Log inSubmit server
JCodemunch MCP logo

JCodemunch MCP

Indexes local repositories by AST symbols so agents can retrieve focused code, inspect structure, and analyze dependency impact with fewer tokens.

InstallRepositoryWebsite
Full README Compare alternatives Embed widget Claim this listing

Install JCodemunch MCP

Pick your client, then add it in one click or paste the config.

STDIOPython
~/Library/Application Support/Claude/claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "jcodemunch-mcp": {
      "command": "uvx",
      "args": [
        "jcodemunch-mcp"
      ]
    }
  }
}

Settings → Developer → Edit Config, merge this into mcpServers, then fully quit and reopen Claude Desktop.

Install not yet sandbox-tested— we haven't run this command through our automated sandbox yet. Not a red flag; we're working through the catalog.

Overview

jCodemunch MCP MCP server indexes local source repositories with tree-sitter and exposes symbol-level code retrieval through MCP. Agents can search functions, classes, methods, constants, and other symbols, then fetch exact implementations instead of reading whole files. Structural queries cover outlines, importers, class hierarchies, dead code, and potential change impact. Reach for it when repository exploration is consuming excessive context or when you need code relationships that basic file and grep tools do not provide.

Use cases

  • Search a large repository for a function and retrieve only its implementation
  • Inspect importers before changing a shared symbol
  • Map class inheritance relationships during codebase exploration
  • Find potentially dead code without reading every source file

Key features

  • Tree-sitter AST parsing for 70+ languages
  • Symbol-level search and source retrieval
  • Local structured repository index
  • Import, hierarchy, dead-code, and change-impact queries
  • Automatic index-refresh hooks

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

ActiveLast checked Today

Recent health check succeeded.

12/12 checks healthy over the last 35d

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 14d ago via OSV.dev · jcodemunch-mcp (PyPI)

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

GitHub stars2.8kStargazers on the source repository.
Last commitTodayMost recent push to the default branch.

How JCodemunch MCP works

What jCodemunch MCP does

jCodemunch MCP turns a source repository into a searchable structural index for AI coding clients. Rather than making an agent open large files and scan unrelated lines, it lets the agent locate symbols and retrieve narrowly scoped source. Supported results include functions, classes, methods, constants, signatures, qualified names, summaries, outlines, trees, imports, and targeted context bundles.

The server is designed for local-first code exploration and supports more than 70 programming languages through tree-sitter parsing. Its symbol-oriented approach is useful for repository navigation, implementation lookup, dependency investigation, and code-change planning.

How it works

A repository is parsed and indexed once. The local index stores structured metadata such as each symbol's kind, signature, qualified name, summary, and byte offsets alongside the raw file content. Subsequent MCP requests use that metadata to find relevant symbols and return the corresponding source ranges without repeatedly transferring entire files.

The server also supports structural questions that are difficult to answer with simple search. find_importers can identify code that imports a target, get_blast_radius examines what may be affected by a change, get_class_hierarchy exposes class relationships, and find_dead_code supports unused-code investigation. Compact MUNCH response encoding can reduce response size further.

A published benchmark across Express, FastAPI, and Gin measured 96.2% fewer tokens than a grep-and-read workflow for the tested lookups. Results vary by query and repository; the benchmark reports different savings for each project.

Setup and configuration

Install the package with uv, then initialize its MCP configuration:

bash
uv tool install jcodemunch-mcp
jcodemunch-mcp init

The initialization command detects Claude Code, Claude Desktop, Cursor, Windsurf, and Continue, and writes the relevant client configuration entries. It can also install hooks that keep the index updated after edits, create a pre-compaction checkpoint, and record a task-complete note without adding those hook messages to the model's context.

VS Code and Cursor provide one-click installation links using uvx and the jcodemunch-mcp package. The server can also be configured in any MCP-compatible client. Personal use is free; commercial use is covered by separate licensing terms.

Tools and capabilities

  • Search indexed symbols by name or related query with search_symbols.
  • Retrieve the exact implementation for a selected symbol with get_symbol_source.
  • Inspect repository outlines, trees, imports, and focused context bundles.
  • Find dependent files with find_importers.
  • Examine likely change impact with get_blast_radius.
  • Investigate class relationships and unused code with get_class_hierarchy and find_dead_code.

Limitations and notes

The server depends on an indexed codebase, so the useful context comes from repositories that have been parsed and stored in its local index. Its token-efficiency figures describe specific benchmark workflows rather than a guaranteed saving for every task. The project identifies itself as local-first and uses a dual-use licensing model, with commercial licensing handled separately from personal use.

Read the full README View source on GitHub

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about JCodemunch MCP

jCodemunch MCP is an MCP server that connects MCP-compatible AI clients to local source-code repositories through a tree-sitter-based index. Its main tools search_symbols and get_symbol_source locate symbols and return implementations, while structural tools such as find_importers and get_blast_radius help analyze dependencies and change impact without sending whole files.

Install it with `uv tool install jcodemunch-mcp`, then run `jcodemunch-mcp init`. The initializer detects supported clients and writes their MCP configuration. VS Code and Cursor also have one-click installation links that invoke the package through `uvx`.

Yes, jCodemunch MCP works with Cursor. Cursor is supported by the initializer and has a one-click installation link. The generated configuration runs the server through `uvx` with the `jcodemunch-mcp` package.

jCodemunch MCP supports Claude Code, Claude Desktop, Cursor, VS Code, Codex CLI, Windsurf, Continue, and other MCP-compatible clients. Its initialization command can detect several of these clients and add their configuration entries automatically.

Related MCP servers

All in Developer Tools All alternatives
Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewJCodemunch MCP AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/jcodemunch-mcp?style=directory)](https://allmcps.com/mcp/jcodemunch-mcp)
HTML Embed
<a href="https://allmcps.com/mcp/jcodemunch-mcp"><img src="https://allmcps.com/api/badge/jcodemunch-mcp?style=directory" alt="JCodemunch MCP on AllMCPs" /></a>