Retrieve a task run
Returns a single task run by ID, including its status, input, output, errors, and interaction state.
Authorizations
Secret key (sk_live_...)
Path Parameters
Unique identifier for the task run (prefixed with trun_).
Query Parameters
Comma-separated list of additional fields to include in the response. Supported values: input, storage, artifacts. Sections are omitted from the response (not returned as null) when their flag is not present.
Response
OK
A task run object representing a single execution of a task.
Unique identifier for the task run, prefixed with trun_.
Always task_run.
Run status: queued, running, canceling, interaction_required, review_required, completed, canceled, or failed.
Run result after completion: success, failure, or unknown. Null while running.
The task that was executed. Prefixed with task_.
The credential used for authentication, if any. Prefixed with cred_.
The session this run executed within, if any. Prefixed with sess_.
The agent that executed this run. Prefixed with agt_.
The source this run connected to, if any. Prefixed with src_.
Execution time in milliseconds.
Output data produced by the task execution.
Error details if the run failed. Same structure as API-level errors.
Interaction details when the run is paused waiting for user input (e.g. MFA code, security question).
ISO 8601 timestamp of when the resource was created.
ISO 8601 timestamp of when the resource was last updated.
Unique identifier for the API request. Include this when contacting support.
Input data provided when the run was created.
Names of input fields whose values were tokenized when the run was created. The listed fields are removed from input. Returned alongside input (only when include=input is requested and one or more fields were vaulted); omitted otherwise.
Files captured during the run, each with inline extraction (if any). Returned only when include=storage is requested.
Run-level visibility artifacts (screenshots). Returned only when include=artifacts is requested.