memU

memU, made by NevaMind AI, provides persistent memory infrastructure for AI agents across multiple sessions, devices, and applications, letting agents retain and recall context, preferences, and decisions without users re-explaining information each time. It offers a shared memory layer accessible across different AI agents including Claude, Codex, Cursor, OpenClaw, Hermes, and WorkBuddy, source-linked and inspectable memory records stored in readable Markdown format, and cross-agent context retrieval so information learned in one tool is available in another. memU includes automatic skill extraction and memory storage, and agent-driven memory decisions where agents themselves choose what is worth remembering. The project maintains an open-source repository, an active Discord community, and regular blog updates, with a recent redesign streamlining its implementation to a compact 474-line core. Pricing is freemium: the core product is 100% free with login required to copy integration instructions, while enterprise deployments use custom, negotiated design-partner engagements.