Rate limits

Limits

• 120 requests per minute per partner, shared by all of your keys, live and test together.
• Each key is also capped at 120 requests per minute on its own.
• The window is a fixed calendar minute (UTC), not a rolling 60 seconds: the count resets at the start of each minute.

What counts

• Every request with a valid key counts, including requests answered 403 or 429. The scope check runs after the rate limiter, so calls a key is not allowed to make still spend budget.
• A request refused with 401 (missing, unknown or revoked key) is not counted.

When you are limited

Over the limit you get 429 with the body { "error": "Rate limit exceeded." } and a Retry-After header of 60 seconds, whatever is left of the minute. Back off before retrying, and add random jitter so your workers do not all retry at the same moment.