [PASTE NOTES] Write a clear outline (H2s + bullets), then a draft of about 1,000 words for beginners. Do not invent sources. Mark any claim you cannot support from the notes as [unverified]. Tone: practical, not hype. 3. Polish grammar and clarity (Grammarly): Paste or open the draft in Grammarly. Accept only edits that keep your meaning. Recheck any rewritten sentence against your sources. #### Hybrid: One paid chat tool plus a workspace Time: About 40–75 minutes Result: A structured draft living next to your notes in Notion - ChatGPT (https://www.launchpadhq.io/platform/chatgpt): starting price Free, from $8/month; free tier yes; data as of 2026-10-24. Balanced choice for outlining and drafting without a separate research tool. Strong free tier for everyday posts. - Notion AI (https://www.launchpadhq.io/platform/notion-ai): starting price Free, from $10; free tier yes; data as of 2026-10-26. Keeps drafts and edits in one workspace. Fits this tier when you want structure and light AI assists in Notion. 1. Draft in chat (ChatGPT): Use ChatGPT (Free or Plus) to outline and write. Search the web yourself for anything that must be cited. Prompt: I am writing a blog post on [TOPIC] for beginners. Ask me 5 clarifying questions, then propose an outline. After I approve the outline, write the draft. Do not invent statistics. 2. Organize and light-edit in Notion (Notion AI): Move the draft into a Notion page. Use trial AI or Agent (Business) only to tighten headings and a short summary — you still own the facts. Prompt: Tighten this draft’s headings and write a 2-sentence intro. Do not add new claims: [PASTE DRAFT] #### Budget: Free chat plus a free grammar checker Time: About 40–80 minutes Result: A readable draft you can publish after a free grammar pass - Gemini (https://www.launchpadhq.io/platform/gemini): starting price Free, from $4.99/month; free tier yes; data as of 2026-10-26. Free and capable for drafting and brainstorming. Good when keeping cost at zero is the priority. - LanguageTool. Free grammar and style checking without a paid editor. Pairs well with any free chat model for cleanup. 1. Draft in Gemini (Gemini): Use the Gemini app on the Free plan. Keep facts you can check yourself. Paid Google AI plans exist; this recipe stays on Free. Prompt: Write a beginner blog post on [TOPIC], about 800–1,000 words, with H2s. No invented stats. End with 3 questions I should verify before publishing. 2. Check grammar (free) (LanguageTool): Paste the draft into LanguageTool’s free checker. The free Basic checker limits how much text you can paste at once, so check long posts in chunks. ### Image generation Canonical: https://www.launchpadhq.io/workflows/image-generation Data: https://www.launchpadhq.io/data/v1/workflows/image-generation.json Concept art, marketing visuals, and iterations from prompt to final asset. #### Optimal: Pro quality - Midjourney (https://www.launchpadhq.io/platform/midjourney): starting price $10/mo; free tier no; data as of not dated. Top-tier aesthetic quality for this workflow. Best when you want standout visuals and accept a paid stack. - Leonardo AI (https://www.launchpadhq.io/platform/leonardo-ai): starting price Free, from $28/mo; free tier yes; data as of not dated. Fast iteration and strong models in-app—ideal second step after Midjourney for more variants or a different aesthetic. #### Hybrid: Balanced - Adobe Firefly (https://www.launchpadhq.io/platform/adobe-firefly): starting price Free, from $199.99/mo; free tier yes; data as of not dated. Solid generative images inside Adobe’s ecosystem. Balances quality with brand-safe workflows. - Canva AI (Magic Studio) (https://www.launchpadhq.io/platform/canva-ai-magic-studio): starting price Free, from $12.99/mo; free tier yes; data as of 2026-10-26. Layouts, templates, and quick social-ready exports. Ideal hybrid path for marketing visuals without deep design tools. #### Budget: Free-first - Leonardo AI (https://www.launchpadhq.io/platform/leonardo-ai): starting price Free, from $28/mo; free tier yes; data as of not dated. Strong free credits and fast iteration for concepts. Fits budget runs before you invest in premium gens. - Ideogram (https://www.launchpadhq.io/platform/ideogram): starting price Free, from $20/month; free tier yes; data as of 2026-10-26. Competitive free tier for text-in-image and general generation. Good default when cost must stay low. ### Voice & narration Canonical: https://www.launchpadhq.io/workflows/voice-clone Data: https://www.launchpadhq.io/data/v1/workflows/voice-clone.json Text-to-speech, cloning, and podcast-ready audio without a studio. #### Optimal: Studio-grade - ElevenLabs (https://www.launchpadhq.io/platform/elevenlabs): starting price Free, from $5 equivalent, Creator $18.33, Pro $82.50); free tier yes; data as of 2026-10-14. Leading natural TTS and cloning when quality is non‑negotiable. Optimal tier for broadcast-style narration. - Descript (https://www.launchpadhq.io/platform/descript): starting price Free, from $24/person/mo monthly or $16 annual; free tier yes; data as of 2026-10-14. Edit audio like a doc and clean up takes fast. Pairs with premium TTS for a full studio-style pipeline. #### Hybrid: Balanced - ElevenLabs (https://www.launchpadhq.io/platform/elevenlabs): starting price Free, from $5 equivalent, Creator $18.33, Pro $82.50); free tier yes; data as of 2026-10-14. Invest in voice quality while saving on other tools. Balanced if narration is the main spend. - Murf AI (https://www.launchpadhq.io/platform/murf-ai): starting price Free, from $29/mo; free tier yes; data as of not dated. Compare a second premium TTS without Audacity—useful for A/B voices before you lock a take. #### Budget: Free-first - Murf AI (https://www.launchpadhq.io/platform/murf-ai): starting price Free, from $29/mo; free tier yes; data as of not dated. Try free tiers for voiceovers before upgrading. Fits narrated explainers on a tight budget. - Inworld AI (https://www.launchpadhq.io/platform/inworld-ai): starting price Free, from $25/month; free tier yes; data as of 2026-10-25. Second voice tool to compare side by side. Its pay-as-you-go plan starts free and includes voice cloning, though it asks for a card on file before you create a professional clone. ### Video editing & shorts Canonical: https://www.launchpadhq.io/workflows/video-editing Data: https://www.launchpadhq.io/data/v1/workflows/video-editing.json Script, generate B-roll, edit, and export for social or web. #### Optimal: Generate clips, then assemble in a template editor Time: About 1–3 hours for a 30–60s short Result: A social-ready short with AI B-roll and captions from a template editor - Runway (https://www.launchpadhq.io/platform/runway): starting price $15/mo monthly or $12/mo billed annually; free tier yes; data as of 2026-10-14. Generative video and effects at pro level. Optimal when AI shots are central to your edits. - InVideo AI (https://www.launchpadhq.io/platform/invideo-ai): starting price $10 per seat/mo; free tier yes; data as of 2026-10-14. Templates and AI-assisted assembly in one product—pairs with Runway for gen-to-edit without a legacy NLE. 1. Script the short (Runway): Write a 20–40 second script first (you can use any notes app). Then generate only the shots you cannot film. Prompt: Turn this idea into a 30-second vertical script with 5 shot descriptions (who/what, camera, duration): [IDEA] Keep on-screen text under 8 words per shot. 2. Generate B-roll (Runway): Generate short clips from each shot line. Stay inside your plan’s monthly credits. 3. Assemble and export (InVideo AI): Upload or describe the sequence in InVideo. Use a short-form template, add captions, export in 9:16. Prompt: Create a 30-second 9:16 video. Hook in the first 2 seconds. Use this script and shot list: [PASTE]. Add burned-in captions. No extra claims beyond the script. #### Hybrid: Templates plus Google generative clips Time: About 45–120 minutes Result: A template-built short with one or two generated clips - InVideo AI (https://www.launchpadhq.io/platform/invideo-ai): starting price $10 per seat/mo; free tier yes; data as of 2026-10-14. Template-heavy edits with AI assists—faster than building from scratch. Balanced cost for social content. - Google Veo 3.1 (https://www.launchpadhq.io/platform/google-veo-3-1): starting price See pricing; free tier yes; data as of not dated. Google’s generative video model for clips and ideas—pairs with InVideo for a full AI-first hybrid stack. 1. Generate a clip in Google’s video tools (Google Veo 3.1): Use Veo from the Gemini / Flow path your Google AI plan includes. Quotas and which Veo variant you get depend on the plan — treat limits as unverified unless you see them in-product. Prompt: 5-second clip, 9:16, natural lighting, [SUBJECT] [ACTION], no on-screen text, no logos. 2. Build the rest in InVideo (InVideo AI): Drop the generated clip into an InVideo short template. Write the hook yourself so the AI does not invent claims. Prompt: Assemble a 20–30s social video around this clip description and script: [PASTE] Keep brand claims only if I wrote them. #### Budget: Free-first generators Time: About 30–90 minutes Result: A few generated clips you can post or drop into any editor - Luma Dream Machine (Ray3) (https://www.launchpadhq.io/platform/luma-dream-machine-ray3): starting price $30/mo; free tier no; data as of 2026-10-14. Generate clips on a generous free path when available. Budget-friendly B-roll before heavier editing. - Google Veo 3.1 (https://www.launchpadhq.io/platform/google-veo-3-1): starting price See pricing; free tier yes; data as of not dated. Extra generative shots on Google’s stack—stays free when your quota allows; good variety next to Luma. 1. Generate on Luma’s free path (Luma Dream Machine (Ray3)): Start from Luma’s try-for-free entry. If the product asks you to pay, stop — this tier stays free. Prompt: Short 5s clip of [SCENE], handheld, daylight, no text, no celebrity likeness. 