MemU is an open-source memory system that gives all of your AI agents one shared, personal memory. Coding assistants forget everything between sessions and tools; MemU fixes that by continuously distilling what your agents learn, your preferences, conventions, and reusable skills, into a readable wiki-style memory that any connected agent can recall. The project has earned more than 14,000 stars on GitHub and is under active development, with updates through late 2026.
MemU works across sessions, agents, and devices. A background process watches your agent activity, captures knowledge worth keeping, and organizes it as inspectable Markdown with links back to its sources, so you can see exactly what your agents remember and correct it. When a new task starts in a different tool, MemU retrieves the relevant memories and skills first: a coding convention learned in Codex stays available when you switch to Cursor or Claude Code. Host adapters cover a wide range of popular agents, including Claude Code, Codex, Cursor, OpenClaw, Hermes, and WorkBuddy, on both desktop and chat surfaces. Despite the scope, the core memory logic is deliberately tiny, around 500 lines, making the system easy to audit, adapt, and self-host. Setup is agent-driven: you get an API key, hand your agent the setup instructions, and it configures itself.
MemU is free for personal use, including cross-device sync through its hosted service, and the core is open source. Paid options exist for teams and enterprise deployments that need custom memory systems and private infrastructure, making it a freemium product.
MemU is for heavy agent users who hop between Claude Code, Cursor, Codex, and other tools and are tired of re-explaining their preferences and project conventions. If your agents keep forgetting how you work, MemU gives them a common, transparent memory to share.