#Rate Limits

Acme applies rate limits at the workspace level, not per-key. Every response carries the live limit state in headers:

X-Acme-Limit: 1000
X-Acme-Remaining: 947
X-Acme-Reset: 1736812800

#Default quotas

#Burst behaviour

Acme runs a token-bucket limiter sized at burst tokens, refilling at sustained per second. A burst of burst requests is allowed, then traffic must slow to the sustained rate or be rejected with 429 Too Many Requests.

#Handling 429

The SDK retries 429 responses with exponential backoff and respects the Retry-After header:

import { client } from '@acme/sdk'

const c = client({
  retry: {
    on: [429, 502, 503, 504],
    max_attempts: 5,
    base_delay_ms: 200,
  },
})

Hand-rolled clients should mirror that behaviour.

#Raising your limit

Pro and Enterprise customers can request a higher limit through the dashboard. Limits are increased without downtime — usually within two business days.