Watch and recover your workflow runs
Find any run, read its status and step timeline, and recover from failures from the Monitor group of your workspace sidebar.
Where monitoring lives
The Monitor section of your workspace sidebar holds Runs and Usage & Cost. Activity and Triggers are tabs inside Workflows. There is no separate monitoring dashboard, so the run list at /[workspaceSlug]/monitoring/workflow-runs is where you start most of the time.
The Runs list shows every execution across your workflows: its Run ID, the workflow that produced it, status, start and completion times, and duration. From there you open a single run to read its step-by-step timeline, or you pause, resume, and cancel a run that is still in flight.
The run detail page polls for updates every 10 seconds while a run is pending or running, and
stops once the run reaches a terminal status. It also pauses polling in background browser tabs,
so a completed run that looks static is not frozen. Use the Refresh button to pull the latest
state at any time.
Find and read a run
Start in the run list, narrow it down, then open the run that interests you.
Filter the run list
The list has two filters: a free-text Run ID field and a Status dropdown (Running, Completed, Failed, Paused, Cancelled). See Find and filter runs.
Open one run
Click a Run ID to open its detail page: the overview metadata, a progress bar, the trigger payload, and the Execution Steps timeline. See Read a run.
Search across runs
When you know what passed through a run but not which run it was, press ⌘K and search the text of step inputs, outputs, and trigger payloads. See Search across runs.
Recover from a failure
When a run fails, the detail page surfaces the first failed step. The timeline shows a Jump to failed step link and auto-expands that step, where you read the error code, the upstream provider status, and whether the error is retryable.
Read the error code, provider status, and retryable flag on the failed step, then decide whether to retry.
Walk a failure end to end in Debug a failed run.
Pause, resume, or cancel a run. Which action is available depends on the run's current status.
Start a new run from the step that failed, or from the start, on your newest published version. Retry is available only for failed runs.
Approvals
Approvals are decided on the run they belong to. When approvals are pending, the workspace sidebar shows how many on the Runs item, and choosing Waiting for approval in the Status filter lists those runs. Open a run and the waiting step shows the approval with Approve… and Reject…, the action it gates, and when it expires. For an AI agent's tool call, that is the agent's step, with one approval per waiting call. A run that is waiting on an approval does not advance until someone decides.
What you cannot do here
The run list filters by Run ID and status only. It does not offer date, cost, or model filters, and the run list and detail pages do not have a cost breakdown or a live log tail. The Status dropdown also omits two real statuses, pending and waiting, even though both appear in step counts on the run detail page. To narrow runs by start time, sort or scan the Started column directly.
Next steps
- Find and filter runs to locate a specific execution.
- Read a run to understand the overview and step timeline.
- Debug a failed run to recover after a failure.