Repomix — Pack Any Repo into One AI-Ready File

What it is

Repomix is a developer tool that packs an entire repository into a single, AI-optimized file. When you need to feed a codebase to an LLM for analysis, code review, or documentation generation, Repomix creates a structured output containing all relevant files with token counts, file metadata, and security scanning.

Repomix is designed for developers working with AI coding assistants who need to provide full project context. It supports Markdown, XML, and plain text output formats, and can process both local repositories and remote GitHub repos.

How it saves time or tokens

Manually copying files into an LLM prompt is tedious and error-prone. Repomix automates this by intelligently selecting files, respecting .gitignore patterns, counting tokens, and scanning for secrets before output. The structured format maximizes LLM comprehension while minimizing wasted tokens on irrelevant files. You can also filter by file type or directory to focus the context on what matters.

How to use

1. Run directly with npx: npx repomix@latest in your project directory.
2. Or install globally: npm install -g repomix.
3. For remote repos: npx repomix --remote https://github.com/user/repo.
4. Choose output format: npx repomix --output repomix-output.md --style markdown.

Example

# Pack current directory
npx repomix@latest

# Pack a remote GitHub repo
npx repomix --remote https://github.com/user/repo

# Output as Markdown with token count
npx repomix --output repomix-output.md --style markdown

# Filter specific directories
npx repomix --include 'src/**' --output context.md

# Exclude test files
npx repomix --ignore 'test/**,*.test.ts'

Related on TokRepo

• AI tools for coding -- explore tools that enhance AI-assisted development workflows.
• Automation tools -- discover CLI tools for developer productivity.

Common pitfalls

• Large repositories may produce output that exceeds LLM context windows. Use --include filters to focus on relevant directories.
• Repomix scans for secrets but is not a replacement for dedicated secret scanning tools. Always review output before sharing with external LLM services.
• Binary files (images, compiled assets) are skipped by default. If you need to reference them, mention them in your prompt separately.