2. Optional second generator (Google Veo 3.1): If Gemini Free (or your existing Google AI plan) offers a video generate control, make one alternate clip. If it requires a paid plan you do not have, skip this step. ### Code writing & shipping Canonical: https://www.launchpadhq.io/workflows/code-writing Data: https://www.launchpadhq.io/data/v1/workflows/code-writing.json From prototype to PR: AI-assisted coding with review and tests. #### Optimal: Ship fast - Cursor (https://www.launchpadhq.io/platform/cursor): starting price Free, from $8/mo; free tier yes; data as of 2026-10-25. Agentic edits inside your repo. Optimal when you want AI to reason across files like an IDE teammate. - GitHub Copilot (https://www.launchpadhq.io/platform/github-copilot): starting price Free, from $10/mo; free tier yes; data as of 2026-10-25. Inline completions and chat in VS Code–style workflows. Best paired with Cursor for speed and safety. #### Hybrid: Balanced - Claude (https://www.launchpadhq.io/platform/claude): starting price Free, from $17/month; free tier yes; data as of 2026-10-26. Long-context reasoning for design and refactors. Hybrid when you want one strong model without full IDE lock-in. - Gemini (https://www.launchpadhq.io/platform/gemini): starting price Free, from $4.99/month; free tier yes; data as of 2026-10-26. Second assistant for quick tries and multimodal snippets—pairs with Claude without listing a non-AI editor. #### Budget: Free-first - Replit Agent (https://www.launchpadhq.io/platform/replit-agent): starting price $20/mo ($18/mo billed annually) including $20 toward more powerful models; free tier yes; data as of 2026-10-14. Hosted coding with AI help on free tiers for small projects. Ideal for prototypes without local setup. - Ollama (https://www.launchpadhq.io/platform/ollama): starting price Free; free tier yes; data as of not dated. Run open models locally at no API cost. Budget pick when privacy and free inference matter most. ### Audio transcription Canonical: https://www.launchpadhq.io/workflows/audio-transcription Data: https://www.launchpadhq.io/data/v1/workflows/audio-transcription.json Turn recordings into searchable text, summaries, and action items. #### Optimal: Accuracy + workflow - Otter AI (https://www.launchpadhq.io/platform/otter-ai): starting price Free, from $16.99/user/mo billed monthly; free tier yes; data as of 2026-10-25. Live transcription with speaker labels and search. Optimal when meeting notes must be reliable and shareable. - Notion AI (https://www.launchpadhq.io/platform/notion-ai): starting price Free, from $10; free tier yes; data as of 2026-10-26. Summarize and organize notes where your team already works. Best for turning transcripts into tasks. #### Hybrid: Balanced - Descript (https://www.launchpadhq.io/platform/descript): starting price Free, from $24/person/mo monthly or $16 annual; free tier yes; data as of 2026-10-14. Edit audio via transcript—great accuracy for podcasts. Hybrid when you need pro cleanup without full enterprise stack. - Google Docs. Free collaboration for cleaned-up text. Pair with any ASR export for a low-cost documentation loop. #### Budget: Free-first - Whisper (local). Open-source ASR you can run offline for free. Budget choice if you can handle setup and GPU/CPU time. - Obsidian. Free markdown vault for storing transcripts and notes. No subscription—ideal for privacy-first workflows. ### Content research Canonical: https://www.launchpadhq.io/workflows/content-research Data: https://www.launchpadhq.io/data/v1/workflows/content-research.json Gather sources, synthesize findings, and export briefs your team can trust. #### Optimal: Cited answers - Perplexity AI (https://www.launchpadhq.io/platform/perplexity-ai): starting price $20/month; free tier yes; data as of 2026-10-26. Web-grounded answers with citations. Optimal when research must be checkable and up to date. - Gemini Notebook (https://www.launchpadhq.io/platform/notebooklm): starting price Free; free tier yes; data as of 2026-10-20. Ground answers in your PDFs and docs. Best for deep synthesis after you’ve gathered sources. #### Hybrid: Balanced - Claude (https://www.launchpadhq.io/platform/claude): starting price Free, from $17/month; free tier yes; data as of 2026-10-26. Long documents and careful reasoning. Hybrid when you want strong analysis without a full research suite. - ChatGPT (https://www.launchpadhq.io/platform/chatgpt): starting price Free, from $8/month; free tier yes; data as of 2026-10-24. Fast iteration and summarization everyone knows. Balanced default for team research with lower friction. #### Budget: Free-first - Microsoft Copilot (https://www.launchpadhq.io/platform/microsoft-copilot): starting price $9.99/month; free tier yes; data as of 2026-10-26. Often bundled with tools you already pay