Our take
- Activity
- 79
- Community
- 57
- Issues
- 81
- Pull requests
- 83
+419 stars in the last 30 days
- Merges more pull requests than 95% of the projects we track
- Closes issues faster than 81% of the projects we track
- Releases more often than 75% of the projects we track
Project history
Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.
Stars
29,271Forks
1,645Change since Nov 14, 2026: +8 stars·+1 forks
Overview
Repomix packages a repository into one structured file so an AI assistant can consider code across files without requiring you to paste them individually. It is designed for developers preparing a codebase for review, explanation, refactoring, or other LLM-assisted work.
Use it when an assistant needs repository-wide context, or when you want to inspect a remote GitHub project without cloning it yourself. The CLI supports local and remote repositories, while the website offers a browser-based option. See the repository README for the full documentation.
Key Features
- Packs local or remote repositories into a single output file.
- Outputs XML, Markdown, JSON, or plain text.
- Selects and excludes files with glob patterns, and respects
.gitignore,.ignore, and.repomixignore. - Reports token counts and can enforce a token budget.
- Offers Tree-sitter-based compression to retain code structure while reducing detail. The README describes this feature as experimental.
- Scans for files matching known credential formats and excludes them from output.
- Provides an MCP server for AI assistant integrations, with an optional workspace sandbox.
- Can include Git history and diffs, split large outputs, and watch local files for changes.
Use Cases
- A developer can package a project before asking an LLM for a code review or refactoring suggestions.
- A maintainer can prepare selected source files and documentation for generating tests or project documentation.
- An engineer can inspect a public GitHub repository remotely without manually cloning and assembling its files.
- An AI-tool user can connect Repomix through MCP to package or search codebase output as part of an assistant workflow.
Getting Started
Run Repomix in a project directory with npx repomix@latest. It writes repomix-output.xml by default. For installation options, configuration, and other workflows, see the README or project website.
Alternatives
- GitNexus: GitNexus indexes code relationships for MCP queries and impact analysis, rather than bundling selected repository files into a prompt.
- codegraph: CodeGraph provides graph-based retrieval of source and call paths through MCP, rather than producing a consolidated file bundle.
- graphify: Graphify builds a queryable graph from code and other project files, rather than formatting repository contents as LLM-ready output.
| Project | Language | License | Stars | Status |
|---|---|---|---|---|
| repomix | TypeScript | MIT | 28.7k | Not checked yet |
| GitNexus | TypeScript | Other | 47.7k | Active |
| codegraph | C | MIT | 73.1k | Active |
| graphify | Python | Apache-2.0 | 124k | Active |
Considerations
- The default output may contain a large amount of code. File filters, token-count tools, compression, and output splitting help tailor it to an AI tool's context or size limits.
- Compression is described as experimental, so check whether its reduced-detail output suits the task.
- Secret scanning is a safeguard based on known credential formats, not a guarantee that all sensitive information will be detected. Review output before sharing it.
- Remote repository configuration is not trusted by default. The README notes that enabling remote config trust can load configuration from that repository, so use it only when appropriate.
Found this useful?
Share it with someone who would like repomix.