> ## Documentation Index
> Fetch the complete documentation index at: https://developers.getbukki.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Rate limits

> Respect key-level and workspace-level request limits.

During the private beta, Bukki allows:

* 60 requests per minute per API key
* 300 requests per minute per workspace

Successful responses include rate-limit metadata. A request that exceeds either allowance receives `429 RATE_LIMITED` and a `Retry-After` header.

## Retry safely

Use bounded exponential backoff with jitter. Do not immediately retry a `429` in a tight loop, and do not create additional API keys to bypass workspace limits.

If the distributed rate-limit service is unavailable, Bukki fails closed with `503 SERVICE_UNAVAILABLE` instead of accepting unmetered traffic. Retry `503` responses with backoff and retain the response's request ID.