Rate Limits

The LLM gateway (/v1/* on the Allternit API, relayed by the Cloud API) enforces two sliding-window limits. When one is exceeded it returns 429 with a Retry-After header.

Layer	Scope	Configured by
Per virtual key	Requests per minute for one `ak-` key	The key’s `rate_limit_rpm` (falls back to a built-in default)
Per organization	Requests per minute across all keys of an org	`organizations.gateway_rate_limit_rpm`

The Cloud API also applies its own rate-limit middleware and sets X-RateLimit-Limit, X-RateLimit-Remaining, and X-RateLimit-Reset on the responses it serves.

Endpoint

Method	Path	Description
`GET`	`/v1/rate-limits`	Requests and budget remaining for the calling virtual key

curl https://api.allternit.com/v1/rate-limits \
  -H "Authorization: Bearer $ALLTERNIT_API_KEY"

{
  "object": "rate_limits",
  "requests_remaining": 58,
  "requests_limit": 60,
  "tokens_remaining": 1200,
  "tokens_limit": 5000,
  "reset_at": "2026-09-30T18:00:00Z"
}

tokens_remaining and tokens_limit are the key’s budget, counted in cents (remaining_cents, limit_cents), not tokens. There is no limit on concurrent sessions or batch jobs.