Rate limits
Quotas per tier, headers, and backoff strategy.
Limits are enforced per workspace on two axes: requests per minute (RPM) and tokens per minute (TPM). Whichever you hit first wins.
Limits by plan
| Plan | RPM | TPM | Concurrent runs |
|---|---|---|---|
| Starter | 60 | 40,000 | 3 |
| Pro | 600 | 500,000 | 25 |
| Enterprise | Custom | Custom | Unlimited |
Response headers
HTTP
x-nexus-limit-requests: 600 x-nexus-remaining-requests: 587 x-nexus-reset-requests: 4.2s x-nexus-limit-tokens: 500000 x-nexus-remaining-tokens: 471233
Handling 429s
The SDK retries 429 and 5xx responses automatically with exponential backoff and full jitter, up to three attempts. If you implement your own client, respect the reset header instead of retrying immediately.
Bursts are fine
Limits are token-bucket based with a 2x burst allowance, so a short spike will not be throttled as long as the one-minute average stays under the ceiling.