FAQ
Straight answers on billing, limits and how the gateway behaves.
Is the API really OpenAI-compatible?
Yes. Point any OpenAI client at https://api.jackz.ai/v1 with your key and it works. Chat completions, embeddings, streaming, tool calling and structured outputs follow the standard shapes.
How is pricing set?
List prices are aligned 1:1 with OpenRouter public list prices. You are billed per token against your prepaid balance — the models page shows the exact numbers.
What happens when my balance hits zero?
Requests return 402 until you top up. Nothing is billed retroactively.
Are failed requests billed?
No. Upstream errors and timeouts are retried automatically across providers, and failed generations are never billed.
Do you have free models?
Yes — models listed at $0.00 on the models page are free at modest rate limits.
What are rate limits?
Defaults suit development and scale with sustained usage. Contact us before large planned bursts.
Do you train on my prompts?
No. Prompts pass through to the serving provider; we do not use your content for training. Provider-side retention varies — see the privacy guides for routing controls.
What is a Compute Contract?
A prepaid capacity product: your prepayment releases as token allowance in a daily drip over a 5-year term. It is a service, not an investment product — see the contracts page for the math.
Where are you based?
Tokyo, Japan. The gateway terminates in Japan with anycast routing across Asia and beyond.
How do I get my app on the rankings page?
Set the attribution headers documented in the app attribution guide; ranked by real token volume.
Something else? Contact support.