GitMCP is a remote Model Context Protocol (MCP) server that turns any public GitHub repository or GitHub Pages site into a documentation source for AI coding assistants. You replace github.com with gitmcp.io in a repository URL (or use {owner}.gitmcp.io/{repo} for a Pages site), add that URL as an MCP server in your assistant, and the assistant can fetch and search the project's current documentation and code instead of relying on what the model was trained on.
Two flavors are available. A repository-specific server (gitmcp.io/{owner}/{repo}) pins the assistant to one project; the generic server (gitmcp.io/docs) lets the assistant choose the target repository per request. Documentation is resolved in priority order: llms.txt, an AI-optimized version of the project's docs, then the README.md and root. The service runs in the cloud with no download, signup, or repository changes, an in-browser chat lets you query a repository's documentation directly, and the project can be self-hosted.
GitMCP requires no authentication, does not store queries or personal information, and does not scrape repositories: it accesses public content only when a user queries it and honors robots.txt on GitHub Pages sites. Configuration snippets are provided for Cursor, Claude Desktop, Windsurf, VS Code, Cline, Highlight AI, Augment Code, and Msty AI.
Features
- URL-based setup: swap the domain in any GitHub or GitHub Pages URL to get an MCP server address
- Repository-specific or generic: pin one project, or use gitmcp.io/docs to let the assistant pick a repository on the fly
- fetch documentation: retrieves the primary documentation, such as llms.txt, for an overview of the project
- search documentation: searches the documentation for a query rather than loading all of it into context
- search code: searches the repository's code through GitHub code search for implementation examples
- fetchurlcontent: retrieves links referenced in the documentation and converts them to a readable format
- Documentation priority: llms.txt, then AI-optimized docs, then README.md and root files
- Embedded chat: an in-browser chat against a repository's documentation
- Client support: any MCP-capable assistant, with SSE URL or npx mcp-remote configuration
- Privacy and self-hosting: no auth, no stored queries, robots.txt respected, and deployable in your own environment
- Badge: an optional badge in the README.md file that counts tool calls against the repository