for. Budget-friendly entry to AI in Word and Edge workflows. - Gemini (https://www.launchpadhq.io/platform/gemini): starting price Free, from $4.99/month; free tier yes; data as of 2026-10-26. Free tier for quick scans and drafts. Good when you need breadth without adding another subscription. ### Social media posts Canonical: https://www.launchpadhq.io/workflows/social-media-posts Data: https://www.launchpadhq.io/data/v1/workflows/social-media-posts.json Batch ideas, hooks, and platform-specific copy with a consistent voice. #### Optimal: Brand-safe scale - Jasper (https://www.launchpadhq.io/platform/jasper): starting price $59/month billed yearly or $69/month billed monthly; free tier no; data as of 2026-10-14. Campaign-focused copy with brand controls. Optimal when marketing needs repeatable, on‑voice output at scale. - Buffer. Schedule and analyze posts across channels. Pairs with Jasper for end-to-end social workflows. #### Hybrid: Balanced - ChatGPT (https://www.launchpadhq.io/platform/chatgpt): starting price Free, from $8/month; free tier yes; data as of 2026-10-24. Hooks and variants in minutes. Hybrid sweet spot: strong copy without a dedicated marketing AI suite. - Canva AI (Magic Studio) (https://www.launchpadhq.io/platform/canva-ai-magic-studio): starting price Free, from $12.99/mo; free tier yes; data as of 2026-10-26. Visuals and templates in one place. Balanced when posts need graphics plus copy without pro design tools. #### Budget: Free-first - Gemini (https://www.launchpadhq.io/platform/gemini): starting price Free, from $4.99/month; free tier yes; data as of 2026-10-26. Free drafting for captions and ideas. Best budget option when every dollar counts. - InVideo AI (https://www.launchpadhq.io/platform/invideo-ai): starting price $10 per seat/mo; free tier yes; data as of 2026-10-14. Template-based short video on a Launchpad profile—avoids non-listed editors while still fitting budget workflows. ### Meeting intelligence Canonical: https://www.launchpadhq.io/workflows/meeting-intelligence Data: https://www.launchpadhq.io/data/v1/workflows/meeting-intelligence.json Capture meetings, extract decisions, and sync tasks to your stack. #### Optimal: Bot-free notes plus automation — or Otter + Zapier Time: Live capture; about 10–20 minutes after the call to file actions Result: Searchable notes plus tasks in the tools you already use - Granola (https://www.launchpadhq.io/platform/granola): starting price Free, from $14; free tier yes; data as of 2026-10-14. Device-audio capture with no bot in the call. The paid Business plan unlocks unlimited history and the Zapier and Notion integrations. - Otter AI (https://www.launchpadhq.io/platform/otter-ai): starting price Free, from $16.99/user/mo billed monthly; free tier yes; data as of 2026-10-25. Accurate transcripts and search for teams when you want a meeting bot. Optimal when meetings are your system of record. - Zapier (https://www.launchpadhq.io/platform/zapier): starting price Free, from $19.99/mo billed annually; free tier yes; data as of 2026-10-25. Automate pushing notes into CRMs and PM tools. Best when workflows must be hands-off after each call. 1. Capture the meeting (Granola): Prefer Granola when you do not want a bot to join. The free Basic plan keeps a limited meeting history; upgrade to Business for unlimited history and Zapier. Alternatively use Otter if you need a notetaker bot. 2. Extract decisions (Granola): Ask the meeting chat for decisions, owners, and due dates. Edit anything that is wrong before it leaves the tool. Prompt: From this meeting, list: (1) decisions, (2) open questions, (3) action items as Owner — task — due date. Quote the transcript if you are unsure. Do not invent owners. 3. Route tasks (Zapier): On Granola Business, use the official Zapier integration, or a Zapier Zap from Otter/Fireflies if you captured there. Zapier's Free plan allows two-step Zaps only and a small monthly task allowance. #### Hybrid: Fireflies capture linked to Notion Time: Live capture; about 10–15 minutes to file the recap Result: A searchable recap sitting on the right Notion page - Fireflies.ai (https://www.launchpadhq.io/platform/fireflies-ai): starting price Free, from $18; free tier yes; data as of 2026-10-14. Searchable meeting memory and summaries. Hybrid when you need solid capture without full enterprise pricing. - Notion AI (https://www.launchpadhq.io/platform/notion-ai): starting price Free, from $10; free tier yes; data as of 2026-10-26. Link summaries to project pages your team already uses. Balanced ops: notes plus light AI assists. 1. Let Fireflies join (with consent) (Fireflies.ai): Connect calendar and tell attendees a notetaker will join. Stay on Free if volume is low; Pro adds video recording and more export/integrations (official billing guide). 