jCodeMunch MCP Server

Content

jCodeMunch MCP is a source-code exploration server for agents that need precise code context without repeatedly loading entire files. It indexes repositories with tree-sitter, stores structured symbol metadata, and lets MCP clients fetch exact functions, classes, methods, outlines, dependency context, repository maps, and token-budgeted context bundles.

The project is local-first and installable from PyPI as jcodemunch-mcp. The upstream license is dual-use: free for non-commercial use, with a paid license required for commercial use.

Source Review

• https://github.com/jgravelle/jcodemunch-mcp
• https://github.com/jgravelle/jcodemunch-mcp/blob/main/pyproject.toml
• https://github.com/jgravelle/jcodemunch-mcp/blob/main/server.json
• https://github.com/jgravelle/jcodemunch-mcp/blob/main/LANGUAGE_SUPPORT.md
• https://github.com/jgravelle/jcodemunch-mcp/blob/main/LICENSE
• https://pypi.org/pypi/jcodemunch-mcp/json

Verified on 2026-06-18. The PyPI package is jcodemunch-mcp, requires Python 3.10 or newer, exposes the jcodemunch-mcp command, and declares tree-sitter, MCP, HTTP, and configuration dependencies. The server manifest declares stdio transport and the GitHub repository source. The license file states that commercial use requires a paid license.

Features

• Index local folders and GitHub repositories for structured code retrieval.
• Search symbols by name, kind, language, fuzzy match, semantic match, or hybrid retrieval when optional dependencies are configured.
• Retrieve exact symbol implementations and file outlines instead of broad file reads.
• Assemble token-budgeted context bundles for coding tasks.
• Inspect importers, references, class hierarchy, call hierarchy, dependency cycles, coupling metrics, and blast radius.
• Find dead code, untested symbols, changed symbols, similar symbols, hotspots, and structural AST patterns.
• Generate repository maps, task context, refactoring plans, symbol provenance, and PR risk profiles.
• Support many source languages through tree-sitter and custom extractors.

Installation

For MCP clients that launch local stdio servers:

{
  "mcpServers": {
    "jcodemunch": {
      "command": "uvx",
      "args": ["jcodemunch-mcp"]
    }
  }
}

You can also install the package directly:

pip install jcodemunch-mcp
jcodemunch-mcp

Review the upstream license before using it in a commercial environment.

Use Cases

• Let Claude Code, Codex, Cursor, or another agent find exact functions without reading whole files.
• Build a small context pack for a bug fix or feature without overloading the model window.
• Ask for blast radius, dependency, or call hierarchy before changing shared code.
• Find dead code, untested symbols, hotspots, or duplicate implementation candidates.
• Analyze a PR branch for symbol-level risk before review.
• Explore unfamiliar repositories through structure-first search and outlines.

Safety and Privacy

jCodeMunch works by indexing source repositories. That means the local index and MCP responses may expose proprietary code, file paths, symbol names, signatures, comments, architecture, dependency relationships, and git history. Use explicit allowlists, avoid indexing repositories outside the intended workspace, and do not connect confidential codebases to hosted summarizers unless that is allowed by policy.

Review optional background features before enabling them. File watching, auto-reindexing, login-service installation, MCP-client hooks, config audits, and anonymous aggregate token-savings telemetry are useful in the right environment but should be deliberate choices.

Duplicate Check

Checked current content/mcp/, content/tools/, content/skills/, open pull requests, and repository-wide content for jgravelle/jcodemunch-mcp, jcodemunch-mcp, jCodeMunch MCP, tree-sitter MCP, token-efficient code search, source code exploration MCP, and blast radius MCP. No dedicated jCodeMunch MCP entry, exact source URL duplicate, target file, or open duplicate PR was found.