Why do AI agents stop without telling me?
Agents lose their place, loop on the same page, hit a wall they can't describe, or run out of time or usage, and the chat doesn't always say so.
How can I check whether my agent finished?
- ›Look at the result itself: the file, the record, the sent message. Don't rely on "done".
- ›Open the agent's activity or task view if it has one, and look for the last step.
- ›Ask the agent to list every step with what happened, and to say plainly which step failed.
- ›For tasks on websites, check the site's own history (orders, uploads, sent items).
What should a run log show?
Five things: the status (running, waiting, done, failed), each step in order with whether it worked, the last step reached, the reason it stopped in plain words, and who or what started it. Without the reason, a failure looks the same as a success that never reported back.
What does runnerID record?
runnerID logs every fill, code and approval in an append-only log you can read in the owner console, and one switch revokes it all. See /guides/human-approval-ai-agent-actions.
When your agent hands runnerID a step it can't finish, which is coming to runnerID, every handed-off step gets a status and a step log. When a run fails or needs you, both your agent and you are told, with the last step reached and a plain reason, such as "the site blocked our browser", "the login failed" or "you declined the payment step". A run can fail; it won't fail silently.
See plans →Questions
My agent said it was done, but nothing happened. Why?
Some agents report the plan as if it were the result. Check the output itself, and ask for each step with its result.
Why does my agent loop on the same page?
Usually it can't get past something on the page, such as a login, a code or a form it can't fill, and keeps retrying. Look at the step it keeps repeating; that's the wall.
Can I get told when a task fails?
Some agents send notices; many don't. Coming to runnerID: when a step your agent handed off fails or waits on you, runnerID emails you, and you can switch those emails off per identity.
Does a log mean my data is used for training?
No. Your data is yours. We never use your content to train AI models: not your records, documents, emails, passwords or codes, and not what your agents read or type on the sites they visit. We do learn from how tasks get done, such as the steps a site's sign-in takes or where it asks for a code, with your content and personal details removed, so agents get better at the work. You can turn this off any time in your settings, and we never sell it or share it for anyone else to train on.
Related guides
The four walls: credentials, codes, blocked browsers and the agent's own limits.
Read →Staff, delegate and service provider seats, and how to take access back.
Read →Site rules, approval by email link, the log and revoke.
Read →