# Rate limits

`POST /v1/invoices` currently has a configured limit of 100 requests per minute. The limiter runs before API-key authentication, so its current bucket is the request IP, not a guaranteed independent allowance per API key. Deployments may also apply limits outside the application.

Do not assume a blanket 5 requests/second login limit or 20 requests/second global limit. Routes opt into application limits individually.

A 429 response uses error code `rate_limited` and includes `error.details.retry_after_seconds`. Wait at least that interval, add jitter and retry sequentially. Preserve the original idempotency key and body for invoice creation.

---