Deploys now say why they failed
October 18, 2026 · platform
A failed deploy used to show an error string and leave you to work out who had to act. It now says which it is.
If your agent throws an error while it loads, the deploy fails with Fix needed in your agent and links to the agent logs where the error is. The new version never takes calls, so the previous version keeps serving them. If the cause is on our side, the deploy says Problem on our side and offers Open a ticket. The button starts an email to support with the ids and times we need to find your deploy, and it leaves out your logs, your code and your secrets. You can read the email before you send it.
Problems that pass by themselves, such as a shortage of capacity, are retried without you. The deploy shows Retrying (attempt 3 of 9) until it works or the last attempt fails.
A broken agent is caught while the deploy loads it. A project that scales to zero between calls gets a short one-off check first, so those deploys take a few seconds longer.
Try it
- Docs: When a deploy fails, with every failure code and what to do about it, and How a deploy checks your agent
- Dashboard: VoiceThere dashboard → your project → Deployments → open a failed deploy
- API:
failureonGET /api/v1/deployments/{id}carries the class, the code and the message