GitHub Codemunch
A token-efficient MCP server for exploring GitHub repositories using tree-sitter AST parsing to extract symbol signatures and summaries. It enables users to search and retrieve specific code components while significantly reducing token consumption c
What is GitHub Codemunch?
GitHub Codemunch is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to token-efficient mcp server for exploring github repositories using tree-sitter ast parsing to extract symbol signatures and summaries. it enables users to search and retrieve specific code components ...
A token-efficient MCP server for exploring GitHub repositories using tree-sitter AST parsing to extract symbol signatures and summaries. It enables users to search and retrieve specific code components while significantly reducing token consumption c
This server falls under the Version Control and Developer Tools categories on MCPgee, the world's largest MCP server directory with 33,000+ servers.
Features
- A token-efficient MCP server for exploring GitHub repositori
Use Cases
Maintainer
Works with
Installation
PIP
pip install jcodemunch-mcpManual Installation
pip install jcodemunch-mcpConfiguration
Configuration Details
claude_desktop_config.json
Performance
Response Metrics
Resource Usage
How to Set Up and Use GitHub Codemunch
jCodeMunch MCP is a token-efficient code intelligence server that indexes GitHub and local repositories using tree-sitter AST parsing, exposing 60+ structured retrieval tools to Claude and other MCP clients. Instead of reading entire files, it lets you extract symbol signatures, call hierarchies, dependency graphs, and blast-radius analyses — achieving up to 95% token savings compared to full-file reads. Developers use it to navigate large codebases, plan safe refactors, find dead code, and understand architectural layers without overwhelming the context window.
Prerequisites
- Python 3.9+ installed (check with `python --version`)
- pip or uv package manager available
- An MCP-compatible client such as Claude Desktop, Claude Code, Cursor, or Windsurf
- Optional: ANTHROPIC_API_KEY, OPENAI_API_KEY, or GOOGLE_API_KEY if you want AI-generated code summaries
Install jcodemunch-mcp
Install the package from PyPI using pip or uv. The uv tool approach is recommended for isolated environments.
pip install jcodemunch-mcp
# or with uv:
uv tool install jcodemunch-mcpInitialize the index store
Run the init command to create the default index directory (~/.code-index/) and generate a global config file at ~/.code-index/config.jsonc.
jcodemunch-mcp initAdd to Claude Code (quickest method)
Register the server with Claude Code using the built-in mcp add command. This writes the server config for you.
claude mcp add -s user jcodemunch uvx jcodemunch-mcpConfigure Claude Desktop manually
Alternatively, add the server block to your Claude Desktop configuration file at ~/Library/Application Support/Claude/claude_desktop_config.json (macOS).
{
"mcpServers": {
"jcodemunch": {
"command": "uvx",
"args": ["jcodemunch-mcp"],
"env": {
"CODE_INDEX_PATH": "~/.code-index/",
"JCODEMUNCH_USE_AI_SUMMARIES": "true",
"ANTHROPIC_API_KEY": "sk-ant-YOUR_KEY_HERE"
}
}
}
}Index a repository
Point jCodeMunch at a local clone of the repository you want to explore. Create a project-level config file in the repo root to scope the index.
# In your project root, create .jcodemunch.jsonc
{
"include": ["src/**", "lib/**"],
"exclude": ["node_modules", ".git", "dist"]
}Verify the server is connected
Restart your MCP client and check that jcodemunch tools appear. You should see tools like search_symbols, get_repo_map, and get_blast_radius available.
GitHub Codemunch Examples
Client configuration
Claude Desktop config with optional AI summaries enabled via Anthropic API key.
{
"mcpServers": {
"jcodemunch": {
"command": "uvx",
"args": ["jcodemunch-mcp"],
"env": {
"CODE_INDEX_PATH": "~/.code-index/",
"JCODEMUNCH_USE_AI_SUMMARIES": "true",
"JCODEMUNCH_LOG_LEVEL": "WARNING",
"ANTHROPIC_API_KEY": "sk-ant-YOUR_KEY_HERE"
}
}
}
}Prompts to try
Example prompts that exercise the 60+ structured retrieval tools exposed by jCodeMunch.
- "Show me the repo map for the src/ directory"
- "Search for all symbols named 'authenticate' in the codebase"
- "What is the blast radius if I change the DatabaseConnection class?"
- "Find any dead code in the utils module"
- "Get the call hierarchy for the processPayment function"
- "Check if it is safe to delete the LegacyExporter class"Troubleshooting GitHub Codemunch
Tools do not appear in Claude after adding the server
Run `jcodemunch-mcp init` to ensure the index directory exists, then fully restart Claude Desktop or reload the MCP session. Check logs with JCODEMUNCH_LOG_LEVEL=DEBUG.
AI summaries are not generated even though JCODEMUNCH_USE_AI_SUMMARIES is true
Make sure at least one provider API key is set (ANTHROPIC_API_KEY, OPENAI_API_KEY, or GOOGLE_API_KEY). The server auto-detects the provider from whichever key is present.
Index is stale after code changes
jCodeMunch performs incremental updates automatically when you call retrieval tools, but you can force a full re-index by deleting ~/.code-index/ and running `jcodemunch-mcp init` again.
Frequently Asked Questions about GitHub Codemunch
What is GitHub Codemunch?
GitHub Codemunch is a Model Context Protocol (MCP) server that token-efficient mcp server for exploring github repositories using tree-sitter ast parsing to extract symbol signatures and summaries. it enables users to search and retrieve specific code components while significantly reducing token consumption c It connects AI assistants to external tools and data sources through a standardized interface.
How do I install GitHub Codemunch?
Install via pip with: pip install jcodemunch-mcp. Then configure your AI client to connect to this MCP server.
Which AI clients work with GitHub Codemunch?
GitHub Codemunch works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.
Is GitHub Codemunch free to use?
Yes, GitHub Codemunch is open source and available under the NOASSERTION license. You can use it freely in both personal and commercial projects.
GitHub Codemunch Alternatives — Similar Version Control Servers
Looking for alternatives to GitHub Codemunch? Here are other popular version control servers you can use with Claude, Cursor, and VS Code.
Ruiyu MA
★ 30.1kConnect AI assistants to GitHub - manage repos, issues, PRs, and workflows through natural language.
GitMCP
★ 8.1kTransforms any GitHub repository or GitHub Pages site into a documentation hub for AI assistants using the Model Context Protocol. It allows AI tools to access real-time code and documentation to prevent hallucinations and provide accurate API usage
Azure DevOps MCP Server
★ 1.7kThis server provides a convenient API for interacting with Azure DevOps services, enabling AI assistants and other tools to manage work items, code repositories, boards, sprints, and more. Built with the Model Context Protocol, it provides a standard
GitHub Official
★ 1.2kOfficial GitHub integration with comprehensive API coverage
Azure DevOps
★ 367Enables interaction with Azure DevOps through Cursor chat, providing tools to manage builds, pipelines, work items, sprints, and board operations. Supports secure authentication via Personal Access Tokens and allows natural language-driven DevOps tas
MCP Git Ingest
★ 306Helps AI read GitHub repository structure and important files. Want to quickly understand what a repo is about? Prompt it with 'read https://github.com/adhikasp/mcp-git-ingest and determine how the code technically works'.
Browse More Version Control MCP Servers
Explore all version control servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.
Set Up GitHub Codemunch in Your Editor
Choose your AI client for step-by-step setup instructions.
Quick Config Preview
Add this to your claude_desktop_config.json or .cursor/mcp.json
Ready to use GitHub Codemunch?
Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.