# Code Review MCP Server

A Model Context Protocol (MCP) server that performs comprehensive code reviews using multiple AI models (O3, Gemini, Claude Opus) and consolidates the results with GPT-4.

## Features

- **Multi-Model Analysis**: Leverages O3, Google Gemini, and Claude Opus for diverse perspectives
- **Consolidated Reviews**: Uses GPT-4 to merge and organize feedback from all models
- **Comprehensive Feedback**: Focuses on bugs, security, performance, and best practices
- **MCP Integration**: Works seamlessly with Claude Code and other MCP-compatible clients

## Installation

1. Clone the repository:
```bash
git clone <repository-url>
cd code_review_mcp
```

2. Install dependencies:
```bash
pip install -r requirements.txt
```

3. Install in development mode:
```bash
pip install -e .
```

## API Keys Required

You'll need API keys from:
- **OpenAI**: For O3-mini (reviews) and GPT-4o (consolidation)
- **Google**: For Gemini 2.5 Pro
- **Anthropic**: For Claude 3 Opus Latest
- **Hugging Face**: For Qwen2.5-Coder model

## Configuration

Choose **one** of these two methods to provide your API keys:

### Method 1: Environment File (Recommended)
1. Set up API keys:
```bash
cp .env.template .env
# Edit .env and add your API keys
```

2. Add to your Claude Code MCP settings:
```json
{
  "mcpServers": {
    "code-review": {
      "command": "python",
      "args": ["-m", "code_review_mcp.server"]
    }
  }
}
```

### Method 2: MCP Settings
Add API keys directly to your Claude Code MCP settings:
```json
{
  "mcpServers": {
    "code-review": {
      "command": "python",
      "args": ["-m", "code_review_mcp.server"],
      "env": {
        "OPENAI_API_KEY": "your-openai-key",
        "GOOGLE_API_KEY": "your-google-key", 
        "ANTHROPIC_API_KEY": "your-anthropic-key",
        "HUGGINGFACE_API_KEY": "your-huggingface-key"
      }
    }
  }
}
```

## Usage

The server provides one tool:

### `multi_model_code_review`

Performs comprehensive code review using multiple AI models.

**Parameters:**
- `code` (required): The source code to review
- `description` (required): Author's description of the code's purpose
- `language` (optional): Programming language (default: auto-detect)

**Example usage in Claude:**
```
Please review this Python function using the multi_model_code_review tool:

Code:
def calculate_average(numbers):
    total = 0
    for num in numbers:
        total += num
    return total / len(numbers)

Description: This function calculates the average of a list of numbers.
```

## Performance Note

This tool can be slow (30-60 seconds) due to:
- Multiple API calls to different models
- High reasoning effort for O3 model
- Consolidation step with GPT-4

## API Requirements

- OpenAI API key (for O3 and GPT-4)
- Google API key (for Gemini)
- Anthropic API key (for Claude Opus)

## Error Handling

The server includes robust error handling:
- Individual model failures won't break the entire review
- Timeout protection (3 minutes total)
- Fallback consolidation if GPT-4 fails
- Clear error messages for missing API keys