Agents overview
What agents are, how they authenticate, and how they show up in your workspace.
Last updated
Agents are long-lived API keys that let automated tools write notes on behalf of your team. Each agent gets its own name, colour, and attribution so notes it produces are clearly marked as agent-authored — not confused with human activity.
Creating an agent
- Open Settings → Agents (workspace admins and owners only).
- Click New agent.
- Pick a name, description, and colour.
- Copy the token that appears — it's shown once. Treat it like a password.
The token starts with pw_agent_. Use it in an Authorization: Bearer <token> header when calling the API.
Attribution
Notes written by an agent show the agent's name and colour wherever authorship is displayed, including the note header, activity feed, and created by / last edited by properties.
Trust levels
By default, every new agent is gated: its writes land in the pending actions inbox until you approve them. Flip the Auto-approve switch next to the agent to let it write directly, only do this for agents you fully trust.
Rotating and revoking
- Rotate issues a new token and invalidates the old one immediately. Use when a token might have leaked.
- Revoke stops the agent from authenticating at all. Notes it already wrote stay attributed to it.