One memory, many agents.
Persistent memory that's shared across sessions and across agents. It outlives resets, restarts, and devices — one key serves every agent, and humans can inspect, import, and export what's stored. The context an agent earned in one run is there for the next.
Memory that follows the work, not the session.
Most agents forget everything when the session ends. mem9 keeps a persistent store that any agent can read and write with one key — so memory survives a reset, moves between devices, and is shared across the different agents on a task. Humans stay in the loop: inspect it, import into it, export from it.
# mem9 — one memory, many agents # • outlives resets, restarts, and devices # • one key serves every agent # • humans inspect / import / export # works with the agents you already run: Claude Code · OpenCode · Codex OpenClaw · Hermes · Dify
Persist. Share. Inspect.
A single memory your whole agent fleet — and the humans behind it — can rely on.
Outlives the session
What an agent learns is written to durable memory that survives resets, restarts, and a switch of device. Pick up a task days later with the context intact.
One key, many agents
Every agent on a task reads and writes the same memory through one key — so a hand-off between agents doesn't start from zero.
Humans stay in the loop
Memory isn't a black box. Inspect what's stored, import existing knowledge, and export it whenever you need to move or audit it.
Native to the agents you already run.
mem9 plugs into the runtimes and tools your agents use today.
Memory, composed with the stack.
Pair mem9 with compute, a shared filesystem, data, and observability — one agent-native stack.
Give your agents a memory that lasts.
mem9 is live. Use-first, claim-later — no signup to start. Install the CLI or point your agent at the sys9 skill.