Agent run state, pending tool approvals, and output available independently of the creation request.
| Field | Type | Required | Description |
|---|---|---|---|
run_id |
str | ✔️ | ID of the run, preserved across approval continuations. |
agent_id |
str | ✔️ | ID of the agent that owns the run. |
state |
models.PlatformAgentRunState | ✔️ | State of the persisted workflow execution. REQUIRES_INPUT is nonterminal. |
created_at |
date | ✔️ | Timestamp when the run was first persisted. |
updated_at |
date | ✔️ | Latest persisted activity timestamp across the execution and its messages, including checkpoint messages. This is not a worker heartbeat or an exact state-transition timestamp. |
expires_at |
date | ➖ | Execution deadline, when one is set. Not the run history retention deadline. |
pending_interactions |
List[models.PlatformAgentRunToolApproval] | ✔️ | Unanswered tool approvals when the run is in REQUIRES_INPUT. Empty in other states. |
output |
Dict[str, Any] | ➖ | Accumulated text-only agent output under the messages key, when available. |
error |
Optional[models.PlatformAgentRunError] | ➖ | Caller-safe details of an agent run execution failure. |