Skip to main content

Troubleshooting Workflow Runs

Identify the workflow surface first

The canvas, connected-step and inserted-output checks below apply to classic workflows. Saved automations created through Chat are a separate surface: check that automation’s saved tools, required inputs and run status instead of looking for a classic canvas.

For either surface, review completed steps and saved outputs before continuing. If a send or update has an unclear result, check the external system before repeating it.

Read the exact error to identify a credit restriction, provider throttle or stalled task. Each needs its own recovery step.

Troubleshooting Workflow Runs

Use this guide when a workflow fails, produces an unexpected result, skips context from a prior step, generates the wrong file, uses unexpected credits, or behaves differently from a previous run.

First checks

  • Confirm the workflow: make sure you are running the intended workflow, not an older copy or a shared template.

  • Check step connections: a later step can only use an earlier step output when the steps are connected and the prior output is inserted into the later prompt.

  • Check required inputs: confirm every required user input, file, URL, source, and tool setting is present.

  • Check the run history: identify which steps completed, which step failed, and whether a generated file or export was created.

  • Check trigger context: for scheduled, webhook, or event-triggered workflows, confirm the trigger fired with the inputs you expected.

Common causes

  • Prior output is missing: the steps may not be connected, or the prior step output may not have been inserted into the later step prompt.

  • Tool did not run: the tool may not be selected, authenticated, available in that surface, or supported by the selected model.

  • Generated file is missing or cannot be downloaded: the run may have completed text output but failed during file creation, export, or storage.

  • Document processing output is wrong: the template mapping, placeholder names, source file, or prior step output may not match what the document-processing step expects.

  • Credits were used before failure: completed AI, file-processing, or tool work can consume credits even if a later step fails.

When behavior may be expected

  • Workflow steps run as separate pieces of work. Later steps need the relevant context from earlier steps passed into them.

  • A workflow can complete some steps before a later step fails.

  • A model can produce different wording on repeat runs, even with the same prompt.

  • Some tools, integrations, files, or generated outputs may require the right permissions, connection status, model capability, or workspace setting.

Collect before contacting support

  • Workflow name and link, if available.

  • Run link or visible run ID, if available.

  • Run time and timezone.

  • Step that failed or produced the unexpected result.

  • Completed steps before the failure.

  • Expected output and actual output.

  • Model selected, tools selected, files used, and trigger type.

  • Business impact, including affected users or tenants, whether production work is blocked, urgency or deadline, and whether there is a workaround.

  • Screenshot of the workflow canvas, run history, error message, generated-file panel, or export/download issue.

  • Whether the issue reproduces with the same inputs.

Share visible workflow, run, and generated-file links when available, plus timestamp, screenshot, and error text. If the issue happened through the API, include the response X-Request-ID.

When to contact Support

  • The same workflow used to work and now fails with the same setup.

  • A documented workflow feature does not behave as described.

  • A generated file, export, or tool call fails even though the settings appear correct.

  • The issue affects multiple users, tenants, workflows, or production work.

  • There is no acceptable workaround and the issue blocks customer work.

Save repeatable work as an automation

After resolving the current run, ask Chat to convert one classic workflow, review the proposal and test its inputs, tools and outputs. Keep the original until the replacement works. See migration guidance.

Reviewing the classic Cedar workflow conversion to an automation.

Watch: convert a classic workflow to an automation

Inspect the latest automation run

For a saved automation, open Workshop → Automations → Activity and choose Open run. Activity shows the latest run for each automation, not a complete run history. In the chat, choose View steps to check the version and completion status, then select a step to inspect its status and timing.

If a step says No details from this run, return to the chat to review the original request and answer. Compare the result with your inputs before using it.

Watch: inspect an automation run

Inspect a completed Cedar Weekly Brief run, then compare the answer with the fictional project notes. This example does not demonstrate a failed run or retry.

Correct a failed Scrape step

Open Preview and read the failed step’s message under Job Run Results. In Cedar Page Check, entering a page name produced “Missing prerequisite valid URL.” The user entered the full HTTPS address and selected Generate. The new run returned the expected sign-in page text. Run History shows both the earlier Error and the later Complete entry.

Did this answer your question?