Memobase
Open-source user-profile memory that gives AI apps long-term memory of their users.
About
Memobase is an open-source backend that adds long-term user memory to generative AI applications. Instead of relying only on embedding retrieval or RAG, it builds structured user profiles plus event timelines from conversations, so an app can remember, understand, and adapt to each user over time.
It is designed for production: per-user buffer zones batch memory writes to cut LLM costs, retrieval runs on SQL over generated indices rather than pure vector search, and it ships with REST APIs plus Python, Node, and Go SDKs. It also exposes an MCP server so coding agents can save, search, and read user profiles directly.
What it can do
- Structured user profiles with topics and sub-topics
- Event timelines for episodic memory
- Buffered writes for low cost and latency
- REST API, SDKs, and MCP server
Setup
Run the self-hosted backend with Docker (default at localhost:8019), then install the SDK with pip install memobase or the Node/Go equivalents.
Prerequisites
Docker for the backend; an LLM API key for memory extraction.
Use cases
Works with
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 116; Trending score 70/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.