LMStudio-MCP
@infinitimeless
A Model Control Protocol (MCP) server that allows Claude to communicate with locally running LLM models via LM Studio.
概要
What is LMStudio-MCP?
LMStudio-MCP is a Model Context Protocol (MCP) server that bridges Claude with locally running LLM models via LM Studio. It is for developers who want to use private, offline AI models through Claude’s interface.
How to use LMStudio-MCP?
Install LMStudio-MCP via one-line script, manual Python setup, Docker, or Docker Compose. Configure it with environment variables for host/port, then add the appropriate MCP configuration to Claude. Start LM Studio with a loaded model, and Claude will be able to call the server’s tools.
Key features of LMStudio-MCP
- Check LM Studio API health
- List and identify loaded models
- Generate chat and raw text completions
- Create vector embeddings for semantic search
- Maintain stateful multi-turn conversations
- Start persistent sessions with a locked system prompt
Use cases of LMStudio-MCP
- Private/offline inference without sending data to external APIs
- Semantic search and RAG workflows via embeddings
- Multi-turn conversational agents with persistent context
- Local coding assistant with consistent behaviour
- Privacy-first document analysis with accurate summarisation
FAQ from LMStudio-MCP
What are the dependencies for LMStudio-MCP?
Python 3.7+, a running LM Studio instance with a model loaded, MCP-compatible Claude client, and the requests, mcp[cli], and openai Python packages.
Where does LMStudio-MCP process data?
All inference happens locally on your machine via LM Studio; no data is sent to external services.
Are there any version requirements for certain tools?
create_response, start_conversation, and continue_conversation require LM Studio v0.3.29+. generate_embeddings requires an embedding-specific model (e.g., text-embedding-nomic-embed-text-v1.5).