2. Park the recap in Notion (Notion AI): Paste or sync the summary onto the project page. Use Notion AI only to format — you confirm owners. Notion’s own AI Meeting Notes require Business. Prompt: Turn this recap into a Notion page with: Summary, Decisions, Action items (Owner, due). Do not add attendees who were not listed: [PASTE] #### Budget: Fathom Free, or paste notes into Gemini Time: Live capture; about 10 minutes to clean the summary Result: A shared Doc with decisions and owners - Fathom (https://www.launchpadhq.io/platform/fathom): starting price Talk to sales; free tier yes; data as of 2026-10-14. Its Free plan covers recordings, transcription and instant AI summaries. - Gemini (https://www.launchpadhq.io/platform/gemini): starting price Free, from $4.99/month; free tier yes; data as of 2026-10-26. Summarize pasted notes on the free tier when you captured audio somewhere else. - Google Docs. Free shared docs for cleaned minutes. Combine with Fathom export or a manual transcript. 1. Record with Fathom Free (Fathom): Use Fathom’s Free plan for recording and transcription. Bot-free capture is listed as beta for Mac. Consent still applies. 2. Optional cleanup in Gemini (Gemini): If the auto-summary is messy, paste it into Gemini Free and ask for a decision log. Then paste into Google Docs for sharing. Prompt: Rewrite this meeting summary as: decisions, owners, due dates, open questions. Do not add people or dates that are not in the text: [PASTE] ### Customer support drafts Canonical: https://www.launchpadhq.io/workflows/customer-support Data: https://www.launchpadhq.io/data/v1/workflows/customer-support.json Draft replies, escalate edge cases, and keep a consistent support voice. #### Optimal: Integrated - Intercom. AI inside the inbox with macros and routing. Optimal when support volume needs automation plus human handoff. - Zendesk. Enterprise ticketing with AI add-ons. Best when SLAs, reporting, and scale are non‑negotiable. #### Hybrid: Balanced - ChatGPT (https://www.launchpadhq.io/platform/chatgpt): starting price Free, from $8/month; free tier yes; data as of 2026-10-24. Draft empathetic replies from bullet facts. Hybrid when you want strong language without a full helpdesk AI. - Gmail. Where replies actually send—free or workspace. Pair with ChatGPT for copy/paste review workflows. #### Budget: Free-first - Claude (https://www.launchpadhq.io/platform/claude): starting price Free, from $17/month; free tier yes; data as of 2026-10-26. Careful wording for sensitive tickets on free/cheap tiers. Budget pick when tone and safety matter most. - Help Scout. Simple shared inbox for small teams. Often affordable—good when you need structure without Zendesk complexity. ## Models Canonical: https://www.launchpadhq.io/models ### Claude (Anthropic) Canonical: https://www.launchpadhq.io/models/claude Data: https://www.launchpadhq.io/data/v1/models/claude.json Last verified: 2026-10-14 - On claude.ai, start with Sonnet for everyday writing, coding, and analysis. It is the default on the free plan. - Use Haiku when you want a fast, cheap answer and the job is simple (summaries, sorting, short rewrites). - Use Opus when the task is long and you need it to keep working without you steering every step. That needs a paid plan. - Use Fable only for the largest, longest jobs. Anthropic positions it above Opus; it uses more of your usage limit. - API developers: Anthropic currently recommends Opus 5.5 as the default, Haiku 5.5 for high-volume cheap work, and Fable 5.1 when Opus is not enough. - Claude Opus 5.5 (API id claude-opus-5-5): current, flagship Context: 1000000 tokens; max output: 128000 tokens API price (from the model registry): input $4 / 1M, output $20 / 1M Best for: Hard, multi-step work you would pay extra for: long coding sessions, research write-ups, and jobs that run for a while. Source: https://platform.claude.com/docs/en/models/opus-5-5/overview - Claude Sonnet 5.5 (API id claude-sonnet-5-5): current, balanced Context: 1000000 tokens; max output: 128000 tokens API price (from the model registry): input $2 / 1M, output $10 / 1M Best for: Day-to-day writing, coding, and analysis — the version most beginners should start with. Source: https://platform.claude.com/docs/en/models/sonnet-5-5/overview - Claude Haiku 5.5 (API id claude-haiku-5-5): current, budget Context: 1000000 tokens; max output: 128000 tokens API price (from the model registry): input $0.1 / 1M, output $0.5 / 1M Best for: Quick answers, sorting, extraction, and anything you run hundreds of times. Source: https://platform.claude.com/docs/en/models/haiku-5-5/overview - Claude Fable 5.1 (API id claude-fable-5-1): current, flagship Context: 1000000 tokens; max output: 128000 tokens API price (from the model registry): input $10 / 1M, output $50 / 1M Best for: The rare job where Opus is not enough — long autonomous work you are willing to pay more for. Source: https://platform.claude.com/docs/en/models/overview Catalog tools: - Claude (https://www.launchpadhq.io/platform/claude): starting price