DeepMem
A self-hostable AI memory layer, API-compatible with Mem0, that stores conversations as searchable long-term memory using a FastAPI service in front of a Qdrant vector store.
DeepMem is an AI memory layer that turns conversations into searchable long-term memory, exposing the same MemoryClient class and method signatures as Mem0 so an existing integration can switch by changing only its import line. It combines a FastAPI HTTP API, a Qdrant vector store, LLM-based fact extraction, and hybrid retrieval that blends vector search, BM25, entity boosting, and time-decay, along with semantic caching and asynchronous batched distillation.
It runs in an open mode with no API key or user registration required, so it can be deployed for a team's own agents directly, with multi-tenant isolation handled through a user ID in each request body; a built-in MCP server and GDPR-oriented deletion controls are also included. A managed cloud version at deepmem.dev exposes the same Mem0-compatible API for teams that prefer not to self-host, and the two are meant to be interchangeable from the client's point of view.
It is written in Python, MIT licensed, and installs as a client package with a companion server component.