Vane MCP Server - GitHub Copilot Integration

A production-ready Model Context Protocol (MCP) server that extends GitHub Copilot with GitHub repository management capabilities. Enables AI-driven issue tracking, repository information retrieval, and seamless GitHub integration.

✨ Features

• 🔌 HTTP-based MCP Server - Universal compatibility with AI tools
• 🚀 GitHub Integration - Native GitHub tool support (list issues, create issues, get repo info)
• 🎯 Agent-Ready Architecture - Built on Vane agent framework with workspace isolation
• 📊 Semantic Caching - Intelligent context management with confidence thresholds
• 🔧 Configurable - Centralized configuration with environment variable support
• 📝 Production-Ready - Railway/Docker deployment guides included
• 🔄 Reusable - Single MCP server instance serves multiple projects

🚀 Quick Start

Local Development (5 minutes)

# 1. Clone and setup
git clone https://github.com/myou260312-eng/skills-integrate-mcp-with-copilot
cd skills-integrate-mcp-with-copilot

# 2. Configure environment
cp .env.example .env
# Edit .env with GITHUB_TOKEN

# 3. Install dependencies
pip install -r requirements.txt

# 4. Start MCP server
python mcp_server.py

Output:

✅ MCP Server started at http://0.0.0.0:3000
🚀 MCP Server is running. Press Ctrl+C to stop.

Verify Server

curl http://localhost:3000/mcp/

🌐 Production URL

https://mdabul-project.railway.app/mcp/

Use this URL in any project's .vscode/mcp.json:

{
  "servers": {
    "github": {
      "type": "http",
      "url": "https://mdabul-project.railway.app/mcp/"
    }
  }
}

📚 API Endpoints

Endpoint	Method	Description
/	GET	Server status and capabilities
/mcp/	GET	MCP tools and resources
/	POST	Execute MCP tools

🛠️ Available Tools

list_issues

List GitHub issues fromany of my repository.

Parameters:

• state (string): "open", "closed", or "all"
• labels (array): Filter by labels
• limit (integer): Max results (default: 10)

Example:

curl -X POST http://localhost:3000/ \
  -H "Content-Type: application/json" \
  -d '{
    "tool": "list_issues",
    "parameters": {"state": "open", "limit": 5}
  }'

create_issue

Create a new GitHub issue.

Parameters:

• title (string, required): Issue title
• body (string): Issue description
• labels (array): Labels to add

Example:

curl -X POST http://localhost:3000/ \
  -H "Content-Type: application/json" \
  -d '{
    "tool": "create_issue",
    "parameters": {
      "title": "New Feature Request",
      "body": "Description here",
      "labels": ["enhancement"]
    }
  }'

get_repository_info

Get information about the configured repository.

Example:

curl -X POST http://localhost:3000/ \
  -H "Content-Type: application/json" \
  -d '{"tool": "get_repository_info"}'

⚙️ Configuration

Environment Variables

# Required
GITHUB_TOKEN=ghp_xxxx                    # GitHub Personal Access Token
GITHUB_REPO_OWNER=myou260312-eng         # Repository owner
GITHUB_REPO_NAME=repo-name               # Repository name

# Optional
MCP_ENABLED=true                         # Enable/disable MCP (default: true)
LLM_MODEL=gpt-3.5-turbo                  # LLM model to use
LOG_LEVEL=INFO                           # DEBUG, INFO, WARNING, ERROR
DEBUG_MODE=false                         # Verbose logging
VANE_ROOT_ID=VANE_ROOT_ID_8A9B3C4D5E6F7G8H  # Workspace ID

Core Settings (config.py)

Setting	Default	Purpose
MCP_ENABLED	true	Enable/disable MCP
TEMPERATURE	0.0	LLM determinism (0=strict)
CONFIDENCE_HIGH_THRESHOLD	0.85	High confidence barrier
API_TIMEOUT	3.5s	External API timeout
MAX_TOKENS	800	Max LLM response tokens

