Terms of Service

These Terms govern your use of the Rhodes inference API, the customer portal, and the documentation (together, the "Service").

The Service is an OpenAI-compatible HTTP API for language-model inference, sold as prepaid credit. You call our endpoint with a key we issue; we forward the request to a model-hosting provider and meter what you consume against your prepaid balance.

By creating an account or calling the API you accept these Terms. If you are accepting on behalf of a company, you confirm you are authorized to bind it.

What the Service is not: There is no subscription, no invoice, no committed volume, and no service-level agreement.

An account is created the first time an authenticated identity reaches the control plane. One account has one credit balance and many API keys. There are no teams, no sub-accounts, and no role-based access control.

You must provide an accurate email address. Sign-in is handled by a third-party identity provider. Linking a new login to an existing account requires a verified email address.

You are responsible for everything done through your account and your keys, including by your own end users.

Keys are created and revoked from the portal. A key's secret is shown exactly once, at creation, and is never shown again. We do not store it in cleartext and it is never written to a log. If you lose a secret, create a new key and revoke the old one.

A key is a bearer credential. Anyone holding it can spend your credit. Keep keys server-side. All of your keys draw on the same single balance.

Do not share, resell, or sublicense keys.

The Service is prepaid. You buy credit denominated in US dollars before you use the Service. There is no subscription and nothing auto-renews. Credit is purchased one top-up at a time. Automatic recharge is deliberately not implemented.

A top-up is a minimum of $5 and a maximum of $5,000 per transaction. Payments are processed by Stripe. We do not see or store your card details.

Credit is granted only after the payment processor confirms the payment, never optimistically. Consumption is metered per request by the gateway, in tokens, at the prices published in the model catalog.

At a zero balance the API refuses requests. This is a hard cap, enforced before a request is dispatched upstream. Your keys stay valid; requests resume when credit lands.

Credits do not currently expire. Prices exclude tax unless stated. Currency is USD only.

We may grant a small trial credit at signup. Trial credit is capped: one grant per account, ever, and a ceiling on the number of grants per signup IP address.

Trial credit is a promotional grant. It has no cash value, is not refundable, and we may change, withhold, or withdraw it at any time. Attempting to obtain more than one grant — including by creating multiple accounts — breaches the Acceptable Use Policy.

There is no SLA, no uptime commitment, and no service credit. We do not promise availability, latency, throughput, or that any particular model remains available.

We may change the model catalog, prices, rate limits, and the providers serving a model. Default rate limits are 60 requests/minute and 100,000 tokens/minute per key.

We resell inference. Your request is forwarded to a third-party model-hosting provider which generates the output. Customer-facing model names are deliberately brand-neutral. We do not disclose which provider serves a given request.

What we retain about a request is metadata, not content: timestamp, model name, token counts, cost, and customer identifier. Prompt and response content is transmitted to the provider to be served. See our Privacy Policy for the precise position.

Output is generated by a machine learning model. It may be inaccurate, incomplete, biased, offensive, or entirely fabricated. It may be substantially similar to output generated for someone else. Do not treat it as fact, as advice, or as a substitute for a professional.

You are responsible for how you use output — for reviewing it, for any decision you base on it, and for any onward publication. Do not use the Service where an incorrect output causes harm.

As between us and you, we do not claim ownership of your inputs or of output generated for you, subject to the terms of our upstream providers.

The Service is provided "as is" and "as available", without warranties of any kind, express or implied, including any implied warranty of merchantability, fitness for a particular purpose, non-infringement, accuracy, or uninterrupted operation.

We specifically do not warrant that output is accurate, that the Service will be available, or that it is fit for any regulated or high-risk use.

We hold no security or privacy certification. We are not SOC 2 audited, not ISO 27001 certified, and not HIPAA compliant. Do not send protected health information or other regulated data.

To the maximum extent permitted by law, our total liability to you for all claims arising from or relating to the Service will not exceed the amount you paid us in the twelve months before the claim arose.

We are not liable for any indirect, incidental, consequential, or special damages, or for any loss of profits, revenue, data, or business opportunities, even if we have been advised of the possibility of such damages.

We may suspend or terminate your account for breach of these Terms or of the Acceptable Use Policy, for non-payment, for suspected fraud or abuse, or where required by law.

Suspension is immediate and total: Your key is blocked at the gateway first, so inference requests stop immediately. There is no grace period in the mechanism.

You may stop using the Service at any time. There is no notice period and nothing to cancel, because there is no subscription.

These Terms are governed by the laws of the State of Delaware, United States, without regard to conflict of law principles.

These Terms incorporate the Acceptable Use Policy and the Privacy Policy.

We may update these Terms from time to time. Continued use of the Service after changes constitutes acceptance of the updated Terms.