Lancelot MCP Server
Vector search and RAG over LanceDB for Claude and other AI assistants. Local-first, type-safe, no cloud retrieval middleman.
Free.
What it does
Lancelot turns any AI assistant into a RAG-enabled partner with vector search backed by LanceDB.
Dual-mode search
Vector search using OpenAI embeddings with automatic text-search fallback when embeddings aren't available.
RAG query engine
Context-aware retrieval with budget management. Pulls only the chunks the assistant actually needs.
LanceDB-native
Direct integration with LanceDB instances. Local-first storage, columnar format, embedded or remote.
Secure credentials
API keys stored in the OS keystore via keytar — Windows Credential Manager, macOS Keychain, Linux Secret Service — with an encrypted-config fallback when a system keychain isn't available.
MCP compatible
Works with Claude Desktop, Claude Code, Cline, and any other MCP-compatible AI assistant. Drop-in stdio server.
Type-safe + production-ready
Full TypeScript implementation with strict mode, runtime validation, structured logging, and graceful error recovery.
Give your AI assistant a memory
Vector search and RAG over your private data, with credentials in the OS keystore. TypeScript, MCP-native.
Download Free