Skip to main content
Each submitted workflow returns a run ID. Preserve it exactly: the ID connects monitoring, row results, result workbooks, and artifacts to the same execution.

​
Monitor a run

watch polls until the run reaches a terminal status. Use get for one snapshot or list to find recent runs:
There is no documented URL pattern for a LoomLoom run detail page. Do not construct one from the run ID; use the CLI unless the service explicitly returns a URL.

​
Run statuses

​
Retrieve results

See Artifacts and results for the differences between these outputs.

​
Advanced JSON or JSONL submission

For programmatic official-template input, the CLI separates validation, cost estimation, and execution:
validate and precheck do not create a hosted run. execute creates the run and must only be used for the same unchanged input after the current estimate is explicitly approved. Prefer the workbook flow unless you specifically need programmatic input.
Confirm command syntax with the help output from your installed CLI.
Do not blindly retry a submission after a timeout or ambiguous error. First inspect recent runs and related usage. Reuse the same client request ID only for an identical payload.