MCP.so
Sign In

mcp-lance-db: A LanceDB MCP server

@kyryl-opens-ml

About mcp-lance-db: A LanceDB MCP server

No overview available yet

Basic information

Category

Databases

Runtime

python

Transports

stdio

Publisher

kyryl-opens-ml

Config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "mcp-server-lancedb": {
      "command": "uv",
      "args": [
        "sync"
      ]
    }
  }
}

Tools

2

Adds a new memory to the vector database

Retrieves semantically similar memories

Overview

What is mcp-lance-db?

mcp-lance-db is a Model Context Protocol server that stores and retrieves memories in the LanceDB vector database. It acts as a semantic memory layer for LLM applications, allowing text to be stored with vector embeddings for later retrieval.

How to use mcp-lance-db?

Configure it in your MCP client (e.g., Claude Desktop) by adding the command uvx mcp-lance-db to the client’s config file. For Claude Desktop, the config file is claude_desktop_config.json. The server then exposes two tools: add-memory and search-memories.

Key features of mcp-lance-db

  • Adds new text memories with vector embeddings
  • Searches memories by semantic similarity
  • Configurable result limit (default 5)
  • Embeds using sentence-transformers (BAAI/bge-small-en-v1.5)
  • Stores data locally in a LanceDB database
  • Notifies clients of resource changes

Use cases of mcp-lance-db

  • Provide a persistent semantic memory for conversational AI assistants
  • Enable retrieval of past context or facts by meaning, not keywords
  • Build a lightweight personal knowledge base for LLM tools
  • Store and recall user preferences or session history semantically

FAQ from mcp-lance-db

What does mcp-lance-db do?

It is a basic MCP server that stores text as vector embeddings in LanceDB and retrieves memories that are semantically similar to a query.

What tools does mcp-lance-db provide?

It provides two tools: add-memory (takes a required content string) and search-memories (takes a required query string and optional limit parameter).

What are the default configuration values?

Database path is ./lancedb, collection name is memories, embedding model is BAAI/bge-small-en-v1.5 on CPU, and the similarity threshold is 0.7.

How do I run mcp-lance-db for debugging?

Use the MCP Inspector: run npx @modelcontextprotocol/inspector uv --directory $(PWD) run mcp-lance-db and open the provided URL in a browser.

How is mcp-lance-db installed?

It is distributed as a Python package and can be run via uvx mcp-lance-db. No manual installation is needed if using uv.

Comments

More Databases MCP servers