MemOS
An open-source AI agent memory system stores context on-device or on your servers, with local plugins for OpenClaw, Hermes Agent and DeepSeek Harness.
MemOS gives AI agents persistent memory across conversations and tasks. Developers can use it for assistants that recall preferences, support history or earlier work, with on-device, self-hosted and hosted storage options.
- Local agent memory: Plugins for OpenClaw, Hermes Agent and DeepSeek Harness retain context in SQLite on your machine, combining full-text and vector search with duplicate removal.
- Reusable task experience: The local plugin turns retained task experience into skills and provides a Memory Viewer for inspecting what the agent remembers.
- Editable memory: The service API lets applications add, retrieve, edit and delete graph-based memories, including text, images, tool traces and personas, and correct stored information through natural-language feedback.
- Shared or isolated context: Separate knowledge bases let users, projects and agents keep memories apart or share selected context.
MemOS is open source under Apache 2.0. The local plugin supports macOS, Linux and Windows and requires Node.js plus a supported agent runtime; its memory storage has no cloud dependency. The self-hosted service runs on your machine or servers, including through Docker, and uses Neo4j and Qdrant.
The hosted API requires an account and API key and stores memory in MemOS Cloud. Cloud plugins for OpenClaw and DeepSeek Harness retrieve context and save new messages through that hosted service.