Create Run
Create a new run.
The workflow must exist and belong to the authenticated organization. If machine_id is not provided, an available machine will be automatically selected. The run will be created with SCHEDULING status and a Temporal workflow will be started asynchronously.
Authorizations
Bearer authentication header of the form Bearer <token>, where <token> is your auth token.
Headers
Unique key for idempotent requests. If provided, the server ensures the request is processed at most once. Retries with the same key return the original response. SDKs auto-generate this for write requests.
"550e8400-e29b-41d4-a716-446655440000"
Body
Schema for creating a run
Optional per-run prompt override. When provided, this run executes the supplied prompt text instead of the workflow's stored main_prompt.
1Optional per-run main agent ModelConfiguration.id override. Stored as run model_metadata.main_agent_model_id and used only for this run.
Optional per-run model configuration metadata. Provided fields override the workflow's model_metadata for this run only.
Machine ID. If not provided, an available machine will be automatically selected.
Pool IDs to filter available machines. Machine must belong to all of these pools (intersection). Ignored when machine_id is provided.
Input values for workflow variables
Files to upload to the machine
Sensitive input values (supports nested objects). These are not stored and will be written to a secure vault and referenced by alias only. In workflow input_schema, sensitive root keys are validated under a '$' prefixed key (for example input key 'api_key' is '$api_key').
Join an existing session; overrides machine_id/pool_ids
Start a new session on the machine used by this run. Session ID will be this run's ID.
Persist outputs under this alias for $ref in this session
Release the session after this run completes successfully
Response
Successful Response
Run response schema
scheduling, running, running_checks, success, cancelled, task_failed, error Per-run prompt override used for execution. Null means the run falls back to the workflow's current main_prompt.
Per-run model metadata overrides used for execution. Null means the run falls back to the workflow's model_metadata.
Arbitrary usage/billing metadata captured during a run (schema is flexible)
Version-tolerant snapshot/results for the run's post-run checks.