Rate limits
Respect organization-wide REST API limits and retry failed requests safely.
Production REST API limits are organization-wide sliding windows of 60 seconds. Paid plans currently offer 120 requests per minute for Hobby, 300 for Standard, 600 for Pro, and 1,000 for Enterprise. Treat the limits as a constraint of your integration. Do not spread your calls across workers that do not coordinate with each other.
A sandbox workspace gets 30 requests per minute. It is the one unpaid plan that reaches the REST API. See REST API: agent sign-up for the rest of that plan's limits.
A free organization has no rate limit of its own, because it never reaches the limiter. The plan gate runs first and answers every endpoint with 403 PLAN_UPGRADE_REQUIRED. Read a 403 on a free organization as missing API access, not as too much traffic.
Successful responses include X-RateLimit-Limit, X-RateLimit-Remaining, and X-RateLimit-Reset. When the API limits your request, it returns 429 with RATE_LIMIT_EXCEEDED and a Retry-After value in seconds. Centralize backoff in the service that owns your API key. Honor the Retry-After value. Add jitter so that concurrent jobs do not resume at the same time.
Reduce unnecessary traffic before you increase concurrency. Cache stable reads. Page through lists instead of restarting them. Poll asynchronous work at a fixed interval. Batch your follow-up processing. Local development does not always enforce production rate limits, because it can lack the backing services. Test the recovery path and the happy path.