Repomix-rs Explorer

Use repomix (the repomix-rs CLI) to pack a codebase into a single AI-friendly file, then analyze it with agent tools.

When to use

• Explore an unfamiliar local or remote repository
• Summarize structure, entry points, or module boundaries
• Find patterns (auth, routes, models, error handling) across many files
• Get token/file metrics before deciding what to read in detail

Use repomix-context-skill instead when a MindMesh index already exists at .mind-mesh/agent/repomix.md.

Installation

npm (recommended)

Package name is repomix-rs; the installed command is repomix.

# Global install — `repomix` available everywhere
npm install -g repomix-rs

# One-off run without global install
npx repomix-rs .

# Run the MCP server on stdio
npx -y repomix-rs --mcp

Supported platforms: Linux (x64, arm64), macOS (x64, arm64), Windows (x64). Node >= 18 required.

If both the original TypeScript repomix and repomix-rs are installed globally, the last install wins the repomix command. Use npx repomix-rs or npx repomix explicitly if both are present.

From source

# Install to ~/.cargo/bin/repomix
cargo install --path crates/cli

# Or build a release binary at ./target/release/repomix
cargo build --release

Prerequisites

Git-related features (--remote, --include-diffs, --include-logs, sort-by-changes, MCP pack_remote_repository) shell out to the git executable on PATH. If git is missing, these features are skipped with a warning rather than failing the pack.

Core commands

Default output file is style-dependent (repomix-output.xml, .md, .json, or .txt). Always use --output to place the file where you expect.

Agent workflow

1. Pack

   • Local: repomix [ROOT]
   • Remote: repomix --remote <URL> --output /tmp/<repo>.xml
   • Note the metrics printed by the command (files, characters, tokens, output path).

2. Inspect structure

   • Read the start of the output file for the file tree and metrics summary.
   • For large outputs, read with offset/limit rather than loading the whole file.

3. Search patterns

   • Use the agent's Grep tool on the output file, not shell grep.
   • Common patterns: export.*function, export.*class, import.*from, router\.|route\.|endpoint, auth|login|jwt, model|schema|database, error|exception|try.*catch.

4. Read slices

   • After locating a relevant ### path or <file path="..."> boundary, read that slice with offset/limit.

5. Summarize

   • Report metrics, top-level structure, and findings.
   • Cite output file paths and line numbers.
   • Suggest next steps for deeper exploration or targeted edits.

Best practices

• Remote repos: always write to /tmp or a temp directory to avoid polluting the current workspace.
• Large repos: use --compress to extract signatures via tree-sitter and reduce token count.
• Narrow scope: use --include / --ignore before reading large outputs.
• Output path: explicitly pass --output so you know exactly where the pack lives.
• Clean up: delete large /tmp output files after analysis unless the user asks to keep them.
• Security: suspicious files are detected and excluded automatically; trust the exclusion list in the output report.

Do not

• Use this skill for targeted file edits — pack, analyze, then switch to direct file tools.
• Read entire multi-megabyte output files into context; grep first, then read slices.
• Run rm on project files unless you are certain they are temporary output files you created.
• Confuse repomix-rs (Rust, npm package repomix-rs) with the original TypeScript repomix (npm package repomix).

Error handling

Resources

• Run repomix --help for the complete flag reference.
• Repository: https://github.com/sopaco/repomix-rs
• npm package: https://www.npmjs.com/package/repomix-rs