Complexipy MCP Server

A Model Context Protocol (MCP) server that wraps complexipy to provide cognitive complexity analysis for Python codebases. This server uses the complexipy Python API for accurate and robust analysis.

Features

• Directory Scanning: Recursively analyze all .py files in a directory and filter by complexity threshold.

• File Scanning: Analyze a single Python file.

• Data Processing: Uses pandas for efficient data filtering and sorting.

• Clean Output: Returns JSON formatted results (list of objects).

• No Emojis: Pure data output suitable for machine consumption.

• Stdio Transport: Uses standard input/output for communication, making it compatible with most MCP clients.

Installation & Usage

Quick Start

You can run the server directly using uvx (no installation required):

Configuration for Claude Desktop

Add the following configuration to your MCP client settings (e.g., claude_desktop_config.json):

Local Development

1. Clone the repository.

2. Install dependencies:

   uv sync

3. Run the server locally:

   uv run complexipy-mcp

Local Configuration for Claude Desktop

If you are developing locally, point the configuration to your local setup:

Tools

scan_directory

Scans a directory for files exceeding a cognitive complexity threshold.

• Arguments:

  • directory_path (string): Absolute path to the directory to scan.

  • threshold (integer): Complexity limit (default: 15).

• Returns: JSON string containing a list of files that exceed the threshold.

  [ { "file_path": "/abs/path/to/file.py", "complexity": 25, "functions": [ { "name": "function_name", "complexity": 10, "line_start": 5, "line_end": 15 } ], "status": "exceeds_threshold" } ]

scan_file

Scans a single file for cognitive complexity.

• Arguments:

  • file_path (string): Absolute path to the file.

  • threshold (integer): Complexity limit (default: 15).

• Returns: JSON string containing the result if it exceeds the threshold.