lance-mcp
56
LanceDB MCP Server for LLMS is designed to enable large language models to dynamically search and interact with locally stored documents through LanceDB. It offers efficient token usage and ensures data security by storing the index locally.
🗄️ LanceDB MCP Server for LLMS
A Model Context Protocol (MCP) server that enables LLMs to interact directly with documents on-disk through agentic RAG and hybrid search in LanceDB. Users can ask LLMs questions about datasets or specific documents.
✨ Features
- LanceDB-powered serverless vector index and document summary catalog.
- Efficient use of LLM tokens, where the LLM looks up necessary data itself.
- Security by local index storage, preventing data transfer to the Cloud when using a local LLM.
🚀 Quick Start
Local directory creation and configuration in the Claude Desktop config file are required.
Prerequisites
- Node.js 18+
- npx
- MCP Client (e.g., Claude Desktop App)
Seed Data
The seed script generates tables in LanceDB for document summaries and vectorized document chunks.
📝 License
This project is licensed under the MIT License.