Frequently asked

Each Relay model has its own price, and that price is what each request is debited. 93 of 100 priced models cost less than the SearchAtlas AI Gateway’s own list price, by up to 60%; the model table shows both prices for every model. We are not a marketplace of other people’s cloud credentials — requests use the SearchAtlas AI Gateway’s own routed capacity.

No. We do not store prompt or response bodies. What we retain is request metadata only — model, token counts, latency, and the amount charged — for billing and abuse prevention. See the Privacy Policy for the full data schedule.

New requests are rejected with a 402 before they reach any model provider — you are never charged for a request that could not be served. Top up any time in the console and the next request goes through immediately.

At launch, the first deposit of $5 or more on a new account is planned to receive a flat $10 bonus — fund $5, get $15 to spend. It will apply once per account, on the first paid deposit only; every deposit after that will be credited at face value. Paid launch is pending and card payments are not live yet, so the bonus is not available yet.

Every account starts at 8 concurrent requests shared across all its API keys. If you’re running a fleet of agents or a production workload that needs more, contact us — raising the account cap is a quick change on our side.

Not today. We run on the same gateway infrastructure that serves SearchAtlas production traffic, but we don’t make a contractual uptime promise in v1. If a request fails before reaching a provider, you are not charged.

Yes — set the base URL and API key as environment variables and the agent routes through Relay with no other config changes. Exact snippets are on the Docs page.

No. Your balance carries forward indefinitely as long as the account stays active — there’s no monthly reset and no use-it-or-lose-it clock.