workflow-completion-hooks

You can use event-driven completion hooks to receive a reliable callback exactly when your workflow finishes. This requires registering a per-workflow hook function that fires automatically with status and timing data, eliminating the need for polling loops.

When a workflow finishes, the hook fires with a completionInfo structure containing WorkflowId, Status, TerminationReason, Mode, ElapsedSec, Steps, FinalMarking, and EndTime. This structured data provides full context for downstream processing.

Yes, you can register a hook after completion. The system supports Immediate firing for hooks registered after a workflow reaches a final state, ensuring you still receive the completion callback without missing the event.

Yes, ClaudeOrchestrator supports completion-driven orchestration across multiple parallel workflows. You can register hooks for each workflow id to coordinate downstream processing exactly when each parallel workflow finishes.

Hook exceptions are isolated by the system to ensure safe execution semantics. This prevents your callback errors from disrupting the workflow completion process or interfering with other registered hooks.

You need to unregister the per-workflow hook function. The system ensures one-time firing per workflow id, so unregistering after the hook fires prevents duplicate callbacks and cleans up the registry.