Medha

•
14
•

Provides persistent, git-versioned memory storage for large language model applications, enabling long-term memory capabilities.

Medha is a Model Context Protocol (MCP) server designed to equip LLM applications with robust, long-term memory. It achieves this by combining Git's powerful version control with graph-based memory associations, ensuring every piece of information is persistent and fully auditable. This allows LLMs to remember context, track changes over time, and understand complex relationships between data points, enhancing their ability to engage in extended, informed interactions.

主要功能

01Multi-database support for SQLite (development) and PostgreSQL (production)
02Graph associations to link memories with typed relationships
03Flexible authentication with local mode or SAML 2.0
04Powerful search capabilities, including semantic search with OpenAI embeddings
05Git-backed storage for complete memory history
0611 GitHub stars

使用案例

01Building knowledge graphs and traversing memory associations for AI agents
02Equipping LLMs with long-term memory for sustained conversations and tasks
03Providing secure, versioned, and searchable context for AI applications
04Integrating persistent memory into MCP-compatible LLM clients like Cursor or Claude Desktop