Workflows and Parallel Agents
Goal — Use a workflow to run several pieces of work at once and get one consolidated result.
By the end — One merged summary produced by several agents working in parallel.
1Understand sub-agents
Claude Code can launch helper agents — sub-agents — that work on their own and report back. The main session stays your point of contact.
Explain what a sub-agent is and why the main session uses them.2Understand a workflow
A workflow launches many sub-agents at the same time, each on a different piece, then merges their results. Instead of researching five things one after another, you research all five at once.
Explain how a workflow runs many agents in parallel and consolidates their output.3Run a parallel research task
Give it something genuinely multi-part. Several agents work at once; you get a single merged answer far faster than doing each in turn.
Research five ways to improve this site — accessibility, performance, layout, color, and copy — in parallel, then give me one consolidated summary.4Read the consolidated result
Notice you received one organized summary, not five scattered ones. The workflow did the gathering and the merging.
5Know when to use it
Workflows shine when work splits into independent parts that can run at once. For a single, linear task, a normal prompt is better.
Given how workflows work, what kinds of tasks are they best and worst for?Recap — You ran one ask across several agents at once and got a single merged result — the workflow gathered the pieces and consolidated them for you.