dispatching-parallel-agents

You coordinate parallel agent workflows by spawning subagents for quick isolated tasks or persistent teammates for long-running pipelines. They stay coordinated through a Task system that manages owner statuses and shared outcomes using inbox-based messaging to synchronize results.

The best way to orchestrate multi-agent code reviews is by dispatching parallel agents for different review areas. This reduces handoffs and wasted waiting time by managing task dependencies and synchronizing shared outcomes across persistent teammates.

You manage dependencies between parallel agents using a dedicated Task system. It tracks owner status and task dependencies while leveraging inbox-based messaging to ensure independent workflows stay coordinated through shared outcomes.

Yes, you can use short-lived subagents for fast, isolated tasks within a larger pipeline. For multi-stage deployments, they work alongside persistent teammates to handle quick execution while the Task system coordinates overall dependencies and shared outcomes.

You should use persistent teammates instead of short-lived subagents when coordinating long-running pipelines or large feature rollouts. Teammates support ongoing collaboration across complex multi-step workflows, whereas subagents are designed for fast, isolated tasks.