OpenMemory

It enables AI applications including Claude desktop, custom agents, and chatbots to store, retrieve, and update structured memories across sessions, giving LLMs access to a personal knowledge base that persists beyond individual context windows without transmitting personal information to external servers.

The system stores memories as structured semantic records with automatic deduplication, contradiction resolution, and relevance scoring, so retrieving memories for a given query surfaces the most contextually relevant information rather than flooding the model context with everything ever stored.

OpenMemory exposes an MCP (Model Context Protocol) server interface, making it compatible with Claude's tool use system and other MCP-compatible AI assistants that can query and update the memory store as part of their reasoning loop.

OpenMemory is open-source and designed for privacy-conscious users who want AI assistant memory functionality without the security implications of cloud storage.

Running locally, it processes all memory operations on-device, making it appropriate for storing sensitive personal information, professional context, and proprietary knowledge that should not leave a user's machine.

The project is compatible with standard vector databases for semantic retrieval and provides a web interface for viewing, editing, and managing stored memories directly.