CodeSentry MCP 🛡️

AI-powered code review assistant for LLM development workflows

CodeSentry is a Model Context Protocol (MCP) server that provides comprehensive code review capabilities through 5 specialized review tools. Built for Cursor, Claude Code, and other MCP-compatible AI assistants.

✨ Features

🔍 Plan Review - Validates implementation plans against codebase context
📊 Implementation Review - Compares completed work vs. original plans
🏗️ Code Review - General codebase analysis with focus areas
🔒 Security Review - Vulnerability assessment and security analysis
📐 Best Practices Review - Code quality and maintainability analysis

🚀 Quick Start

For Cursor

1. Install & Build:

   1git clone https://github.com/crazyrabbitLTC/mcp-codesentry.git
   2cd mcp-codesentry
   3npm install && npm run build

2. Add to Cursor Settings (Cmd/Ctrl + , → Extensions → MCP):

   1{
   2  "mcpServers": {
   3    "codesentry": {
   4      "command": "node",
   5      "args": ["/path/to/mcp-codesentry/dist/index.js"],
   6      "env": {
   7        "GEMINI_API_KEY": "your-api-key-here"
   8      }
   9    }
   10  }
   11}

3. Get API Key: Google AI Studio

4. Use in Chat:

   @codesentry Please review this codebase for security issues
   

For Claude Code

1. Complete steps 1-3 above

2. Add to MCP Settings (~/.claude/mcp_servers.json):

   1{
   2  "codesentry": {
   3    "command": "node",
   4    "args": ["/path/to/mcp-codesentry/dist/index.js"],
   5    "env": {
   6      "GEMINI_API_KEY": "your-api-key-here"
   7    }
   8  }
   9}

🔧 Available Tools

security_review

1// Comprehensive security vulnerability assessment
2{
3  "codebasePath": "./src",
4  "securityFocus": "authentication" // optional
5}

best_practices_review

1// Code quality and maintainability analysis
2{
3  "codebasePath": "./src", 
4  "practicesFocus": "testing", // optional
5  "language": "TypeScript" // optional
6}

code_review

1// General codebase analysis
2{
3  "codebasePath": "./src",
4  "reviewFocus": "performance" // optional
5}

review_plan

1// Pre-task planning validation
2{
3  "taskId": "feature-123",
4  "taskDescription": "Add user authentication",
5  "implementationPlan": "Use JWT with refresh tokens...",
6  "codebasePath": "./src"
7}

review_implementation

1// Post-task implementation review
2{
3  "taskId": "feature-123",
4  "taskDescription": "Add user authentication", 
5  "originalPlan": "Use JWT with refresh tokens...",
6  "implementationSummary": "Implemented JWT auth with Redis...",
7  "beforePath": "./before",
8  "afterPath": "./after"
9}

🏗️ How It Works

AI Assistant → MCP Tool → Repomix Analysis → Gemini Review → Structured Feedback

Benefits:

• ✅ Handles large codebases without context window limits
• ✅ Leverages Gemini's 2M+ token context for full analysis
• ✅ Returns concise, actionable feedback
• ✅ Maintains context between planning and implementation

🛡️ Security & Privacy

• 🔐 Local Processing - Your code stays on your machine
• 🔑 API Key Security - Environment variables only
• 🗑️ Auto Cleanup - Temporary files automatically deleted
• 📝 Audit Logging - Comprehensive operation logs

🧪 Development

1npm run dev          # Development mode
2npm test             # Run test suite (32 tests)
3npm run build        # TypeScript build
4npm run lint         # Code linting

📋 Requirements

• Node.js 18+
• Google Gemini API Key (Get one free)
• MCP-compatible AI Assistant (Cursor, Claude Code, etc.)

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Add tests for new functionality
4. Submit a pull request

👨‍💻 Authors

Dennison Bertram - Creator & Maintainer
GitHub: @crazyrabbitLTC
Email: dennison@tally.xyz

Claude (Anthropic) - AI Development Partner
Assisted with architecture, implementation, and testing

📄 License

MIT License - see LICENSE file for details.

---

⭐ Star this repo if CodeSentry helps improve your code quality!