Analyze a failed workflow run
Explains why a run failed, in plain language.
This only works on runs whose status is failed. Anything else returns a 400.
The first call runs a language model over the run’s steps and its definition, then stores the result, so later calls for the same run return the stored text without paying for it again.
Check that explanation is non-empty before showing it. If the model is unavailable you still get a 200, with explanation and generated_at both empty rather than an error.
This endpoint is limited to 5 requests per minute.
Authorizations
Personal API key.
Path Parameters
ID of the app whose workflows you want to work with.
ID of the workflow, as returned in id by List workflows.
ID of the run, as returned in run_id by List runs for a workflow.
Response
The explanation. Check that it is non-empty.
AI-generated failure analysis.
Why the run failed, in plain language. Empty when the model could not be reached.
"The email step failed because the recipient address was missing from the trigger payload."
When the explanation was produced, as an ISO 8601 timestamp. Empty when none was produced.
"2026-08-25T09:20:11Z"