MCP_SETUP_GUIDE.md

MCP Tools Setup Guide for AI Code Analyzer Project

🎯 Configured MCP Tools

I've set up the following MCP tools in your ~/.cursor/mcp.json file:

Essential for ML/AI Projects:

1. 🤗 Hugging Face - Model management and dataset access

   • API Key Location: https://huggingface.co/settings/tokens
   • Required Scope: read (minimum), write (if you want to upload models)
   • Replace: YOUR_HUGGINGFACE_API_KEY_HERE

2. 🗄️ DuckDB - Local analytics database (no API key needed)

   • Perfect for: Analyzing your analyst_dataset.jsonl file
   • No setup required - ready to use!

Development & Deployment:

3. 🐙 GitHub - Version control and collaboration

   • API Key Location: https://github.com/settings/tokens
   • Required Scope: repo, workflow, read:org
   • Replace: YOUR_GITHUB_TOKEN_HERE

4. 🚀 Vercel - Deploy web interfaces

   • API Key Location: https://vercel.com/account/tokens
   • Replace: YOUR_VERCEL_TOKEN_HERE

5. 🚂 Railway - Full-stack deployment

   • API Key Location: https://railway.app/account/tokens
   • Replace: YOUR_RAILWAY_TOKEN_HERE

Data Storage:

6. 🍃 MongoDB - Database for structured data
   • Connection String: Get from MongoDB Atlas (https://cloud.mongodb.com/)
   • Format: mongodb+srv://username:password@cluster.mongodb.net/database
   • Replace: YOUR_MONGODB_CONNECTION_STRING_HERE

Monitoring & Error Tracking:

7. 🚨 Sentry - Error tracking and performance monitoring
   • Auth Token: https://sentry.io/settings/account/api/auth-tokens/
   • Organization: Your Sentry org slug
   • Project: Your project slug
   • Replace:
     • YOUR_SENTRY_AUTH_TOKEN_HERE
     • YOUR_SENTRY_ORG_HERE
     • YOUR_SENTRY_PROJECT_HERE

Security & Code Quality:

8. 🔒 Snyk - Vulnerability scanning

   • API Key Location: https://app.snyk.io/account
   • Replace: YOUR_SNYK_TOKEN_HERE

9. 🔍 Semgrep - Static analysis (no API key needed)

   • Ready to use - no setup required!

10. 🏗️ SonarQube - Code analysis

    • Setup: You need a SonarQube instance (cloud or self-hosted)
    • Token: Generate in your SonarQube instance
    • Replace:
      • YOUR_SONARQUBE_URL_HERE (e.g., https://your-org.sonarcloud.io)
      • YOUR_SONARQUBE_TOKEN_HERE

🚀 Quick Start (Recommended Order)

Phase 1: Essential Tools (Start Here)

1. Hugging Face - Most important for your ML project
2. DuckDB - Already ready, no API key needed
3. GitHub - For version control

Phase 2: Development Tools

4. Vercel or Railway - For deployment
5. Sentry - For monitoring

Phase 3: Advanced Tools (Optional)

6. MongoDB - If you need structured data storage
7. Snyk - For security scanning
8. Semgrep - For static analysis
9. SonarQube - For comprehensive code analysis

📝 How to Add API Keys

1. Open your MCP config: c:\Users\arunk\.cursor\mcp.json
2. Find the placeholder (e.g., YOUR_HUGGINGFACE_API_KEY_HERE)
3. Replace with your actual API key/token
4. Save the file
5. Restart Cursor for changes to take effect

🔧 Installation Commands

Run these commands to ensure all MCP servers are installed:

# Install all MCP servers
npm install -g @modelcontextprotocol/server-huggingface
npm install -g @modelcontextprotocol/server-github
npm install -g @modelcontextprotocol/server-duckdb
npm install -g @modelcontextprotocol/server-sentry
npm install -g @modelcontextprotocol/server-vercel
npm install -g @modelcontextprotocol/server-mongodb
npm install -g @modelcontextprotocol/server-railway
npm install -g @modelcontextprotocol/server-snyk
npm install -g @modelcontextprotocol/server-semgrep
npm install -g @modelcontextprotocol/server-sonarqube

✅ Testing Your Setup

After adding API keys, test each tool in Cursor Composer:

• Hugging Face: "Search for code analysis models on Hugging Face"
• DuckDB: "Analyze my training dataset using DuckDB"
• GitHub: "Show me my recent commits"
• Sentry: "Check for errors in my project"

🛡️ Security Notes

• Never commit API keys to version control
• Use environment variables for sensitive data
• Regularly rotate your API keys
• Use minimal required permissions for each service

🆘 Troubleshooting

If a tool doesn't work:

1. Check if the MCP server is installed: npm list -g @modelcontextprotocol/server-*
2. Verify API keys are correct and have proper permissions
3. Restart Cursor after making changes
4. Check Cursor's MCP settings in Settings > Features > MCP

📚 Useful Commands for Your Project

Once set up, you can use these commands in Cursor Composer:

• "Use Hugging Face to find the best code analysis models"
• "Analyze my analyst_dataset.jsonl with DuckDB"
• "Check my code for vulnerabilities with Snyk"
• "Deploy my analyzer to Vercel"
• "Monitor errors with Sentry"