Code Agent

An AI-powered code review agent that leverages Model Context Protocol (MCP) and Claude to provide intelligent, context-aware code reviews.

✨ Features

• 🤖 AI-Powered Reviews: Uses Claude to provide intelligent code analysis
• 🔧 MCP Integration: Built on Model Context Protocol for tool-augmented LLM reasoning
• 📊 Code Metrics: Cyclomatic complexity, line counts, maintainability index
• 🎨 Style Checking: Detects common anti-patterns and style issues
• 📝 Git Integration: Review diffs, check history, understand context
• ⚡ Streaming Output: Real-time feedback as the AI analyzes your code

🚀 Quick Start

Installation

# Clone the repository
git clone https://github.com/geethikakoya/mcp-code-agent.git
cd mcp-code-agent

# Create virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -e .

Set up API Key

# Get your API key from https://console.anthropic.com/
export ANTHROPIC_API_KEY=your_api_key_here

Usage

# Review a file
mcp-review review src/main.py

# Review with streaming output
mcp-review review --stream src/main.py

# Review git changes
mcp-review diff

# Explain what a file does
mcp-review explain src/utils.py

# Quick analysis (no AI, just metrics)
mcp-review analyze src/main.py

📖 Architecture

┌─────────────────────────────────────────────────────────────┐
│                    Code Review Agent                        │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│  ┌──────────┐      ┌──────────────┐      ┌──────────────┐  │
│  │  CLI     │─────▶│  Agent       │─────▶│  Claude API  │  │
│  │  (typer) │◀─────│  (reviewer)  │◀─────│  (LLM)       │  │
│  └──────────┘      └──────────────┘      └──────────────┘  │
│                           │                                 │
│                           ▼                                 │
│                    ┌──────────────┐                        │
│                    │  MCP Tools   │                        │
│                    └──────────────┘                        │
│                           │                                 │
│              ┌────────────┼────────────┐                   │
│              ▼            ▼            ▼                   │
│        ┌─────────┐  ┌─────────┐  ┌─────────┐              │
│        │  File   │  │   Git   │  │  Code   │              │
│        │  Tools  │  │  Tools  │  │ Analysis│              │
│        └─────────┘  └─────────┘  └─────────┘              │
│                                                             │
└─────────────────────────────────────────────────────────────┘

🔧 MCP Tools

File Operations

Tool	Description
read_file	Read file contents with optional line range
list_directory	List files and directories
search_code	Regex search across codebase

Git Integration

Tool	Description
git_diff	Get diff of changes
git_log	Get commit history
get_changed_files	List modified files

Code Analysis

Tool	Description
calculate_complexity	Cyclomatic complexity (Python)
count_lines	LOC, comments, blanks
check_code_style	Style issues and anti-patterns

📝 Example Output

🔍 Reviewing: src/api/handler.py

📊 Metrics Summary
• Lines of Code: 142
• Cyclomatic Complexity: 8 (Moderate)
• Maintainability Index: 65.2

🔴 Critical Issues
1. Line 45: SQL injection vulnerability - use parameterized queries

⚠️ Warnings
1. Line 23: Function too long (>50 lines) - consider splitting
2. Line 67: Magic number - extract to constant

💡 Suggestions
1. Add type hints to function parameters
2. Consider using dataclass for UserData

✅ What's Good
• Good error handling in API endpoints
• Consistent naming conventions

🏗️ Project Structure

mcp-code-agent/
├── src/
│   ├── mcp_server/           # MCP Server (exposes tools)
│   │   ├── server.py         # Main MCP server
│   │   └── tools/
│   │       ├── file_tools.py
│   │       ├── git_tools.py
│   │       └── analysis_tools.py
│   │
│   ├── agent/                # AI Agent
│   │   ├── reviewer.py       # Code review logic
│   │   └── prompts.py        # System prompts
│   │
│   └── cli/                  # CLI interface
│       └── main.py
│
├── tests/
├── pyproject.toml
├── requirements.txt
└── README.md

🔌 Using as MCP Server

You can use this as a standalone MCP server with any MCP-compatible client:

# Run the MCP server
mcp-review server

Configure in your MCP client (e.g., Claude Desktop):

{
  "mcpServers": {
    "code-review": {
      "command": "python",
      "args": ["-m", "src.mcp_server.server"],
      "cwd": "/path/to/mcp-code-agent"
    }
  }
}

🧪 Development

# Install dev dependencies
pip install -e ".[dev]"

# Run tests
pytest

# Format code
black src/
ruff check src/

📄 License

MIT License - see LICENSE for details.

👤 Author

Geethika Koya

• GitHub: @geethikakoya
• LinkedIn: geethika-koya

---

Built with ❤️ using Model Context Protocol and Claude