Start here
Almost every problem is diagnosed the same way: open Agents, select your Agent, expand the execution and read the list of actions — or click Read chat in the Activity log. The Agent explains what it was doing at every step. See Agent executions and activity log.
I do not see Agents in the menu
Agents must be enabled for your workspace. Contact your Crono representative.
My Agent never runs
Check it is not disabled. A greyed-out card means Disable was used; switch it back on from the ⋯ menu.
Check you are in Associated users. An Agent only runs for the people on that list.
Trigger-based: check the trigger really matches. An Updated trigger limited to one field only fires when that field changes, and — if you set values — only for those values.
Trigger-based: it fires for the owner of the record. If a colleague owns the contact and is not associated with the Agent, nothing happens.
Scheduled: it runs at the chosen slot in each user's own time zone, not yours.
Check your credits — see below.
My Agent stopped and is asking me something
That is Supervised mode: it pauses on its main actions and waits for Confirm or Reject. See Supervised and Autonomous Agents.
It asks for confirmation at the wrong moment
The action Crono treats as the Agent's main one is worked out from your description and instructions. Rewrite the first lines so the outcome you want is unmistakable, and save again — every save re-decides it.
It says it has run out of credits
The Agent has hit its own monthly AI credit limit, or the workspace has used its allowance. Raise the limit in the editor, or wait for the refresh date shown on the card. See Agents and AI credits.
It consumed far more credits than expected
Almost always a missing stop condition, or unbounded web research. Open the execution and look at how many actions it took, then add an explicit stop line and a research boundary — see How to write effective instructions for your Agent.
It did nothing at all
The criteria were not met and your instructions said to do nothing. Check the chat of the execution to confirm.
Or the action you wanted was never described, so the Agent was not allowed to take it. An Agent only uses actions your instructions imply: if you never mention searching the web, it will not.
It worked on the wrong record
Scheduled Agents get no record handed to them. If the instructions do not say which records to work on, the Agent chooses. Name the list or the filter.
Never refer to “the contact I have open” in the saved instructions — there is no screen during automatic runs. In the chat, you can refer to the page you have open.
It created a duplicate list
Asking the Agent to create a list if it does not exist can produce a new one each run, especially with a common name. Create the list yourself, give it a unique name, and name it in the instructions.
I replied to the Agent in the chat and it started a new run
Replying in the chat starts a new execution rather than continuing the one waiting for you. To resolve an execution that is pending approval, open the Agent's executions and use Confirm or Reject there.
Still stuck?
Note the Agent's name and the date and time of the execution, and contact support. Those two details let us find the exact run.