jcodemunch-mcp

Symbol-level code retrieval for AI agents: fetch the function you need, not the whole file. Runs locally. 96.2% fewer tokens per lookup on average than grep-and

What is jcodemunch-mcp?

Symbol-level code retrieval for AI agents: fetch the function you need, not the whole file. Runs locally. 96.2% fewer tokens per lookup on average than grep-and-read on our 15-lookup benchmark. Works with Claude Code, Cursor, Codex and any MCP client. Free for non-commercial use; commercial use needs a one-time license.

Overview of jcodemunch-mcp

What it is

jCodeMunch MCP is a Model Context Protocol server that retrieves precise source code from GitHub repositories using tree-sitter AST parsing. It indexes a codebase once and allows AI agents to fetch only the exact code they need—functions, classes, methods, constants, outlines, and tightly scoped context bundles—instead of reading entire files repeatedly. The project reports 838B+ tokens saved, 136,000+ reporting installs, $4.2M+ in AI spend avoided, and 100,000+ kg CO₂ prevented (figures as of 2026-08-17, valued at the $5/MTok Claude Opus input rate). It works with Claude Code, Cursor, VS Code, Codex CLI, Windsurf, Continue, and any MCP-compatible client.

Key features

  • Symbol-level retrieval: search for symbols and fetch exact implementations with byte-level precision.
  • Tree-sitter AST parsing: indexes source code into structured symbol metadata (signature, kind, qualified name, summary, byte offsets).
  • Token efficiency: benchmarked at 28.9x fewer tokens than a grep-and-read agent (96.5% reduction) across three public repos.
  • Structural queries: includes tools like get_blast_radius, find_importers, get_class_hierarchy, and find_dead_code that native tools cannot answer.
  • Compact MUNCH wire encoding: trims a median 45.5% more bytes off responses.
  • One-command install and init: uv tool install jcodemunch-mcp then jcodemunch-mcp init auto-detects MCP clients and writes config entries.

Use cases

  • Finding a function: search symbol, fetch exact implementation instead of opening and scanning large files.
  • Understanding a module: pull only relevant symbols and imports rather than reading broad file regions.
  • Exploring repository structure: query outlines, trees, and targeted bundles instead of traversing file after file.
  • Impact analysis: use get_blast_radius to determine what breaks if a change is made.
  • Orphaned file detection: use find_importers to detect files that are not imported anywhere.

Getting started

bashuv tool install jcodemunch-mcpjcodemunch-mcp initVerify with:bashjcodemunch-mcp --versionFor manual Claude Code setup:bashclaude mcp add -s user jcodemunch -- uvx jcodemunch-mcpThen add to CLAUDE.md:markdownCall the jcodemunch_guide tool and strictly follow its instructions.

License and pricing

The license is not stated in the metadata (NOASSERTION). The README states:

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/jgravellejcodemunch-mcp.svg)](https://yunodb.com/tools/jgravellejcodemunch-mcp?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/jgravellejcodemunch-mcp?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/jgravellejcodemunch-mcp.svg" alt="Featured on YunoDB" height="20"></a>