jCodeMunch MCP

Install

npx skillfoxx add mcp/jcodemunch-mcp

claude mcp add --transport stdio jcodemunch-mcp -- uvx jcodemunch-mcp==1.108.319

{
  "mcpServers": {
    "jcodemunch-mcp": {
      "command": "uvx",
      "args": [
        "jcodemunch-mcp==1.108.319"
      ]
    }
  }
}

{
  "mcpServers": {
    "jcodemunch-mcp": {
      "command": "uvx",
      "args": [
        "jcodemunch-mcp==1.108.319"
      ]
    }
  }
}

code --add-mcp '{"name":"jcodemunch-mcp","type":"stdio","command":"uvx","args":["jcodemunch-mcp==1.108.319"]}'

{
  "servers": {
    "jcodemunch-mcp": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "jcodemunch-mcp==1.108.319"
      ]
    }
  }
}

codex mcp add jcodemunch-mcp -- uvx jcodemunch-mcp==1.108.319

[mcp_servers.jcodemunch-mcp]
command = "uvx"
args = ["jcodemunch-mcp==1.108.319"]

gemini mcp add -s user jcodemunch-mcp uvx jcodemunch-mcp==1.108.319

{
  "mcpServers": {
    "jcodemunch-mcp": {
      "command": "uvx",
      "args": [
        "jcodemunch-mcp==1.108.319"
      ]
    }
  }
}

{
  "mcpServers": {
    "jcodemunch-mcp": {
      "command": "uvx",
      "args": [
        "jcodemunch-mcp==1.108.319"
      ]
    }
  }
}

{
  "mcpServers": {
    "jcodemunch-mcp": {
      "command": "uvx",
      "args": [
        "jcodemunch-mcp==1.108.319"
      ]
    }
  }
}

{
  "mcpServers": {
    "jcodemunch-mcp": {
      "command": "uvx",
      "args": [
        "jcodemunch-mcp==1.108.319"
      ]
    }
  }
}

{
  "mcp": {
    "jcodemunch-mcp": {
      "type": "local",
      "command": [
        "uvx",
        "jcodemunch-mcp==1.108.319"
      ]
    }
  }
}

{
  "context_servers": {
    "jcodemunch-mcp": {
      "command": "uvx",
      "args": [
        "jcodemunch-mcp==1.108.319"
      ]
    }
  }
}

{
  "mcpServers": {
    "jcodemunch-mcp": {
      "command": "uvx",
      "args": [
        "jcodemunch-mcp==1.108.319"
      ]
    }
  }
}

uv tool install jcodemunch-mcp
jcodemunch-mcp init

This is third-party code. Review the repository files before installing.

What it does

The server parses a repository once with tree-sitter and builds a local symbol index: signatures, kinds, qualified names, byte offsets. After that the agent requests a specific function, class, method or a scoped context bundle rather than a full file. It answers structural questions ordinary search cannot: who imports a symbol, what breaks on a change, the call hierarchy, dead code detection. The index lives on disk and refreshes as edits happen, with support for more than 70 languages. The point is to let the agent read less noise and spend fewer tokens on code exploration.

Who it is for. For developers and teams who drive an agent through large or unfamiliar repositories and want to cut token spend on reading code.

Example request

Index this repository with jcodemunch and show me the source of the function that handles authorization

Limitations

Dual-use license: free for personal use, commercial use needs a paid license from the author. It installs via the uv tool manager or PyPI and needs Python. By default the server sends an anonymous saved-tokens counter, which you can turn off with the share_savings setting. Some tree-sitter grammars are fetched over the network on first parse, so an offline install cannot parse those languages. Optional extras such as semantic search and AI summaries pull in external services and are enabled separately.

How to disable. Remove the server entry from your MCP client config. If you installed it as a uv tool, remove it with uv tool uninstall jcodemunch-mcp. The local index lives in the ~/.code-index/ directory.

MCP

Transport

stdio, http, sse

Authentication

not required

Security check

• Reads and indexes the project's source files locally
• Sends an anonymous statistics counter by default, with some network features optional

README in short

The README describes an MCP server for precise source code retrieval: it indexes a repository with tree-sitter and returns symbols instead of files to cut token usage during agent work. It lists structural tools including importer search, change impact analysis, call hierarchy and dead code detection, plus assembling a task's context in one call. Installation goes through uv, uvx, pipx or pip, and the init command writes config for common clients on its own. The server runs locally, the index is stored on disk, and some network features are optional. The license is dual-use: free for personal use, paid for commercial use.

Related