Multi-agent

Parallel sectioning

Split independent parts, run them at once, then merge.

N agentsCore

How it works

1Divide the task into independent sections with no cross-dependencies.
2Run each section concurrently as its own call.
3Each works in isolation with only its slice of context.
4Combine the outputs with a final merge/format step.

Use it when

A task cleanly divides into pieces with no dependencies — sections of a report, files in a batch — and you want speed or separate focus per part.

Reach for something else when

Parts depend on each other. The merge becomes the hard part and can reintroduce the work you saved.

Where you stay in the loop

You define the section boundaries; the agents work in isolation. The merge is yours to scrutinize — independent parts can quietly contradict each other.

In the wild

Summarize a long report by sending each chapter to a separate call simultaneously, then stitch the summaries together.

Hand this to your agent

Use parallel sectioning.

Split my task into independent sections that can each be done without knowing
the others' results. List the sections first.

Then produce each section's output separately and clearly labeled. Finally,
merge them into one deliverable — smooth the transitions and remove
redundancy.

Task: <...>

Replace the <…> placeholders, paste it into your agent, and it'll scaffold the workflow with you.