project-analysis-mcp

An MCP server that helps an LLM understand and analyze a codebase. It exposes four read-only tools:

Transport: stdio (JSON-RPC). Python ≥ 3.11.

Install & run

Option A — from PyPI (recommended)

uvx project-analysis-mcp
# or
pipx install project-analysis-mcp

Option B — from this repo (editable)

cd mcp-servers/project_analysis_mcp_pkg
uv pip install -e .
# then run:
project-analysis-mcp

Option C — run the module directly (no install)

uv run --directory <repo-root> python mcp-servers/project_analysis_mcp.py

Configure a client

Add the server to your client's MCP config (the exact file depends on the client: mcp.json, settings.json, claude_desktop_config.json, etc.).

Using the console script (after install):

{
  "mcpServers": {
    "Project Analysis": {
      "command": "project-analysis-mcp"
    }
  }
}

Using uvx (auto-fetches the package):

{
  "mcpServers": {
    "Project Analysis": {
      "command": "uvx",
      "args": ["project-analysis-mcp"]
    }
  }
}

Using the repo directly (no publish):

{
  "mcpServers": {
    "Project Analysis": {
      "command": "uv",
      "args": [
        "run", "--directory", "/abs/path/to/repo",
        "python", "/abs/path/to/repo/mcp-servers/project_analysis_mcp.py"
      ],
      "env": {}
    }
  }
}

Notes

• All tools are read-only and operate on a project_path you pass in.
• Binary/large/ignored files are skipped automatically.
• Compatible with any MCP client that supports stdio servers (Claude Code, Claude Desktop, CodeBuddy, custom clients using the official mcp SDK, etc.).

Project layout

mcp-servers/
├── project_analysis_mcp.py          # thin launcher (kept for repo-local use)
└── project_analysis_mcp_pkg/        # publishable package (single source of truth)
    ├── pyproject.toml               # build + console entry point
    ├── README.md
    └── project_analysis_mcp/
        ├── __init__.py              # server + tools + main()
        └── __main__.py              # python -m project_analysis_mcp