APIs, websites and AI providers reject requests when a workflow sends too many too fast. You usually see it as an error such as 429 Too Many Requests, “rate limited” or “quota exceeded”, often only after the first few items worked.
Use Split in Batches
Section titled “Use Split in Batches”The fix is almost always Split in Batches: it sends the items through the next steps a few at a time instead of all at once.
- Put Split in Batches right before the node that calls the service (for example Slack, Google Sheets or HTTP Request).
- Set Batch size below the service’s limit, for example 10. The repeated steps start from its Loop output.
- Add a Wait inside the batch loop to space the batches out, for example 1 second.
- Connect the last repeated step to the Loop Back input, and continue the workflow from the Done output.
Send fewer requests
Section titled “Send fewer requests”- Filter and Limit cut down how many items reach the service at all.
- Aggregate combines items into one, so you send one message or row instead of many.
- Run scheduled workflows less often: a Schedule every hour instead of every minute.
If it still fails
Section titled “If it still fails”- Check the account’s plan and quota with the provider. Free tiers often have much lower limits.
- For HTTP Request, set a timeout and decide what happens on an error, so one rejected call doesn’t stop the run.
- AI providers limit tokens as well as requests: shorter prompts and smaller inputs help, or use an on-device model, which has no rate limit.