# Rate limits and sizes

Every key has its own limits:

| Limit | Default |
|---|---|
| Requests per minute | 300 |
| Requests per hour | 10000 |
| Write requests (POST, PUT, PATCH, DELETE) per minute | 90 |
| File upload size | 25 MB |
| File read size | 5 MB |
| Active keys per account | 50 |

A bulk upload counts as ONE write request however many files are in it, so the archive upload
(POST on `/sites/<id>/files/archive`) is the way to put a whole built site up - not one file write per file.

Every answer carries `X-RateLimit-Limit`, `X-RateLimit-Remaining` and `X-RateLimit-Reset` (Unix time the minute window
ends). Over the limit you get `429 rate_limited` with `Retry-After` (seconds): wait that long and retry.
`GET /limits` returns the limits of the key making the call, and says whether they are the standard ones:
`overridden` is `true` (with `override_scope` set to `key` or `account`) when support has adjusted them for you.

## Need a higher limit?

These limits suit almost every dashboard, script and AI assistant. If your tool genuinely needs more - a large
account, a migration, a nightly sync over thousands of sites - [ask support](https://app.pbn.ltd/support/), say what
you are building and roughly how many calls a minute you need. We can raise the limit on a single key or on the whole
account, and it applies to your next call - nothing to change in your code.

Some actions also have the panel's own limits, whatever the key: measuring usage (one per site every few minutes),
error logs (20 fetches per 10 minutes per account), access logs (30 per 10 minutes), health re-checks (10 an hour),
integrity checks and repairs (12 per 10 minutes per site), admin login links (a few per minute).

## Good practice

- Cache what does not change often; use `ETag` / `If-None-Match` on GET requests: an unchanged answer is
  `304 Not Modified` with no body.
- Poll jobs every 5-10 seconds, not faster.
- On `503 unavailable` (a server did not answer) retry after a short wait.