dispatching-parallel-agents

Parallel task delegation works by identifying independent sub-tasks with no shared mutable state or output dependencies, then dispatching them concurrently using the delegate tool. This mechanism executes self-contained operations simultaneously, reducing total processing time for parallel research or data gathering.

Dispatch parallel agents when you have two or more independent tasks with no shared mutable state or output dependencies, such as parallel research or data gathering. Sequential execution remains preferable for tasks with strict output dependencies or shared state requirements.

Executing multi-tasking and dispatching parallel agents requires available agents (explore or general), a compliant task definition for each independent sub-task, and a coordinated collection and synthesis process for the results after concurrent execution completes.

Synthesize results from parallel research tasks by dispatching independent data gathering operations concurrently via the delegate tool, then performing coordinated collection and synthesis to merge outputs into a unified result after all parallel agents complete their assigned tasks.

Parallel task delegation does not work for tasks with shared mutable state or output dependencies. It is designed exclusively for independent, self-contained sub-tasks like parallel research or data gathering where concurrent execution occurs without state interference.