Mcp Automem
MCP client for AutoMem — give Claude, Cursor, Codex, and other MCP tools durable graph+vector memory across conversations.
AutoMem stores memories in both a graph for relationships and a vector index for meaning. AutoMem runs in the background to consolidate new memories into a retrieval graph. Agents connect to AutoMem through the MCP protocol.
All deployment options for AutoMem expose the same MCP endpoint. On the BEAM benchmark, AutoMem achieves 57.4% accuracy at 10 million tokens. AutoMem uses only 2.6 to 4.8 thousand context tokens per answer regardless of scale.
Summary drafted from the project's own website. Every sentence is backed by text on that page and was reviewed before publishing.