Crazyrabbitltc MCP Code Review Server

🚀 Code Review Server

A custom MCP server that leverages Repomix and Large Language Models (LLMs) to perform code reviews.

✨ Features

• Flatten code repositories using Repomix.
• Analyze code using large language models.
• Obtain structured code comments with specific issues and suggestions.
• Support multiple LLM providers (OpenAI, Anthropic, Gemini).
• Handle chunking for large code repositories.

📦 Installation

# Clone the repository
git clone https://github.com/yourusername/code-review-server.git
cd code-review-server

# Install dependencies
npm install

# Build the server
npm run build

🛠️ Configuration

Create a .env file in the root directory based on the .env.example template:

cp .env.example .env

Edit the .env file to set your preferred LLM provider and API key:

# LLM provider configuration
LLM_PROVIDER=OPEN_AI
OPENAI_API_KEY=your_openai_api_key_here

💻 Usage Examples

Basic Usage

As an MCP Server: The code review server implements the Model Context Protocol (MCP) and can be used with any MCP client:

# Start the server
node build/index.js

The server provides two main tools:

1. analyze_repo: Flatten the code repository using Repomix.
2. code_review: Perform code review using LLM.

Advanced Usage

This server offers two different tools for different code analysis needs:

analyze_repo

Use this tool when you need to:

• Get a high - level overview and structure of the code repository.
• Flatten the code repository using Repomix for initial analysis.
• Understand the directory structure and file contents without a detailed review.
• Prepare for a more in - depth code review.
• Quickly scan the code.

For example:

# Example command
node analyze_repo --path /your/repository/path

code_review

Use this tool when you need to:

• Conduct an in - depth analysis of the code and obtain detailed review results.
• Handle chunking for large code repositories.
• Generate detailed review reports based on focus areas and levels of detail.

🔧 Technical Details

How LLM Integration Works

1. The code_review tool uses Repomix to process the code and flatten the repository structure.
2. The code is formatted and chunked (if necessary) to fit the LLM context limit.
3. Detailed prompts are generated based on the focus area and level of detail.
4. The prompts and code are sent directly to the API of your selected LLM provider.
5. The LLM response is parsed into a structured format.
6. The review results are returned as a JSON object, including issues, strengths, and suggestions. The implementation includes retry logic to enhance resilience against API errors and ensures proper code formatting to include the most relevant code in the review.

Code Review Output Format

The code review is returned in a structured JSON format:

{
  "summary": "A brief overview of the code and its purpose",
  "issues": [
    {
      "type": "SECURITY|PERFORMANCE|QUALITY|MAINTAINABILITY",
      "severity": "HIGH|MEDIUM|LOW",
      "description": "Description of the issue",
      "line_numbers": [12, 15],
      "recommendation": "Recommended fix"
    }
  ],
  "strengths": ["List of code strengths"],
  "recommendations": ["List of overall suggestions"]
}

📄 License

MIT