An agent missed its run
What you're seeing
A finding like:
Cruz bot has shown no activity for 31h; its owner expects it at least every 24h (last seen 2026-10-26T08:12Z)
It is a Missed run finding. On Home the source-by-category grid counts it in the Workflows at risk row for the source the agent belongs to. The same row counts rising error rates and latency.
What it means
Someone, usually the agent's owner, told Provantix how often this agent should run, and it has been quiet for longer than that. This is a plain comparison of the agent's last activity with the interval you set. It is not a forecast, and it does not say why the agent stopped. The cause can be in the workflow, in the emitter that reports it, or in nothing at all if the schedule changed.
Severity is medium once the agent is late and high when it has been quiet for three times its interval, because at that point it is no longer a delay.
Nothing is measured for an agent with no interval set, an agent never seen, or an agent that is proposed, paused or retired. An agent that is meant to be quiet never raises this.
What to do
- Tell the workflow failing from the reporting failing. Check the source's row on Connectors and look for an emitter silent finding or batch gaps. If the emitter is silent, the workflow may be fine and only its reporting is broken.
- If the workflow stopped, find out why in the system that runs it (the scheduler, the host, a credential that expired) and restart it. The finding closes by itself on the next run once the agent shows activity.
- If the schedule changed on purpose, open the agent, choose Expected to run, and set the new interval, or choose Stop measuring if it should no longer be watched.
- Resolve the finding with the cause if you want it on the record. It reopens if the agent goes quiet again.
Setting the interval
Open the agent on the Agents page, choose Expected to run, and enter a whole number of hours from 1 to 720 (30 days). Setting it is written to the admin audit log.
Choose an interval a little longer than the real gap between runs. A workflow that runs every day at 09:00 should be set to 26 or 30 hours, not 24, so a run that starts a few minutes late does not raise a finding.
What good looks like
Every agent that is meant to run on a schedule has an interval, and the finding opens only when something has really stopped. See Findings for how findings open and resolve.