Rate Limits

Limits are applied per API key, per minute, and per day. Response headers make consumption observable.

Limits by Tier

Tier	Requests / min	Requests / day	Concurrent
Orbit & Fabric	100	10,000	5
Core	500	50,000	20
Atlas	Custom	Custom	Custom

Response Headers

Header	Meaning
`X-RateLimit-Limit`	Total requests allowed in the current window.
`X-RateLimit-Remaining`	Requests remaining before the limit resets.
`X-RateLimit-Reset`	Unix timestamp when the window resets.
`Retry-After`	Seconds to wait after a 429 response before retrying.

Handling 429 with Exponential Backoff

import time, random, requests

def call_with_backoff(url, headers, max_retries=5):
    for attempt in range(max_retries):
        r = requests.get(url, headers=headers)
        if r.status_code != 429:
            return r
        retry_after = int(r.headers.get("Retry-After", 0))
        delay = retry_after or min(60, 2 ** attempt) + random.random()
        time.sleep(delay)
    r.raise_for_status()

Best Practices

Read X-RateLimit-Remaining and slow down before you hit zero.
Always honor Retry-After on 429 responses.
Use batch endpoints for high-volume workloads instead of per-item calls.
Cache stable results (historical data, model metadata) to reduce request volume.
Add jitter to backoff to avoid thundering-herd retries.

Tier

Requests / min

Requests / day

Concurrent

Orbit & Fabric

100

10,000

Core

500

50,000

Atlas

Custom

Header

Meaning

X-RateLimit-Limit

Total requests allowed in the current window.

X-RateLimit-Remaining

Requests remaining before the limit resets.

X-RateLimit-Reset

Unix timestamp when the window resets.

Retry-After

Seconds to wait after a 429 response before retrying.

Handling 429 with Exponential Backoff

import time, random, requests

def call_with_backoff(url, headers, max_retries=5):
    for attempt in range(max_retries):
        r = requests.get(url, headers=headers)
        if r.status_code != 429:
            return r
        retry_after = int(r.headers.get("Retry-After", 0))
        delay = retry_after or min(60, 2 ** attempt) + random.random()
        time.sleep(delay)
    r.raise_for_status()

Best Practices

Read X-RateLimit-Remaining and slow down before you hit zero.
Always honor Retry-After on 429 responses.
Use batch endpoints for high-volume workloads instead of per-item calls.
Cache stable results (historical data, model metadata) to reduce request volume.
Add jitter to backoff to avoid thundering-herd retries.