Keys & limits

Each key is an independent set of permissions and limits, not just a password. Use one key per application so you can revoke or cap them separately.

Spend limit

Caps total spend for one key. Once reached, that key returns 402 while your other keys keep working. This is the single most useful guard against a runaway loop draining your balance.

Model allowlist

Restrict a key to specific model IDs. A request for anything else returns 403. Useful for pinning a production app to a model you have tested.

IP allowlist

Accepts exact addresses and CIDR ranges. A leaked key is useless from outside your ranges. Leave empty to allow any IP.

203.0.113.42
198.51.100.0/24
2001:db8::/32

Rate limit

Requests per minute, measured with a sliding window. Set 0 to use your tier default.

Concurrency

In-flight requests are capped per account. This matters more than RPM for streaming: one stream can occupy a slot for minutes.

Expiry

Optional. Handy for keys you hand to a contractor or use in a demo.

Rotating a key

Plaintext is shown once at creation and never again — we store only a hash, so a database leak does not expose usable keys. To rotate, create the new key, deploy it, then delete the old one.