@samscarrow/lmstudio-mcp

LM Studio MCP Server with concurrent multi-agent support and memory safety features.

Quick Start

# Run directly with NPX (recommended)
npx @samscarrow/lmstudio-mcp

# Or install globally
npm install -g @samscarrow/lmstudio-mcp
lmstudio-mcp

Requirements

• Node.js 16+ - To run this NPM package
• Python 3.7+ - For the MCP server itself
• LM Studio - Running with a model loaded

The Python components are automatically downloaded and set up on first run.

Configuration

Environment Variables

# LM Studio API endpoint (default: http://localhost:1234/v1)
export LMSTUDIO_API_BASE="http://localhost:1234/v1"

# Logging level
export LOG_LEVEL="INFO"

Command Line Options

# Custom API base
npx @samscarrow/lmstudio-mcp --api-base http://192.168.1.100:1234/v1

# Custom log level
npx @samscarrow/lmstudio-mcp --log-level DEBUG

# Show help
npx @samscarrow/lmstudio-mcp --help

# Show version
npx @samscarrow/lmstudio-mcp --version

Features

• 🚀 Concurrent Multi-Agent Support - Handle multiple requests simultaneously
• 🛡️ Memory Safety - Intelligent memory management and limits
• 📊 Real-time Monitoring - Track performance and resource usage
• 🔄 Auto-retry Logic - Robust error handling and recovery
• 🎯 Load Balancing - Distribute requests across available resources
• 📈 Scaling Support - Handle increased load automatically

Programmatic Usage

You can also use this package programmatically in your Node.js applications:

const LMStudioMCP = require('@samscarrow/lmstudio-mcp');

// Create instance with custom options
const server = new LMStudioMCP({
    apiBase: 'http://localhost:1234/v1',
    logLevel: 'INFO'
});

// Start the server
server.start()
    .then(() => console.log('Server started'))
    .catch(console.error);

Troubleshooting

Python Not Found

If you get Python-related errors:

# Install Python 3 (macOS with Homebrew)
brew install python3

# Install Python 3 (Ubuntu/Debian)
sudo apt update && sudo apt install python3 python3-pip

# Windows - download from python.org

LM Studio Connection Issues

• Ensure LM Studio is running
• Check that a model is loaded in LM Studio
• Verify the API endpoint is correct (default: http://localhost:1234/v1)
• Try using 127.0.0.1 instead of localhost

Port Already in Use

The MCP server uses stdio communication, not HTTP ports. If you see port errors, they're likely from LM Studio itself.

Installation Locations

Files are installed to:

• macOS/Linux: ~/.lmstudio-mcp/
• Windows: %USERPROFILE%\.lmstudio-mcp\

Development

# Clone the repository
git clone https://github.com/samscarrow/lmstudio-mcp.git

# Install dependencies  
npm install

# Run locally
node bin/lmstudio-mcp.js

License

MIT - see LICENSE file for details.

Support

• 📖 Documentation: https://github.com/samscarrow/lmstudio-mcp
• 🐛 Issues: https://github.com/samscarrow/lmstudio-mcp/issues
• 💬 Discussions: https://github.com/samscarrow/lmstudio-mcp/discussions

Related

• LM Studio - Local LLM runtime
• Model Context Protocol - MCP specification
• Claude Code - AI coding assistant