Free, from $17/month; free tier yes; data as of 2026-10-26 - Anthropic API (https://www.launchpadhq.io/platform/anthropic-api): starting price $0.10 input / $0.50 output per million tokens; free tier no; data as of 2026-10-26 ### GPT (ChatGPT / OpenAI) (OpenAI) Canonical: https://www.launchpadhq.io/models/gpt Data: https://www.launchpadhq.io/data/v1/models/gpt.json Last verified: 2026-10-24 - In the ChatGPT app, what you can pick depends on the plan. OpenAI’s help center: GPT-6 Pro (powered by GPT-6 Astra) is on Pro, Business, and Enterprise; Plus gets Astra in ChatGPT Work and Codex, not as GPT-6 Pro in regular Chat. Free and Go do not get Astra. - For API work, OpenAI currently tells developers to start with GPT-6 Astra for hard jobs, GPT-6.1 Sol when you want similar quality for less money, and GPT-6 Luna for cheap high-volume work. - GPT-5.5 is the previous flagship. Use it only if your API work already depends on it. - GPT-6 Astra (API id gpt-6-astra): current, flagship Context: 1050000 tokens; max output: 128000 tokens API price (from the model registry): input $10 / 1M, output $50 / 1M Best for: Paid ChatGPT users and API developers with the hardest work — research, long coding, and multi-step computer use. Source: https://developers.openai.com/api/docs/models/gpt-6-astra - GPT-6.1 Sol (API id gpt-6.1-sol): current, balanced Context: 1050000 tokens; max output: 128000 tokens API price (from the model registry): input $2 / 1M, output $10 / 1M Best for: API and ChatGPT Work/Codex users who want strong results without Astra’s price. Source: https://developers.openai.com/api/docs/models/gpt-6.1-sol - GPT-6 Sol (API id gpt-6-sol): current, prior Context: 1050000 tokens; max output: 128000 tokens API price (from the model registry): input $2 / 1M, output $10 / 1M Best for: Existing API work built on GPT-6 Sol; new work should start on GPT-6.1 Sol. Source: https://developers.openai.com/api/docs/models/gpt-6-sol - GPT-6 Luna (API id gpt-6-luna): current, budget Context: 1050000 tokens; max output: 128000 tokens API price (from the model registry): input $0.1 / 1M, output $0.5 / 1M Best for: Lots of short, cheap jobs — classification, drafts, and bulk API calls. Source: https://developers.openai.com/api/docs/models/gpt-6-luna - GPT-5.5 (API id gpt-5.5): legacy, prior Context: 1050000 tokens; max output: 128000 tokens API price (from the model registry): input $5 / 1M, output $30 / 1M Best for: Existing API integrations that are not ready to move to GPT-6. Source: https://developers.openai.com/api/docs/models/gpt-5.5 Catalog tools: - ChatGPT (https://www.launchpadhq.io/platform/chatgpt): starting price Free, from $8/month; free tier yes; data as of 2026-10-24 - OpenAI API (https://www.launchpadhq.io/platform/openai-api): starting price $2.50/$10 per 1M tokens); free tier yes; data as of 2026-10-26 ### Gemini (Google) Canonical: https://www.launchpadhq.io/models/gemini Data: https://www.launchpadhq.io/data/v1/models/gemini.json Last verified: 2026-10-24 - Exact Gemini-app picker for 3.8 Flash is unverified on Google's plan pages. - For the Gemini API, use 3.8 Flash for most production work (it is on the API free tier). Use 4 Argon only when you need the flagship and can pay (not on the API free tier). - 3.6 Flash is the previous Flash generation at the same listed paid price as 3.8 Flash through 2026 — Google still documents it as previous-generation. - Gemini 4 Argon (API id gemini-4-argon): current, flagship Context: 1048576 tokens; max output: 65536 tokens API price (from the model registry): input $2 / 1M, output $10 / 1M Best for: Hard coding, math, science, and long agent jobs. Source: https://ai.google.dev/gemini-api/docs/models/gemini-4-argon - Gemini 3.8 Flash (API id gemini-3.8-flash): current, balanced Context: 1048576 tokens; max output: 65536 tokens API price (from the model registry): input $0.75 / 1M, output $3.75 / 1M Best for: Most Gemini API projects, and everyday Gemini use if your app plan includes it. Source: https://ai.google.dev/gemini-api/docs/models/gemini-3.8-flash - Gemini 3.6 Flash (API id gemini-3.6-flash): legacy, prior Context: unverified; max output: unverified API price (from the model registry): input $0.75 / 1M, output $3.75 / 1M Best for: Existing apps that already call gemini-3.6-flash. Source: https://ai.google.dev/gemini-api/docs/models Catalog tools: - Gemini (https://www.launchpadhq.io/platform/gemini): starting price Free, from $4.99/month; free tier yes; data as of 2026-10-26 ### Grok (xAI) Canonical: https://www.launchpadhq.io/models/grok Data: https://www.launchpadhq.io/data/v1/models/grok.json Last verified: 2026-10-24 - xAI’s models page: for code and chat, use Grok 4.7. It is described as the most capable model they have built. - Grok 4.6 is still listed at the same token price as 4.7 — treat it as the previous 4.x snapshot, not the one xAI tells you to pick. - Grok 4.7 (API id grok-4.7): current, flagship