600 requests/minute per API key across the whole surface. Over the line:
HTTP/1.1 429 Too Many RequestsRetry-After: 37{ "name": "rate_limited", "code": "rate_limited", "message": "Rate limit exceeded.", "fix": "Back off and retry." }
Send-heavy workloads should fan out across project-scoped keys; webhook
deliveries to your endpoints retry separately (Effect schedule, 5 attempts,
jittered backoff) and never consume your quota.
⌘I
Assistant
Responses are generated using AI and may contain mistakes.