MCP GitHub Reader
@skydeckai
Bring github repos into context using this mcp server
Overview
What is MCP GitHub Reader?
MCP GitHub Reader is a lightweight Model Context Protocol (MCP) server that brings GitHub repositories into context for large language models. It works directly with GitHub’s API, requires no local cloning, and is compatible with any LLM client supporting MCP.
How to use MCP GitHub Reader?
Install globally (npm install -g mcp-github-reader) or locally. Run as a standalone server with mcp-github-reader (global) or npx mcp-github-reader (local). For Claude Desktop, add a JSON entry with the appropriate command and arguments.
Key features of MCP GitHub Reader
- API-based: no local repository cloning required
- Repository analysis with statistics and language breakdown
- File access: retrieve individual file contents or entire repo structures
- Smart filtering with glob and regex patterns
- Cache-enabled to help avoid GitHub API limits
- Search capabilities for files matching specific patterns
- Ready-to-use prompt templates for common analysis tasks
Use cases of MCP GitHub Reader
- Analyze a public repository’s structure and purpose
- Retrieve the content of a specific file without cloning
- Search for files matching a pattern within a repository
- Get repository statistics and language breakdown
- Filter repository contents to include or exclude certain files
FAQ from MCP GitHub Reader
What is MCP GitHub Reader?
It is an MCP server that provides tools and prompts for LLMs to access and analyze public GitHub repositories via the GitHub API.
Can I use MCP GitHub Reader with private repositories?
No, currently it only works with public GitHub repositories and does not support authentication or private repos.
What are the rate limits?
Unauthenticated requests are subject to GitHub’s default rate limit of 60 requests per hour.
Do I need to install extra software?
You need Node.js and npm. The server is installed via npm and runs as a standalone process.
Does MCP GitHub Reader clone repositories?
No, it works entirely through the GitHub API and does not perform any local cloning.
