🤖 AI Code Review MCP Server

What is AI Code Review MCP Server?

An AI-powered code review and quality management server built on the Model Context Protocol (MCP). It integrates with GitHub pull requests and the Claude API to analyze code changes by commit type, compute quality scores, detect security vulnerabilities, and assess performance impact. Designed for developers using GitHub who want automated, context-aware code reviews.

How to use AI Code Review MCP Server?

Clone the repository, install dependencies (npm install), set environment variables (CLAUDE_API_KEY, GITHUB_TOKEN, etc.) in .env, then start the server with npm run dev or npm start. Access the web UI at http://localhost:3001, enter an owner/repo and PR number, select a commit type, and click “PR 분석” to run analysis. Alternatively, call the MCP tool analyze_pr or use HTTP POST to /analyze.

Key features of AI Code Review MCP Server

• Commit‑type‑specific analysis (feat, fix, refactor, etc.)
• Quality score calculation (complexity, maintainability, security)
• Automatic security vulnerability detection
• Performance impact assessment
• Integrated review checklist and markdown report generation
• Review guidelines with best practices and common pitfalls

Use cases of AI Code Review MCP Server

• Automate PR reviews in a GitHub repository with commit‑type tailored feedback
• Generate comprehensive review guides and checklists for each commit type
• Compute and track code quality scores across multiple PRs
• Identify security issues and performance bottlenecks before merging

FAQ from AI Code Review MCP Server

What API keys are required?

Claude API key (CLAUDE_API_KEY) and GitHub Personal Access Token (GITHUB_TOKEN) with repo permissions are required.

How do I connect to a GitHub repository?

Set up a GitHub webhook pointing to https://your-domain.com/webhook with content type application/json and events set to Pull requests. Provide the webhook secret in GITHUB_WEBHOOK_SECRET.

What happens if a PR analysis fails?

Check that the PR exists, the repository is accessible with the configured token, and that the Claude API key is valid with sufficient quota.

Can I customize the review templates or quality scoring?

Yes – edit src/handlers/claude-client.ts for prompt templates and src/handlers/pr-analyzer.ts for quality score calculation logic.

What transports and authentication methods are supported?

The server provides both HTTP APIs (REST over port 3001) and MCP tool calls. Authentication is via API keys set in environment variables.