Context: 500000 tokens; max output: unverified API price (from the model registry): input $2 / 1M, output $6 / 1M Best for: Grok users and API callers who want the current flagship for coding and chat. Source: https://docs.x.ai/developers/models/grok-4.7 - Grok 4.3 (API id grok-4.3): current, budget Context: 1000000 tokens; max output: unverified API price (from the model registry): input $1.25 / 1M, output $2.5 / 1M Best for: Longer documents on a budget, when you do not need 4.7’s flagship quality. Source: https://docs.x.ai/developers/models - Grok 4.6 (API id grok-4.6): legacy, prior Context: 500000 tokens; max output: unverified API price (from the model registry): input $2 / 1M, output $6 / 1M Best for: Existing callers that have not moved to 4.7. Source: https://docs.x.ai/developers/models Catalog tools: - Grok (https://www.launchpadhq.io/platform/grok): starting price Free, from $8/month; free tier yes; data as of 2026-10-14 ### Mistral (Vibe) (Mistral AI) Canonical: https://www.launchpadhq.io/models/mistral Data: https://www.launchpadhq.io/data/v1/models/mistral.json Last verified: 2026-10-14 - Which of these names appear in free vs Pro Vibe was not confirmed on docs.mistral.ai in this pass. - Mistral Large 4 (API id mistral-large-4): current, flagship Context: 1000000 tokens; max output: unverified API price (from the model registry): input $0.68 / 1M, output $2.09 / 1M Best for: People who want a strong European-lab model, especially via the API or open weights. Source: https://docs.mistral.ai/models/mistral-large - Mistral Small 4 (API id mistral-small-4): current, budget Context: unverified; max output: unverified API price (from the model registry): input $0.15 / 1M, output $0.6 / 1M Best for: High-volume, cheaper Mistral API calls. Source: https://docs.mistral.ai/inference/pricing - Mistral Large 3 (API id mistral-large-3): legacy, prior Context: unverified; max output: unverified API price (from the model registry): input $0.5 / 1M, output $1.5 / 1M Best for: Existing Large 3 API callers. Source: https://docs.mistral.ai/inference/pricing Catalog tools: - Mistral Vibe (https://www.launchpadhq.io/platform/mistral-vibe): starting price Free, from $14.99/month; free tier yes; data as of 2026-10-25 ### Llama (Meta) (Meta) Canonical: https://www.launchpadhq.io/models/llama Data: https://www.launchpadhq.io/data/v1/models/llama.json Last verified: 2026-10-14 - There is no Platform.slug in this repo’s tutorials or seed scripts for Meta’s Llama product (llama-cpp is a separate local engine). This family is documented for /models only. - Llama 4 Scout: long context (10M tokens on the model card) and can run on a single H100 when quantized. - Llama 4 Maverick: larger expert count (400B total / 17B active) and a 1M context window. - Meta does not list a first-party per-1M-token API price on the pages checked — you download weights or pay a host. API prices are unverified here. - Llama 4 is still Meta's newest Llama-branded release, but llama.com now leads with Meta's newer Muse models (checked 2026-10-24). See the Muse family for Muse Spark Open and Muse Glimmer 30B. - Llama 4 Maverick: current, flagship Context: 1000000 tokens; max output: unverified API price (from the model registry): input Unverified, output Unverified Best for: Teams that can host Llama themselves and want the stronger Llama 4. Source: https://dev.meta.ai/llama/docs/model-cards-and-prompt-formats/llama4 - Llama 4 Scout: current, budget Context: 10000000 tokens; max output: unverified API price (from the model registry): input Unverified, output Unverified Best for: Very long documents on hardware you control, when you do not need Maverick. Source: https://dev.meta.ai/llama/docs/model-cards-and-prompt-formats/llama4 ### Muse (Meta) (Meta) Canonical: https://www.launchpadhq.io/models/muse-spark Data: https://www.launchpadhq.io/data/v1/models/muse-spark.json Last verified: 2026-10-24 - Meta's open models are now the Muse family; llama.com leads with Muse. Muse Spark is the model named in the consumer Meta AI assistant, and Muse Spark Open is its downloadable open-weight release. We don't treat them as the same checkpoint. - Pick Muse Spark Open for a frontier open-weight model if you have data-center hardware or use a hosted provider. Pick Muse Glimmer 30B if you want one that runs on a single high-end consumer GPU. - The licenses differ: Muse Spark Open uses Meta's Muse Model License, while Muse Glimmer 30B is Apache 2.0. - Muse Spark: current, flagship Context: unverified; max output: unverified API price (from the model registry): input Unverified, output Unverified Best for: People using the consumer Meta AI assistant, not self-hosting weights. Source: https://ai.meta.com/meta-ai/assistant/ - Muse Spark Open: current, balanced Context: 1048576 tokens; max output: unverified API price (from the model registry): input Unverified, output Unverified Best for: Teams that want