Cloud memory, however you work.
Your claude-mem observations DB, synced to the cloud — on your own, in your own isolated container, or rolled out to your whole team.
What teams build for months. Yours in one command.
A real memory layer is a database, a vector index, a sync service and an MCP server. cmem is all of it — open-source engine, hosted cloud sync.
Questions, answered
What is claude-mem?
claude-mem is an open-source memory plugin for Claude Code and other AI agents. While one AI works, a second AI takes notes about what it does — building a searchable memory that persists across every session.
Is my data private?
Yes. claude-mem is local-first: your observations live in a SQLite database on your own machine, with privacy stripping applied as memory is generated. Cloud sync is strictly opt-in.
What's an MCP link?
An MCP link is a personal MCP server URL that gives any MCP client — Claude Code, Cursor, Codex CLI, and more — secure access to your memory from anywhere.
Does it slow my agent down?
No. Observation generation runs out-of-band, so it doesn't block your agent. Thanks to progressive disclosure, sessions actually get cheaper as your memory grows.
What is RAD?
RAG is how agents retrieve knowledge. RAD is how they remember experience. RAD — Retrieval-Augmented Dreaming — is our term for live observation generation that captures what happened while you work. Dreaming ships in claude-mem today; RAD as an open standard is on the way.
What does cloud cost?
CMEM Pro is 7 days free, then $30/month for solo developers — memory generation runs on our observer model instead of your Claude plan, and every observation syncs to the cloud behind one private MCP link. Team Cloud is $333 per seat per month with FDE onboarding.