Agent Profiles
Click any card on the AI Agents screen to open its profile. The profile has two views: Operator (for most users) and Developer (for power users and admins).
Operator view
The operator view answers practical questions about the assistant.
| Section | Tells you |
|---|---|
| Status | Healthy, Needs attention, or Paused. Last run timestamp. Next scheduled run. |
| Recent activity | Last 20 events: signals detected, responses proposed, recordings ingested. |
| What it watches | The signal types it can detect and the threshold rules it relies on. |
| Specialist chain | For supervisors: which specialists it routes to and in what order. |
| Effective policies | What this assistant is allowed to do per workflow, after organisation rules and overrides. |
| Cooldown | When it last ran and when it can run again. |
This view is the right place for “why didn’t it catch X?” or “why did it act on Y?”. Recent activity and effective policies together explain almost everything.
Developer view
The developer view exposes the underlying configuration.
| Section | Shows |
|---|---|
| Slug | Stable identifier used in policies. |
| Trigger | Schedule (cron expression) or event type. |
| Cooldown | Minimum gap between runs. |
| System prompt | The assistant’s role description (read-only for managed assistants). |
| Tool access | Which specialists or services this assistant can call. |
| Memory store | Link to the assistant’s pattern memory. |
| Delivery channels | Where this assistant sends its output (Delivery Channels). |
Common operations
- Pause. Stops the trigger. In-flight runs complete.
- Run now. Manual trigger, useful for debugging cadence issues.
- Override autonomy tier. Change the assistant’s default policy bundle.
- Reset cooldown. Clear the cooldown timer (admin only).