Skip to main content
When a run doesn’t do what you expected, the Runs tab has everything you need: a run-level banner for run-wide problems, and per-step detail for everything else.

​
Reading a failed run

  1. Check the run banner. Run-wide problems — out of credits, a credit limit stop, skipped records — appear as a notice above the graph with the specifics (for a credit-limit stop: the budget, what was spent, and which step it stopped before).
  2. Find the failed step. The run graph highlights each executed step with its status. Click a failed step to open its detail panel — the error message, an expandable stack trace, and the inputs the step received.
  3. Check per-record outcomes. A step can fail for some records and succeed for others. The detail panel lists each processed contact/company with its outcome, so you can see exactly which records were affected.
Failed vs. Stopped vs. Completed: a run is Failed only when nothing progressed at all. If some records made it through, the run Completes — with the per-record failures visible in the step details. Stopped means the run halted itself deliberately, most commonly at a run credit limit.

​
Common errors

​
Retries and re-running

  • Transient failures retry themselves. Infrastructure hiccups are retried automatically with backoff; you’ll rarely see them. Business failures (the table above) do not retry — they need a fix.
  • There is no retry button. To process again after fixing the cause, start a new run. With “Process each contact only once” enabled, the new run skips records the previous run already handled — so re-running doesn’t double-process or double-message anyone.
  • Cancelling a running run stops new work; steps already dispatched may still finish and spend their credits.

​
How runs spend credits

  • Credits are reserved per step, per record, just before the work happens, and released again if the step doesn’t complete. Steps that use external data providers or AI are the ones that cost credits — each step’s reference entry says whether it does.
  • AI Agent steps are billed by actual usage as the agent works, attributed to the run.
  • The run credit limit (Settings) is checked before every reservation: a run never starts a step that would push it over budget — it stops instead, showing budget and spend on the run.
  • Where to look: the run itself shows credits spent (and the budget, when a limit is set); your workspace balance and overall usage live under Usage in the app.
For recurring scheduled workflows, the combination of “Process each contact only once” and a run credit limit is the safest default: no repeated spend on the same records, and a hard ceiling per run even if a source suddenly returns far more records than usual.