📦 Project Structure

skills-integrate-mcp-with-copilot/
├── config.py                      # Core configuration with MCP settings
├── mcp_server.py                  # MCP HTTP server implementation
├── requirements.txt               # Python dependencies
├── .env.example                   # Environment variable template
├── .vscode/
│   └── mcp.json                   # VS Code MCP configuration
├── MCP_DEPLOYMENT_GUIDE.md        # Complete deployment guide
└── README.md                      # This file

🔌 Using with GitHub Copilot

In VS Code or Codespace

1. Update .vscode/mcp.json with my MCP server URL
2. Click Start button and authenticate
3. Open Copilot Chat and use:

   @github List all open issues
   @github Create issue: Bug fix needed
   @github Show repository information
   

🚀 Deploy to Production

Railway.app (Recommended)

1. Visit Railway.app
2. Create new project from GitHub repo
3. Name service: mdabul-project
4. Add environment variables
5. Deploy! My URL: https://mdabul-project.railway.app/mcp/

Docker

docker build -t vane-mcp-server .
docker run -p 3000:3000 \
  -e GITHUB_TOKEN=$GITHUB_TOKEN \
  -e MCP_ENABLED=true \
  vane-mcp-server

Heroku

heroku create mdabul-mcp-server
git push heroku main
heroku config:set GITHUB_TOKEN=ghp_xxxx

📖 Complete Documentation

See MCP_DEPLOYMENT_GUIDE.md for:

• Detailed setup instructions
• Multiple deployment options
• Troubleshooting guide
• Testing procedures
• Cross-project usage examples

🧪 Testing

Test Locally

# Server status
curl http://localhost:3000/mcp/

# List issues
curl -X POST http://localhost:3000/ \
  -H "Content-Type: application/json" \
  -d '{"tool": "list_issues", "parameters": {"state": "open"}}'

Test Production

curl https://mdabul-project.railway.app/mcp/

Test with Copilot

In VS Code Copilot Chat:

@github What are the open issues?

🔄 Reuse Across Projects

We can use this MCP server in any GitHub project:

1. Copy .vscode/mcp.json to your project
2. Update .env with your repo credentials
3. Start using: @github commands in Copilot Chat

One server. Unlimited projects. 🎯

📋 Checklist

• [x] MCP server implementation
• [x] GitHub tool integration
• [x] Vane agent framework integration
• [x] Local testing
• [x] Production deployment guide
• [x] VS Code configuration
• [x] Documentation
• [x] Railway URL configured

🛠️ Architecture

Core Components

config.py - Centralized configuration

• Workspace isolation via VANE_ROOT_ID
• MCP server settings
• LLM orchestration parameters
• Confidence thresholds
• Logging configuration

mcp_server.py - HTTP MCP Server

• MCPRequestHandler - HTTP request processing
• MCPServer - Server lifecycle management
• Tool execution engine
• GitHub integration interface

Vane Agent Framework

• Semantic caching with Chroma DB
• Confidence-based filtering
• Structured workspace management

🔐 Security

• GitHub token stored in .env (never committed)
• HTTPS enforced in production
• Request validation on all endpoints
• Error handling prevents information leakage

📊 Logging

Logs stored in: workspaces/{VANE_ROOT_ID}/logs/vane_agent.log

Configure verbosity:

LOG_LEVEL=DEBUG DEBUG_MODE=true python mcp_server.py

🤝 Contributing

This is a production MCP server. To extend:

1. Add new tools in MCPRequestHandler.CAPABILITIES
2. Implement handler methods: handle_<tool_name>()
3. Update documentation
4. Test locally before deploying

📝 License

MIT License - See LICENSE file for details

🔗 Resources

• Model Context Protocol
• GitHub Copilot Extensions
• GitHub API Reference
• Railway Documentation

📞 Support

For issues or questions:

1. Check MCP_DEPLOYMENT_GUIDE.md troubleshooting section
2. Review logs: tail -f workspaces/*/logs/vane_agent.log
3. Test connectivity: curl https://mdabul-project.railway.app/mcp/

---

My MCP Server: https://mdabul-project.railway.app/mcp/

Built by MD ABUL HOSSAIN (myou260312-eng) | 2026

Copyright © 2026 MD ABUL HOSSAIN (MCP SERVER). All Rights Reserved.