API Platform: Arrow 2, Open Responses, and sandbox keys
By Steven Van ·
Arrow 2 models are billed by measured token usage, and new sandbox keys let developers test against the API without reaching a model.
QuiverAI's API platform now has a POST /v1/responses endpoint that runs on Arrow's Open Responses contract, alongside token-based pricing for the new Arrow 2 models and sandbox keys for testing. The changes are listed in the original announcement for QuiverAI.
Open Responses endpoint
Developers can declare function and custom tools, execute function_call items themselves, and point a stock Open Responses client at the endpoint unmodified. The endpoint is stateless. previous_response_id and store: true are rejected rather than silently ignored.
Token-based pricing for Arrow 2
Arrow 2 and Arrow 2 Telos are priced by measured token usage rather than fixed credits. The rates are reported as billing.rates on GET /v1/models. Arrow 1, Arrow 1.1 and Arrow 1.1 Max remain available through the API.
Sandbox keys
Sandbox keys reach the same routes as production keys but never reach a model. They return deterministic output marked x-quiver-environment: test. A documented inline marker slows a streamed response down on demand, which lets developers exercise client-side timeout handling.
Rate-limit headers and key rotation
- Every response carries rate-limit headers for limit, remaining, reset, scope, subject and dimension, so a client can back off before it is throttled.
- API key issuance and rotation are now crash-recoverable. A failure mid-rotation can no longer strand an organization or leave a phantom live secret.
