Search documentation

Search for a topic, a control, or an event type.

    Docs / Workflows / Logs

    Logs

    Logs records what each run of a workflow did: which jobs ran, the events they received and published, their messages, model and tool calls, duration and cost.

    Where logs appear

    In the workflow builder, the Logs button beside Run workflow opens a panel under the canvas. After a run in the builder, the panel opens on the first job that needs attention: a failed job first, then a waiting, skipped or running one. While the panel is closed, the Logs button changes colour when a run finishes with a failed or skipped job. Full page gives the panel the whole screen.

    Show Results in a job’s panel opens Logs on that job. Show job in Logs opens the job’s configuration. A link to a workflow’s logs opens the workflow in the builder with the linked run showing.

    Log Details is a page for one recorded job, opened from links such as the runs listed on an event’s page. It shows the job name, status, start and finish times and the executed workflow version, with the rest of its run beside it. Its actions are Open in workflow, Re-run step and Explain in chat. Runs explains what a re-run does. The Runs tab in Activity lists runs across all workflows.

    The run list

    Logs opens on the list of the workflow’s runs, including test runs. Each row shows when the run started and how it began, such as Triggered run or Test run. It also shows the run status, a summary, the duration and the cost. The summary names the job where a run stopped or reports that all steps completed. Its second line reports an emitted output event or the number of completed steps.

    Live Feed
    On by default. The list adds new runs and updates running ones while the panel is open. Turning it off freezes the list and adds a Refresh button.
    Status filter
    All statuses, Succeeded, Failed, Running, Rejected or Timed out.
    Date filter
    Last 7 days, Last 30 days, Last 90 days or All time. The list starts at Last 30 days.

    Pointing at a row shows that run’s job statuses on the canvas. Selecting it opens the run. Runs explains what a run is and what each status means.

    A run and its jobs

    An open run shows its overall status and start time above a Steps list. Each step is one job execution, with its status, start time and a Re-run badge when it came from a re-run. A re-run also links to its original execution. Pointing at a step highlights its job on the canvas. Selecting a step opens the job detail, which has three tabs.

    Execution
    The status, execution time and cost of the job. When the job did not succeed, a box states the reason, such as the first sentence of the error. Execution logs below list what happened in time order. Files the job produced appear under Artifacts.
    Input & output
    The input event the job received and the output event it published, with their properties and files. A job that published several output events has a selector to choose one. No output event means the job published nothing.
    Cost
    Total cost, total tokens and the number of model calls. Token usage separates input, output, cached input and thinking tokens. Cost by resource lists each priced resource. Cost is recorded separately and can appear a moment after the job finishes.

    Execution logs mark each model call with an AI badge and show the model, token counts, duration and cost, and the AI input, AI output, Thought and Final answer text of the call. Each tool or integration call has its status, Tool input and Tool output, or Tool error when it failed. Other entries are messages from the action, code or system.

    A waiting Human Approval job shows its pending approval on the Execution tab, where you can decide it. Approvals explains that decision. A Wait for Event job shows its deadline and the events it waits for.

    Explain in chat

    Explain in chat appears for a job that failed, timed out, was rejected, was skipped or did not run. It starts a new conversation in Auto mode and asks why the job ended that way. In the builder the chat opens beside the canvas. On Log Details it opens on the Chat page.

    The chat receives the job, its execution log and its input event as context. When a test run saved no execution log or input event, the message carries those details itself: the job and action type, the status and reason, up to 4,000 characters of input properties, the error and the last 20 log entries.

    Troubleshooting

    A job shows Filters not met
    The job received the event, but the event did not match its Filters. Compare the input event on Input & output with the Filters, or run the test with another event.
    A job shows Event mismatch
    The test event has a different type from the event the job listens for. Choose a matching event with Change event, or change the job’s trigger.
    A job shows Did not run
    In a test, no event reached the job because the job before it published no output event. Select the earlier job in Steps to see whether it failed, was skipped or has no output event. In a live run, Did not run also marks a job whose Filters did not match the event.
    No Triggered run appears after a live event
    The workflow is a draft or paused, so events start nothing. Choose Go live for a draft or Resume for a paused workflow. Workflows explains both states.
    A property is missing in Input & output
    Only properties configured in the earlier job’s output event reach the next job. Add the value to that job under New properties from this action or Forwarding properties.
    An integration write did nothing in a test
    The job succeeded and its output event holds example values, but the connected service did not change. Write was off in the run’s Permissions. Turn it on to send real changes.
    A job Failed
    The box on the Execution tab gives the first sentence of the error, and Execution logs show the entry that failed, such as a Tool error. Fix the job and run it again, or use Explain in chat to ask for a diagnosis.