@ai-gentik/lm-studio-mcp

MCP server that bridges Claude to LM Studio for local model inference via HTTP API.

Installation

npm install @ai-gentik/lm-studio-mcp

Usage

With Claude Code

Add to your user scope (available globally):

claude mcp add --transport stdio lm-studio "npx @ai-gentik/lm-studio-mcp" --scope user

Or to a specific project:

cd your-project
claude mcp add --transport stdio lm-studio "npx @ai-gentik/lm-studio-mcp" --scope project

Direct

npx @ai-gentik/lm-studio-mcp

Requirements

• Node.js 18+
• LM Studio running locally (default: http://localhost:1234)
• LM Studio API server enabled

Configuration

Set environment variables to customize:

• LM_STUDIO_API_URL — LM Studio API endpoint (default: http://localhost:1234/v1/)
• LM_STUDIO_TIMEOUT — Request timeout in milliseconds (default: 300000)
• LM_STUDIO_CHAR_LIMIT — Max characters in response (default: 8000)

LM_STUDIO_API_URL=http://localhost:1234/v1/ npx @ai-gentik/lm-studio-mcp

Tools

• generate_completion — Text generation from a prompt
• chat_completion — Chat-based conversation
• list_models — List available models
• load_model — Load a model into memory
• unload_model — Unload a model from memory

Development

npm install
npm run build
npm start

Build

npm run build

Outputs bundled server to dist/index.js ready for distribution.

License

MIT