Monitoring Runs
Last updated:
Every time an agent task executes, Oscova records a run. Run history helps you confirm what happened, troubleshoot failures, and decide whether a task needs adjustment.
Note
Run means one execution of an agent task, including status, timestamps, result summary, error text, token count, and stored conversation messages when available.
Read the task directory
The Agent Tasks page shows summary metrics for total tasks, running tasks, completed runs, and failed or attention-needed work. Each row shows task name, agent, model, schedule, latest run, enabled state, and actions.
Status filters include running, completed, failed, orphan, never run, and disabled. An orphan task is linked to an agent that no longer exists.
Pin task history
Click a task row to open its details. The details view shows task summary fields, the current state, and recent executions. Use View all or View runs to browse a larger recent run list.
Task outcome notifications can deep-link back to the task and focused run. When you open a notification for a task run, Oscova can select that task and load more recent runs so the reported run is visible when it is still in recent history.
Inspect a run
Click Inspect when a run has stored messages. The inspector shows the full agent conversation for that execution, which is usually the best evidence for why the task succeeded, failed, or skipped work.
When a task fails, read the first model, tool, credential, or workspace error before changing the schedule. A missing credential or blocked tool usually needs configuration, not a different interval.
Respond to failures
- Fix provider or credential settings when the run fails before useful work starts.
- Adjust the instruction if the agent asks for missing details.
- Review tool approvals when automation blocks a tool call.
- Reduce schedule frequency if tasks run longer than their interval.
- Use Run now only when the task is not already executing.
- Reassign orphan tasks when the backing agent was deleted.
- Check webhook API key scope and allowed slugs when an external trigger never starts a run.
- Check application logs when a run completed but its outcome notification did not arrive.
Tip
A failed one-shot task remains enabled but does not retry automatically. Fix the failure and use Run now to retry it.
Live versus historical state
The row status tells you what is happening now. Run history tells you what happened before. During long-running work, wait for the current run to finish or inspect the active messages before editing the task.