Automem

AutoMem is an open source long term memory service for AI assistants and coding agents. It gives tools such as Claude, Cursor, Copilot, Codex and ChatGPT a shared memory that survives between sessions, saving decisions, preferences, notes and context so an agent stops asking about your stack and conventions every time it starts. Agents connect to it through the Model Context Protocol. It is built for developers and power users who work across several AI tools and want one memory behind all of them. AutoMem stores each memory in two layers: a FalkorDB knowledge graph that records entities, typed relationships and time, and a Qdrant vector index that captures meaning. Recall combines both, so asking why a decision was made can return the decision, the alternatives considered and the work that followed. Retrieval does not call a generative model, which keeps lookups fast and avoids extra charges; embeddings come from Voyage, OpenAI or a local provider. A background consolidation process clusters related memories, strengthens connections you revisit and lets noise decay. A local MCP bridge serves desktop tools, while remote MCP over HTTPS connects cloud agents such as ChatGPT Developer Mode, Claude.ai and ElevenLabs. A REST API is also exposed. AutoMem is free under the MIT license. You can run it locally with Docker, on your own infrastructure or Kubernetes, or as a Railway service group, and it supports macOS, Linux and Windows through WSL2. No hosted subscription pricing is published; running costs are your own infrastructure and embedding provider. The project publishes its benchmark methodology and raw results openly so they can be reproduced.