Repomix
✓ Editorially verifiedPacks an entire codebase into a single AI-friendly file so LLMs can actually read your repo.
Pick Repomix if you regularly hand whole projects to Claude or ChatGPT and want one clean, token-counted bundle instead of copy-pasting files.
Skip it if your workflow is already inside an IDE agent like Cursor or Claude Code that indexes the repo for you.
Repomix is an open-source utility that flattens a whole repository into one structured file (XML, Markdown, JSON, or plain text) optimized for ingestion by ChatGPT, Claude, Gemini, and other LLMs. Instead of pasting files one at a time or wiring up a custom RAG pipeline just to ask 'how does auth work in this project,' you run `npx repomix@latest` and hand the resulting bundle to the model. It respects .gitignore, counts tokens per file, runs Secretlint to strip secrets, and can compress code with Tree-sitter for roughly 70% token reduction.
It's aimed squarely at developers who use AI assistants for code review, refactoring, bug hunts, and onboarding into unfamiliar codebases. Pricing is the easiest part of the review: Repomix is MIT-licensed and completely free, distributed as a CLI, Node library, Docker image, and Chrome extension. There's also a hosted web UI at repomix.com for one-off packs of public repos.
It's not an AI product in the strict sense - there's no model behind it - but it sits in the daily loop of anyone pairing with a frontier LLM on a real codebase. The main caveats are the usual ones for context-packing tools: very large monorepos still blow past context windows even after compression, and you need to actually read the security-scan output rather than trusting it blindly before pasting code into a third-party model.
Repomix is one of those small, sharp tools that quietly becomes part of your daily AI workflow. It doesn't try to be clever - it just produces the single artifact every frontier model actually wants. The Tree-sitter compression and Secretlint pass are what elevate it above the dozen weekend-project clones on GitHub.
— The AI Tool Bible editorial team
Pros
- ✅ Free, MIT-licensed, and runs locally with no account required
- ✅ Tree-sitter compression cuts tokens by ~70% on most codebases
- ✅ Secretlint pass strips API keys and secrets before you paste
- ✅ Works as CLI, Node lib, Docker image, Chrome extension, and hosted UI
- ✅ Respects .gitignore and supports remote repo URLs out of the box
Cons
- ⚠️ Not an AI itself - you still need a capable LLM to do the actual work
- ⚠️ Huge monorepos can exceed context windows even after compression
- ⚠️ Hosted web UI is convenient but means uploading code to a third party
Use cases
Explore related
Compare with similar tools
All in Coding →Cursor
FeaturedAI-first VS Code fork — chat, edit, and agentic coding in one IDE.
GitHub Copilot
FeaturedThe original AI pair programmer, now with chat and agents.
Replit Agent
FeaturedBuild & deploy a full app from a single prompt.
Aider
Terminal-based AI pair programmer that writes commits.
Codeium
Free, fast AI autocomplete + chat across 70+ editors.
Cody
Sourcegraph's AI coding assistant — codebase-aware via their search index.