Rate Limits

When you get a 429

Every rate-limit refusal, on every endpoint, has the same shape: HTTP 429 with error_code "RATE_LIMITED", retry_after (seconds) in the body, and a Retry-After header.

The Retry-After header and the body's retry_after are the number of seconds until the window of the limit you hit resets. When more than one limit fired, it is the longest of their resets, so a retry at that moment is not refused by any of them. Wait that long; do not retry sooner. (Before 2026-10-09 an hourly limit always answered 300 seconds, however much of the hour was left.)

HTTP/1.1 429 Too Many Requests
Retry-After: 1742

{
  "error": "rate_limit_exceeded",
  "error_code": "RATE_LIMITED",
  "message": "Request rate limit exceeded. Please slow down and try again.",
  "retry_after": 1742,
  "limit_type": "rate_limit"
}

Match on error_code, not on the message. Some send-safety checks also answer 429 with their own error value (for example an approval or claim that is temporarily locked); they also carry Retry-After. Honour the header whatever the body says.

What your code should do on any 429: wait the number of seconds in Retry-After (or the body's retry_after if the header is not available to you), then retry the same request. Do not hard-code the limits on this page into your client: they can change, and the 429 always tells you how long to wait.