Rate limits

Limits protect the platform and keep latency predictable. Design clients to read the limit headers and back off.

Response headers

Rate-limited responses carry your current budget so you can throttle proactively:

http

RateLimit-Limit: 60
RateLimit-Remaining: 54
RateLimit-Reset: 1792710460

RateLimit-Reset is the Unix timestamp when the limiting window resets. Effective limits depend on your program and the endpoint's policy. Only anonymous public metadata routes such as health and OpenAPI are exempt; authenticated identity, usage, log, provider, and resource routes remain rate limited.

When you exceed a limit

You receive 429 with the same headers. Back off and retry after the reset, adding jitter so retrying clients don't synchronise:

python

import time, random, requests

def call_with_backoff(fn, attempts=5):
    for i in range(attempts):
        r = fn()
        if r.status_code != 429:
            return r
        reset = int(r.headers.get("RateLimit-Reset", "0"))
        wait = max(0, reset - int(time.time())) or (2 ** i)
        time.sleep(wait + random.uniform(0, 1))
    return r

Cache reads that change slowly and prefer webhooks over polling once webhook delivery is enabled for your organization.