Your First Parallel Task

Parallel execution is one of the primary reasons to use agent teams. When two tasks have no dependency on each other, running them sequentially wastes time. A researcher and a data analyst working the same problem from different angles can both start immediately and return results at roughly the same time. The parent agent dispatches both, waits for results, and synthesizes.

Let us build a concrete example. You want a briefing on a new competitor. Agent A (Researcher) searches the web for the competitor's website, recent news, and product announcements. Agent B (Analyst) pulls any available financial data, team size, and funding information. Both run simultaneously. When they return, the parent combines the research findings with the analytical data into a single competitor profile. What would take one agent two sequential passes takes the team one parallel pass.

1. 1. Identify Independent Subtasks Map out the full task and draw the dependency graph. Any subtasks without dependencies on each other can run in parallel. In our example, web research and data analysis are independent.
2. 2. Dispatch Simultaneously The parent agent spawns both child agents in the same message. Each child receives its own prompt with a clear deliverable. They begin executing at the same time.
3. 3. Collect and Synthesize When both agents return results, the parent reviews both outputs, resolves any contradictions, fills gaps where one agent found something the other missed, and assembles the combined deliverable.