The list
Every deployment you’ve created, with its current state and how many of your allowed running agents are in use. Deploying from inside a Space takes you here too — the surface is the same one.A single deployment
Opening one shows:- Recent runs — what it’s been asked to do and how each ended.
- Last reported — when the agent last checked in.
- Its run URL and the run key, masked.
- Health — whether its model and connector credentials are still good.
”Last reported” is not a heartbeat
The time shown is when the agent last told us it was healthy. Nothing pings its URL to check. A recent timestamp means it was working then; it is not a live liveness probe, and the page says so rather than showing a green dot that implies more than it knows.Re-reading the run key
The key shown once at create can be read again from the deployment detail screen. It is masked by default.Redeploying
Changed the agent? Update its.vio, then redeploy onto the same deployment. The URL
and the run key survive, so anything already calling it keeps working.
Stopping
Stop tears the agent down and frees its slot against your limit.When something’s wrong
Runs fail with an authorisation error — a connector’s access expired or was revoked. Use Reconnect on the deployment to re-authorise; the agent picks it up without redeploying. Runs fail immediately on a tool — an API key the agent’s tools need wasn’t supplied at deploy time. The deployment reports which one. Stuck not-yet-running — a deploy that never finished. Cancel it to release the slot.Deploy an agent
Build a
.vio and put an agent on its own URL.