mnemos-setup

mnemos Setup

Persistent memory for AI agents. This skill helps you set up mnemos with any agent platform.

Prerequisites

You need a running mnemo-server instance. See the server README for deployment instructions.

Step 1: Deploy mnemo-server

cd mnemos/server
MNEMO_DSN="user:pass@tcp(host:4000)/mnemos?parseTime=true" go run ./cmd/mnemo-server

Step 2: Provision a tenant

curl -s -X POST http://localhost:8080/v1alpha1/mem9s | jq .
# → { "id": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx", "claim_url": "..." }

Save the returned id.

• For OpenClaw, this is the value you should store as apiKey (preferred).
• Legacy OpenClaw config can still store the same value as tenantID, but the plugin will still use v1alpha2.
• For Claude Code / OpenCode env vars, this remains the tenant ID value used by the current server API.

Step 3: Configure your agent platform

Pick your platform and follow the instructions:

---

OpenClaw

Add to openclaw.json:

{
  "plugins": {
    "slots": { "memory": "mem9" },
    "entries": {
      "mem9": {
        "enabled": true,
        "hooks": {
          "allowConversationAccess": true
        },
        "config": {
          "apiUrl": "http://localhost:8080",
          "apiKey": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx"
        }
      }
    }
  }
}

Restart OpenClaw. You should see:

[mem9] Server mode (v1alpha2)

Compatibility note:

• Preferred config: apiKey -> plugin uses v1alpha2 with X-API-Key.
• Legacy config: tenantID -> plugin treats it as an alias for apiKey and still uses v1alpha2.
• OpenClaw 4.23+ / 2026.4.22+ requires plugins.entries.mem9.hooks.allowConversationAccess = true so mem9 can read conversation messages in agent_end and upload them. Older OpenClaw builds that reject this hook policy should omit the hooks block and upgrade for full automatic conversation upload.
• The underlying value is the same UUID either way.