Your code's got something to say!
A few commands that make your repo talk back
Ask questions about your codebase in plain English and get instant answers with actual code snippets. The REPL stays open so you can keep asking follow-ups without losing context. It's like pair programming with someone who's read your entire repo.
Generate context files for Cursor, Windsurf, Cline, Continue, and Aider all at once. Your AI tools instantly understand your architecture, patterns, and dependencies without you explaining anything. One command, five tools, zero manual setup.
Create a professional README.md directly from your actual code. Detects your tech stack, finds entry points, extracts features. No more outdated docs that lie about what your code actually does.
Auto-update context files as you code. Run it in the background and your AI assistants always have fresh knowledge of your project. Change a file, context updates automatically. No stale info, ever.
See who owns what code with commit counts, lines changed, and file ownership stats. Track recent changes across any time period or get a quick repo snapshot. Perfect for code reviews and catching up after time off.
Use OpenAI, Claude, Gemini, Grok, DeepSeek, or Mistral for answers. Switch providers on the fly to compare responses. All embeddings stay 100% local, only your questions hit the API.
Everything you need to know about QuackStack
No! QuackStack generates embeddings entirely locally on your machine. Your code never leaves your computer during indexing. Only your natural language queries (like 'where is auth handled?') and small relevant code snippets are sent to the AI provider for generating conversational answers.
You need: (1) A PostgreSQL database - free options like Neon or Supabase work great, and (2) ONE API key for conversational answers (OpenAI, Claude, Gemini, DeepSeek, or Mistral). Gemini has a free tier! Embeddings are generated locally with no API calls.
First-time indexing depends on your codebase size. A typical project (1000-5000 files) takes 2-5 minutes. After that, only changed files are re-indexed, making updates nearly instant. Large monorepos may take 10-15 minutes initially.
Copilot and Cursor suggest code based on immediate context. QuackStack gives them FULL codebase understanding. It's complementary - QuackStack generates the .cursorrules file that makes Cursor smarter about YOUR specific project, patterns, and architecture.
Yes! Each project gets its own isolated namespace in the database (based on directory name). Just run 'quack' in any project folder and it automatically manages separate indexes. Switch between projects freely without conflicts.
QuackStack itself is 100% free and open source. Your only costs are: (1) AI provider API usage for conversational answers (Gemini has a free tier!), and (2) database hosting (Neon/Supabase offer free tiers). Most developers spend $0-5/month depending on usage.