Guides
Rate limits
Build an integration that plays well at any pace.
| Bucket | Requests per minute |
|---|---|
| Per API key | 60 |
| Per user across all keys | 120 |
| AI, Jev, voice refresh/customization per user | 10 |
These are atomic shared fixed-minute windows, not rolling windows. Requests near a window boundary may burst. Headers X-RateLimit-Limit/Remaining/Reset/Policy describe the last checked bucket (or the bucket that blocked the request); Reset uses Unix seconds. Failed requests can consume quota. Additional edge-IP, business, provider and concurrency limits still apply and may produce 429 even with remaining public quota. Respect Retry-After with jitter; never rotate keys to retry around a limit.