yergin/local-mcp-hub

Local MCP Hub

A portable MCP (Model Context Protocol) hub that runs locally on development machines while connecting to a remote Ollama server. Provides AI coding assistance with access to local code analysis and documentation tools.

Architecture

Dev Machine                    Remote Server
┌─────────────────────────┐    ┌─────────────────┐
│ Local MCP Hub           │    │ Ollama Server   │
│ ├── Serena (Code)       │◄──►│ qwen2.5:latest  │
│ ├── Context7 (Docs)     │    │ Port 11434      │
│ └── OpenAI API (3002)   │    └─────────────────┘
└─────────────────────────┘
         ▲
         │
┌─────────────────────────┐
│ Continue Extension      │
│ VS Code                 │
└─────────────────────────┘

Features

• 🔧 Local Code Analysis: 18 Serena tools for semantic code understanding
• 📚 Documentation Search: 2 Context7 tools for up-to-date library docs
• 🌐 Remote AI Processing: Connects to remote Ollama server for model inference
• 🔌 Continue Integration: OpenAI-compatible API for VS Code Continue extension
• 📦 Auto-Installer: One-command setup that downloads and configures all dependencies
• 🖥️ Cross-Platform: Works on Mac, Windows, and Linux
• ⚡ Portable: Easy deployment across multiple development machines

Available Tools

Serena (18 tools)

• list_dir, find_file, symbol_overview, find_symbol
• get_symbol_definition, list_symbols_in_file, find_references
• replace_symbol_body, search_for_pattern, read_file_content
• get_workspace_overview, search_symbols_in_workspace
• get_class_hierarchy, find_implementations, get_function_calls
• analyze_dependencies, find_similar_code, extract_interfaces

Context7 (2 tools)

• resolve-library-id, get-library-docs

Quick Start

Prerequisites

• Node.js (v18+)
• Python 3.8+
• Git
• Remote Ollama server running qwen2.5:latest

Installation

1. Clone the repository:

   git clone <repository-url>
   cd local-mcp-hub
   

2. Run the installer for your platform:

   Linux:

   ./install.sh
   

   macOS:

   ./install-mac.sh
   

   Windows:

   install.bat
   

3. Configure the Ollama server connection:

   Edit config.json and update the Ollama host to match your server:

   {
     "ollama": {
       "host": "http://YOUR_OLLAMA_SERVER:11434",
       "model": "qwen2.5:latest"
     }
   }
   

4. Start the hub:

   npm start
   

5. Configure Continue extension in VS Code:

   models:
     - name: "Local MCP Hub + Qwen2.5"
       provider: openai
       model: "qwen2.5:latest" 
       apiBase: "http://localhost:3002/v1"
       apiKey: "dummy-key"
   

Configuration

Environment Variables

• PORT: Override default port (3002)
• Example: PORT=3003 npm start

Config File (config.json)

{
  "ollama": {
    "host": "http://10.0.0.24:11434",
    "model": "qwen2.5:latest"
  },
  "hub": {
    "port": 3002,
    "log_level": "info",
    "cors_origins": ["*"]
  },
  "mcps": {
    "enabled": ["serena", "context7"]
  }
}

Usage Examples

Ask Continue these questions to see the MCP tools in action:

• "Can you analyze the LocalMCPHub class in my codebase? Show me its methods."
• "What files are in my src directory and how is the project structured?"
• "How does the sendToOllama method work and what error handling does it have?"
• "Find documentation for the latest React hooks API" (uses Context7)

API Endpoints

• GET /health - Health check
• GET /v1/models - Available models
• POST /v1/chat/completions - OpenAI-compatible chat endpoint
• GET /v1/tools - List available MCP tools

Development

Project Structure

local-mcp-hub/
├── src/
│   └── hub.ts              # Main hub implementation
├── mcps/                   # Auto-downloaded MCPs
│   ├── serena/            # Code analysis toolkit
│   └── context7/          # Documentation search
├── install.sh             # Auto-installer
├── config.json           # Configuration
└── continue-config.yaml   # Continue extension template

Running in Development

npm run dev     # Start with ts-node
npm run build   # Build TypeScript
npm test        # Run tests

Logs

Check local-mcp-hub.log for detailed logs including:

• MCP tool usage
• Ollama communication
• Continue extension requests

Troubleshooting

Port Already in Use

PORT=3003 npm start

Continue Extension Not Connecting

1. Ensure hub is running: curl http://localhost:3002/health
2. Check Continue config uses correct port
3. Verify Continue extension is reloaded

MCP Tools Not Working

1. Check mcps/ directory exists with serena and context7
2. Run ./install.sh again to re-download
3. Verify Python virtual environment: mcps/serena/.venv/

Ollama Connection Issues

1. Update config.json with correct Ollama server address:

   {
     "ollama": {
       "host": "http://YOUR_OLLAMA_SERVER:11434",
       "model": "qwen2.5:latest"
     }
   }
   

2. Test direct connection: curl http://your-server:11434/api/tags
3. Ensure qwen2.5:latest model is installed on server
4. Check firewall settings allow access to port 11434

Contributing

1. Fork the repository
2. Create a feature branch
3. Make changes
4. Test on multiple platforms
5. Submit a pull request

License

MIT License - see LICENSE file for details.

Acknowledgments

• Serena - Semantic code analysis toolkit
• Context7 - Documentation search MCP
• Ollama - Local language model server
• Continue - VS Code AI coding assistant