MCP Memory libSQL

MCP Memory libSQL is an MCP server that provides persistent graph and vector memory for AI assistants using libSQL. It connects local Model Context Protocol clients, such as Claude Desktop or Cline, directly to local SQLite files or remote libSQL database instances like Turso. Software engineers and AI developers use this server to give LLM agents cross-session memory without relying on external cloud vector databases. By managing both an entities table with embeddings and a relations table, the server supports structured semantic recall and knowledge graph traversal. AI clients can store facts, link entities through relations, and retrieve context using libSQL's built-in vector similarity search operations. It eliminates state loss between individual chat sessions while supporting secure token authentication for distributed deployments.

Category: AI Memory & Context

Tags: embeddings, libsql, persistence, sqlite, vector-search

Visit MCP Memory libSQL

How to install and configure MCP Memory libSQL

1. Configure your MCP client settings file (such as Cline or Claude Desktop). 2. Add the server entry with the command npx and arguments ["-y", "mcp-memory-libsql"]. 3. Configure the environment variables to point to your storage location. For a local database, set LIBSQL_URL to file:/path/to/your/database.db. If left blank, it defaults to file:/memory-tool.db in the working directory. 4. For remote instances like Turso, set LIBSQL_URL to your remote database URL (e.g., libsql://your-database.turso.io) and supply LIBSQL_AUTH_TOKEN with your authorization token. 5. Restart your MCP client to initialize the libSQL persistent memory store.

What you can do with MCP Memory libSQL

• Storing project facts and architectural decisions across coding sessions so AI coding assistants retain workspace context. * Querying stored entities through libSQL vector search to retrieve relevant notes based on semantic similarity rather than exact keywords. * Building dynamic knowledge graphs by creating and deleting directional relationships between stored user, code, or task entities. * Centralizing shared agent memory across multiple development machines using a remote Turso database instance with secure token authentication.

Key facts

• Open Source
• https://github.com/joleyline/mcp-memory-libsql
• AI Memory & Context, Databases & Data Stores
• embeddings, libsql, persistence, sqlite, vector-search

Part of MCP Servers

What can MCP Memory libSQL do?

MCP Memory libSQL provides persistent memory for MCP clients by implementing standard memory interfaces alongside vector search capabilities. It allows AI models to create, update, delete, and search entities using vector embeddings, as well as create and query relationships between entities. It can store data in local SQLite database files or remote libSQL servers.

How do I install MCP Memory libSQL?

You can run MCP Memory libSQL directly via npx by specifying npx with the arguments -y and mcp-memory-libsql in your MCP client configuration. You also need to configure the LIBSQL_URL environment variable to point to a local file or a remote database.

Which MCP clients work with MCP Memory libSQL?

The server works with any standard Model Context Protocol client. The documentation explicitly provides configuration examples for Cline and Claude Desktop, including instructions for running under Windows Subsystem for Linux environments using WSL and Node.js.

What database does MCP Memory libSQL require?

The server uses libSQL as its backing store. It supports both local SQLite files referenced with file: URIs and remote libSQL deployments such as Turso databases via libsql: URIs with token-based authentication. If no URL is set, it defaults to a local file.

Is MCP Memory libSQL open source?

Yes, MCP Memory libSQL is open source software. It is released under the MIT License, and its source code is publicly hosted on GitHub.