A context engine that gives agents persistent memory, user profiles, and hybrid RAG
Brings memory extraction, user/project profiles, contradiction- and time-aware forgetting, hybrid RAG + memory, and connectors together in a single context engine. Its choice between the convenience of hosted MCP and fully self-hosted operation with a local binary, Ollama, and Docker makes it more flexible than SaaS-only services. The README reports strong retrieval results on benchmarks such as LongMemEval and LoCoMo.
People who want to connect personal preferences, project knowledge, and document RAG in a single profile.
Hosted MCP sends your code and personal context to a third party; for sensitive projects, use local mode and disable telemetry. The project reports its own benchmark results.
https://mcp.supermemory.ai/mcp npx supermemory local