Repomix

Overview

Repomix is a tool that transforms an entire repository into a single AI-friendly file. It formats the codebase for AI comprehension and can produce XML, Markdown, JSON, or plain text output. It is designed for sharing full repository context with large language models and other AI tools.

Key Features

• AI-optimized output formats codebase content for easier AI understanding and processing.
• Token counting reports token counts for each file and the entire repository.
• Git-aware behavior automatically respects .gitignore, .ignore, and .repomixignore files.
• Security-focused scanning uses Secretlint to detect files matching known credential formats and leave them out of output.
• Code compression uses Tree-sitter to extract key code elements and reduce token usage.
• Customization supports include and ignore patterns, output styles, persistent configuration, and remote repository packing.

Use Cases

• Code review and refactoring with AI assistants using the packed repository as context.
• Implementation planning, bug investigation, and third-party library security checks.
• Documentation generation and test case generation from the whole codebase.
• Creating LLM-compatible knowledge datapacks from written content such as blogs, documentation, and books.

Getting Started

• Run npx repomix@latest in a project directory to try Repomix without installation.
• Install globally with npm install -g repomix, yarn global add repomix, bun add -g repomix, or Homebrew, then run repomix.
• Repomix generates repomix-output.xml in the current directory containing the repository in an AI-friendly format.
• Run repomix --init to create a repomix.config.json file for persistent settings.
• Use repomix --remote with a repository shorthand or URL to pack a remote repository.

Deployment & Requirements

• Repomix can be run through npx, a global npm, yarn, or bun installation, Homebrew, or Docker.
• Container use is supported with the ghcr.io/yamadashy/repomix image.
• The CLI can access private repositories because it uses the locally installed git.

Before You Adopt

• License: MIT. Review its terms before using, modifying, or distributing the project.
• Credential-file exclusion depends on Secretlint matching known credential formats.
• Enabling output escaping according to the chosen style schema can increase token count.