Rate limits

How limits work

Limits are enforced per workspace on two axes: the number of concurrent in-flight jobs and the number of jobs per hour. Both scale with your plan tier. Exceeding either returns 429 with a rate_limit_error.

Retrying

On a 429, wait and retry with exponential backoff. Because limits are per workspace (not per key), spreading load across multiple keys does not raise your ceiling.

Raising your limits

Upgrade your plan for higher concurrency and hourly throughput, or contact support for a custom limit.