spences10/mcp-memory-libsql

View on GitHub ↗

🧠 High-performance persistent memory system for Model Context Protocol (MCP) powered by libSQL. Features vector search, semantic knowledge storage, and efficient relationship management - perfect for AI agents and knowledge graph applications.

87 ★18 forksTypeScriptUpdated 24d ago

What you need to know

High-performance, persistent memory system for MCP backed by libSQL with ranking-aware text search for entities, observations, and relations.

Install

npx -y mcp-memory-libsql
Add {"command":"npx","args":["-y","mcp-memory-libsql"],"env":{"LIBSQL_URL":"file:/path/to/database.db"}} to mcpServers

Usage

  • Set LIBSQL_URL to a local SQLite file or a remote libSQL (Turso) URL
  • For remote databases also set LIBSQL_AUTH_TOKEN
  • Default: file:/memory-tool.db in the current directory

Key features

  • Persistent storage of entities and relations
  • High-performance text search with relevance ranking and fuzzy matching
  • Knowledge graph management with relationship queries
  • Flexible text search scoped to name, type, and observation fields
  • Local SQLite and remote libSQL (Turso) compatibility with token authentication

Best for

Adding durable, searchable memory and a knowledge graph to an MCP assistant with minimal setup

Caveats

  • Requires an LIBSQL_URL pointing to a database (local or remote)
  • WSL users must use Linux-style filesystem paths
  • npm publishing is manual due to 2FA requirements
Platforms: macOS · Linux · WindowsClients: Claude Desktop · Cline
Documentation ↗

Reviewed 2026-08-11

Topics

Stars
87★
Forks
18
Language
TypeScript
License
MIT
Created
2025-01-14
Last push
2026-08-11