Skip to main content
Every time a workflow starts, HITL creates an execution. It records the input, every step it passed, every decision people made and every action it took.

Status

What an execution holds

A finished run of the refund workflow, from GET /executions/:id:
  • result: the last decision. decisions: every review step’s answer, with the reviewer’s name (never their email) and where they answered (mobile, web_portal, telegram, …).
  • action_results: what each action step got back, starting with the status code.
  • trail: the path the run took.
  • triggered_via: api, webhook, email, schedule, manual or test.
In my.hitl.sh, the same run is on the workflow’s Executions tab:
The Executions tab listing runs marked Waiting for review and Completed, with one completed run expanded to show its trigger data and the webhook action's response

Getting the result

GET /executions/:id?wait=30 answers as soon as the run has finished, or after 30 seconds with the run as it is (max 60). Call it again until finished is true. One call covers up to a minute, so this is cheap.
The CLI’s hitl run <id> --wait and the MCP tool get_execution with wait do this for you.

Limits

  • Input: up to 64 KB of JSON per run.
  • /workflows and /executions routes: 1,000 requests an hour per API key.
  • wait: up to 60 seconds per call.