Meta's downloadable open-weight Muse Spark release or a third-party hosted API — not the consumer assistant app. Source: https://dev.meta.ai/docs/muse-spark-open - Muse Glimmer 30B: current, budget Context: 131072 tokens; max output: unverified API price (from the model registry): input Unverified, output Unverified Best for: People who want a capable Meta model running privately on their own computer. Source: https://huggingface.co/meta-models/Muse-Glimmer-30B ### DeepSeek (DeepSeek) Canonical: https://www.launchpadhq.io/models/deepseek Data: https://www.launchpadhq.io/data/v1/models/deepseek.json Last verified: 2026-10-17 - API: use deepseek-flash (DeepSeek-V4.1-Flash-1017) for most work. It is cheaper, has vision, and is the name DeepSeek currently documents as the Flash model. - Use deepseek-v4-pro when you want the Pro checkpoint. DeepSeek says it will keep serving V4 Pro after 14 Sep 2026 with the same billing until further notice. - Consumer chat-app daily limits are unverified on the API docs. Hosted DeepSeek processes data in China — see the catalog privacy note. - DeepSeek V4.1 Flash (API id deepseek-flash): current, budget Context: 1000000 tokens; max output: 384000 tokens API price (from the model registry): input $0.15 / 1M, output $0.6 / 1M Best for: Budget API use and anyone comparing open-weight / China-hosted options, after reading the privacy note. Source: https://api-docs.deepseek.com/quick_start/pricing - DeepSeek V4 Pro (API id deepseek-v4-pro): current, flagship Context: 1000000 tokens; max output: 384000 tokens API price (from the model registry): input $0.66 / 1M, output $1.98 / 1M Best for: API jobs where you specifically want the Pro checkpoint. Source: https://api-docs.deepseek.com/quick_start/pricing Catalog tools: - DeepSeek (https://www.launchpadhq.io/platform/deepseek): starting price Free, from $0.30/$1.20 per 1M tokens in/out; free tier yes; data as of 2026-10-24 ### Qwen (Alibaba (Qwen / Tongyi)) Canonical: https://www.launchpadhq.io/models/qwen Data: https://www.launchpadhq.io/data/v1/models/qwen.json Last verified: 2026-10-14 - Qwen Chat being free with no subscription is stated in our catalog seed; exact chat limits remain unverified on Model Studio docs. - Qwen3.8-Max (API id qwen3.8-max): current, flagship Context: 1000000 tokens; max output: 131072 tokens API price (from the model registry): input $2 / 1M, output $6 / 1M Best for: Hard Qwen API jobs — long coding and office-style work — after you pick a region. Source: https://www.alibabacloud.com/help/en/model-studio/qwen3-8-max - Qwen3.8-Flash (API id qwen3.8-flash): current, budget Context: 1000000 tokens; max output: 131072 tokens API price (from the model registry): input $0.15 / 1M, output $0.47 / 1M Best for: Everyday Qwen API use and high-volume jobs. Source: https://www.alibabacloud.com/help/en/model-studio/qwen3-8-flash - Qwen3.7-Max (API id qwen3.7-max): legacy, prior Context: 1000000 tokens; max output: unverified API price (from the model registry): input $2.5 / 1M, output $7.5 / 1M Best for: Existing 3.7-Max callers. Source: https://www.alibabacloud.com/help/en/model-studio/model-pricing Catalog tools: - Qwen (https://www.launchpadhq.io/platform/qwen): starting price $0.05/$0.40 per 1M tokens; free tier yes; data as of 2026-10-14 ### Claude vs GPT models Canonical: https://www.launchpadhq.io/models/compare/claude-vs-gpt Data: https://www.launchpadhq.io/data/v1/models/compare/claude-vs-gpt.json Flagship (API default / hardest work): - Claude Opus 5.5 (API id claude-opus-5-5): current, flagship Context: 1000000 tokens; max output: 128000 tokens API price (from the model registry): input $4 / 1M, output $20 / 1M Best for: Hard, multi-step work you would pay extra for: long coding sessions, research write-ups, and jobs that run for a while. Source: https://platform.claude.com/docs/en/models/opus-5-5/overview - GPT-6 Astra (API id gpt-6-astra): current, flagship Context: 1050000 tokens; max output: 128000 tokens API price (from the model registry): input $10 / 1M, output $50 / 1M Best for: Paid ChatGPT users and API developers with the hardest work — research, long coding, and multi-step computer use. Source: https://developers.openai.com/api/docs/models/gpt-6-astra Budget (cheap, fast, high volume): - Claude Haiku 5.5 (API id claude-haiku-5-5): current, budget Context: 1000000 tokens; max output: 128000 tokens API price (from the model registry): input $0.1 / 1M, output $0.5 / 1M Best for: Quick answers, sorting, extraction, and anything you run hundreds of times. Source: https://platform.claude.com/docs/en/models/haiku-5-5/overview - GPT-6 Luna (API id gpt-6-luna): current, budget Context: 1050000 tokens; max output: 128000 tokens API price (from the model registry): input $0.1 / 1M, output $0.5 / 1M Best for: Lots of short, cheap jobs — classification, drafts, and bulk API calls. Source: https://developers.openai.com/api/docs/models/gpt-6-luna