Automation & triggers

Batch processing

Run the same workflow across many items efficiently.

N agentsStarter

How it works

1Collect the items into a queue.
2Apply the same workflow to each, in parallel batches within rate limits.
3Capture per-item results and errors separately.
4Retry failures, then aggregate and report at the end.

Use it when

Large volumes of similar items — score 10k reviews, tag a dataset, clear a backlog — where throughput and cost matter more than latency.

Reach for something else when

Items each need different handling (route first), or you need real-time results.

Where you stay in the loop

You define the per-item workflow and the failure policy; the agent grinds through the volume. Review the errors list and a sample of the successes — scale hides individual mistakes.

In the wild

Classify a year of support tickets by theme overnight, with a failure list for the few that error.

Hand this to your agent

Design a batch job.

For processing many items the same way, define: (1) the per-item workflow,
(2) the output schema for each item, (3) how to handle and retry failures
without stopping the batch, (4) the final aggregate summary.

Then process the items I paste (one per line) and return a table plus an
errors list.

Item type + task: <...>

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