OpenClaw Plugin. Personal memory, auto-wired.

Install

openclaw plugins install @octamem/octamem-openclaw

Configure

Get your API key from platform.octamem.com, then:

openclaw octamem configure <your-api-key>

or set the environment variable and run:

export OCTAMEM_PERSONAL_KEY=<your-api-key>
openclaw octamem configure

Restart OpenClaw after configuring.

What it does

• Auto-recall — Before each turn, relevant personal memory is fetched and injected so the LLM can answer from your notes.
• Auto-capture— After each turn, the assistant’s reply is stored in your personal memory (plain text).
• Tools — octamem_get (search) and octamem_add (add) for the agent to query or store memory when needed.
• CLI — openclaw octamem status | search <query> | details | add <content>.

Use only OctaMem (disable local/file memory)

openclaw octamem configure automatically adds tools.deny to your OpenClaw config so only OctaMem is used. No manual config edit needed.

If you ever need to set it yourself, add to ~/.openclaw/openclaw.json under "tools":

"tools": {
  "profile": "coding",
  "deny": ["memory_search", "memory_get"]
}

Restart OpenClaw after changing config.

Optional config

• memories.personal.apiKeyAPI key (supports ${OCTAMEM_PERSONAL_KEY}).
• toolDescriptionOverride the description for the get-memory tool.

What you actually get

• Continuity. An agent that talked to you yesterday remembers your preferences today, without you plumbing context through your own state machine.
• Decisions that stick. When the agent commits to an action, that fact lands in personal memory automatically. No more asking the same thing twice.
• One memory, many surfaces. The same OctaMem account backs Claude, Cursor, and OpenClaw, your notes follow you across clients.