Documentation
Authentication & scopes
Rental API requests use the same prlc_ key as inference, but the key must be granted the rental scope before it can spend credits on GPUs.
The Rental API authenticates with the same prlc_ API key as the Inference API, passed as a bearer token:
Authorization: Bearer prlc_your_key_here
The rental scope
A key can call inference out of the box, but it cannot start rentals until you grant it the rental scope. This is deliberate: rentals spend real credits, so a key that leaks into a notebook or CI job shouldn't be able to drain your balance.
Every key you already have is inference-only by default. To enable rentals:
- Open the Console and go to API keys.
- Click the settings icon on the key you want, and turn on GPU Rentals.
- Optionally set Max rentals running at once (see below).
You can create a fresh key with the rental scope from the start, or toggle it on an existing key at any time. Turning it off immediately stops that key from starting new rentals.
Requests without the scope get:
{ "error": "insufficient_scope",
"message": "This API key does not have the 'rental' scope. Enable it for this key first." }
Per-key rental limit
Each key has a max_active_rentals limit (default 2, settable 0–10 in the Console). It caps how many rentals that key can hold at the same time — a safety net against a runaway script. Hitting it returns 409 rental_limit_reached.
Setting the limit to 0 pauses rentals on a key without revoking it: existing rentals keep running, but the key can't start new ones.
Your credit balance is the real ceiling on spend;
max_active_rentalsjust bounds concurrency. Keep keys private, use a separate key per app, and delete a key the moment you suspect it leaked.