LanceDB MCP Server

Example Use Cases

• 1

  Search and summarize a local document collection using vector search (LanceDB) to support RAG workflows.

• 2

  Fetch relevant document chunks on demand to minimize LLM token usage and provide precise context.

• 3

  Run entirely local/offline document indexing and retrieval for privacy-sensitive applications.

Description

This MCP server exposes a serverless LanceDB-backed vector index and a document summary catalog so LLMs can look up relevant documents and chunks on demand. It provides catalog and chunk search tools (catalog_search, chunks_search, all_chunks_search), a seed script to create indexes, and local development instructions. The design emphasizes efficient token usage by letting the LLM fetch only what it needs and supports fully local/offline operation for improved privacy and security.