Rate limits

Each API key may make 200 requests per 10 seconds. The limit is the same for every plan and every operation under /v1.

How it works#

• Each key has a bucket that holds up to 200 requests. A request takes one. The bucket refills continuously at 20 requests per second.
• A burst of up to 200 requests succeeds at once. After that you can sustain 20 requests per second.
• The limit is per key, not per account or per IP address.
• The public endpoints /health, /plans, /openapi.json and /llms.txt are not rate limited.
• Requests that fail authentication are not counted.
• Each API server process keeps its own buckets in memory. If more than one process serves the API, the limit applies to each process separately.

Response headers#

Every authenticated response carries these headers.

When you are limited#

The API answers HTTP 429 with the rate_limited code and retryable: true. Wait for Retry-After seconds and retry. See Errors.

{
  "error": {
    "code": "rate_limited",
    "message": "Too many requests for this API key.",
    "fix": "Wait the number of seconds in the Retry-After header, then retry. Spread requests out to stay under the X-RateLimit-Limit value.",
    "docs_url": "/docs/errors#rate_limited",
    "retryable": true,
    "request_id": "req_Qx7RtVw3ZbN1pEs9fKd2Lm"
  }
}

async function withRetry(send, attempts = 4) {
  for (let i = 0; i < attempts; i += 1) {
    const res = await send();
    if (res.status !== 429) return res;
    const wait = Number(res.headers.get('retry-after') ?? 1);
    await new Promise((resolve) => setTimeout(resolve, wait * 1000));
  }
  throw new Error('still rate limited');
}

Other limits#