automem

Why we included this project

If your assistant keeps forgetting what you decided last week, AutoMem is built to fix that. It's a self-hosted service that stores decisions, preferences, and project context in a graph-plus-vector index, so a later session can retrieve not just similar text but the reasoning behind it: the choice made, the alternatives weighed, and the work that followed. Recall runs straight against the index with no generative-LLM call in the middle, which keeps it fast and avoids an extra charge every time the assistant needs context. Embeddings still come from providers like Voyage or OpenAI. An MCP bridge lets it plug into Claude, Cursor, Codex, and similar tools rather than replacing them. If you're building agents that need durable, relational memory across sessions, this is a practical option.