MCP Code Expert System

A Python-based code review system using the Model Context Protocol (MCP). It provides code review capabilities through simulated expert personas like Martin Fowler and Robert C. Martin (Uncle Bob).

Features

• Code review based on Martin Fowler's refactoring principles

• Code review based on Robert C. Martin's Clean Code principles

• Knowledge graph storage of code, reviews, and relationships

• Integration with Ollama for AI-powered reviews

• Server-side Event (SSE) support for web integration

Related MCP server: MAGI MCP Server

Prerequisites

Python 3.10+

This project requires Python 3.10 or higher.

Ollama

Ollama is required for AI-powered code reviews.

1. Install Ollama for your platform:

   • macOS: Download from ollama.com

   • Linux: curl -fsSL https://ollama.com/install.sh | sh

   • Windows: Windows WSL2 support via Linux instructions

2. Pull a recommended model:

   ollama pull llama3:8b

3. Start the Ollama server:

   ollama serve

Installation

Run the setup script to install dependencies and create the virtual environment:

Configuration

Edit the .env file to configure (create from .env.example if needed):

Usage

Running the Server

Standard Mode (for Cursor Integration)

HTTP/SSE Mode (for Web Integration)

This will start the server at http://localhost:8000/sse for SSE transport.

For custom port:

Installing in Cursor

To install in Cursor IDE:

Available Tools

The server exposes these tools:

• ask_martin: Ask Martin Fowler to review code and suggest refactorings

• ask_bob: Ask Robert C. Martin (Uncle Bob) to review code based on Clean Code principles

• read_graph: Read the entire knowledge graph

• search_nodes: Search for nodes in the knowledge graph

• open_nodes: Open specific nodes by their names

Example Usage

To review a code snippet with Martin Fowler:

Project Structure

• server.py: Main server implementation with MCP integration

• experts/: Expert modules implementing the code review capabilities

  • __init__.py: Shared models and interfaces

  • martin_fowler/: Martin Fowler expert implementation

  • robert_c_martin/: Robert C. Martin expert implementation

• knowledge_graph.py: Knowledge graph for storing code and reviews

• ollama_service.py: Integration with Ollama for AI-powered reviews

• examples/: Example code for review in different languages

• requirements.txt: Python dependencies

• setup.sh: Setup script

Architecture

The system follows a modular architecture:

1. Server Layer: Handles MCP protocol communication and routes requests

2. Expert Layer: Encapsulates code review logic for each expert

3. Service Layer: Provides AI integration and knowledge graph functionality

Each expert implements a standard interface allowing for consistent handling and easy addition of new experts.

License

MIT