Rate limits

300 a minute per key, a burst of 50, and 1,200 a minute per tenant.

Limit	Value
Per key	300 requests a minute
Burst, per key	At most 50 in any 10 seconds
Per tenant	1,200 requests a minute, across all of its keys

The burst limit stops a key spending its whole minute in one second. The tenant limit is shared by every key the tenant holds, so a batch job and an interactive integration each get a key, and neither can starve the other of its own 300, but together they stay within 1,200.

Every authenticated response carries:

• RateLimit-Limit: the key’s requests a minute.
• RateLimit-Remaining: what is left in the tightest window.
• RateLimit-Reset: seconds until that window resets.

Over the limit, the answer is 429 with the problem type rate-limited and Retry-After, in seconds. A refused request still counts, so wait for Retry-After instead of retrying at once: a client that keeps hammering stays limited.

import time, requests

def get(url, **kw):
    while True:
        r = requests.get(url, **kw)
        if r.status_code != 429:
            return r
        time.sleep(int(r.headers.get("Retry-After", "1")))

Reading a large directory is cheap at the right page size: ask for limit=500 and a first full read of 10,000 people is 20 requests. See Pagination.