Get compute

Start with available supply

Use GET /v1/compute/offers and current limits. An offer is discovery, not a capacity hold. Inspect execution mode, availability, rate status and storage options. Quote the intended request before creating resources.

Reserve and create

POST /v1/compute/reservations creates a time-bounded hold. Pass its returned ID to POST /v1/compute/instances. Both require separate idempotency keys. Reuse a key only for the same operation and body.

{"offerId":"<returned-offer-uuid>","units":1,"requestedEndAt":"<future-UTC-time>"}

{"reservationId":"<returned-reservation-uuid>"}

Observe readiness and cleanup

A 202 response accepts intent. Poll the instance and events; it does not mean compute is ready. Use allocation and infrastructure operation state to investigate progress. Request release with DELETE /v1/compute/instances/{id} and verify cleanup. Capacity release and provider billing termination are separate observations.

Compute API reference → · Legacy simulated compatibility routes →