orchestrator

Cross-provider agent orchestration coordinates background tasks across different providers by providing durable task IDs and status tracking while enforcing a strict safety profile that defaults to read-only execution.

Safe multi-provider agent tasks enforce read-only execution by default and require explicit approval before initiating any write operations, ensuring strict security and error handling for complex workflows.

Yes, cross-provider orchestration manages durable background tasks specifically designed for scenarios requiring collaboration between different services, such as integrating third-party APIs or handling cross-provider data synchronization.

To launch a bounded read-only task, run the command `orchestrator launch codex --name "<task name>" "<bounded read-only task>" to initiate safe execution with status tracking.

Cross-provider orchestration is needed for complex workflows that exceed the capabilities of native subagents, particularly when coordinating durable, long-running background tasks across multiple service providers.

The safety profile limitations include restricted runtime models and a strict default to read-only operations, meaning any write operation requires explicit approval before execution within the orchestrated workflow.