The English text is legally binding. Other languages are convenience translations.

ExecuteAI ApS · CVR 45798798

Terms of Service

Effective 14 September 2026. These terms govern use of tokenmaxing.online and api.tokenmaxing.online.

1. Who we are

Tokenmaxing.online is operated by ExecuteAI ApS, Nikolajgade 22, 1068 København K, Denmark, CVR 45798798. Contact: info@tokenmaxing.online.

The service is offered to businesses. If you are a consumer, mandatory consumer-protection rules in your country still apply and take precedence where they cannot be waived.

2. What the service is

Tokenmaxing is a prepaid, OpenAI-compatible HTTP gateway. You send chat-completion requests with an API key. We forward a supported request to the live backend for that alias (Azure OpenAI, Azure AI Foundry, or Google Gemini), return the provider’s response, and debit your prepaid Stripe balance using the provider’s measured token usage.

The public calculator and the public price feed are illustrative cost scenarios for the models that are live today. They are not a savings guarantee. The authoritative list of models your key can call is GET /v1/models.

3. What is live today

Inference is currently available for these aliases (dots and hyphens are treated as the same name). The authoritative list is always GET /v1/models:

The public catalogue lists this current generation. A few older OpenAI aliases remain deployed and may still appear on GET /v1/models. Muse is not live. Anthropic /v1/messages is not implemented; Claude is called through the OpenAI-compatible /v1/chat/completions gateway.

4. Accounts and API keys

Onboarding collects a company name and work email. Email ownership must be verified before a session and API key are issued. The API key is emailed to you; only its hash is stored in our database. The account starts with zero prepaid balance. Sign in with an email code to add credit or replace a lost API key. Replacement immediately revokes the previous key.

You are responsible for keeping the key secret and for all use under it. Tell us at info@tokenmaxing.online if it is lost or leaked so we can revoke it.

You may also store your own Anthropic, OpenAI or Google API keys in Account. Those keys are encrypted at rest and used only to call that provider for your requests. The portal never shows a stored key again. Removing a key stops further use immediately. Claude is not served without your Anthropic key. If you add an OpenAI or Google key, those families use your key instead of the platform credential for that call. The provider bills the account that owns the key; our prepaid balance still covers the Tokenmaxing fee and any platform-served models.

5. Prepaid billing

Usage requires a positive prepaid balance. You add funds through Stripe Checkout. Balance is credited only after Stripe reports a settled payment, either by a signed webhook to our gateway or by the browser returning to the site so we can verify the session.

Before an inference call we reserve a conservative maximum against your balance. After the provider responds we settle to the measured input, output and cache price plus a performance fee of 30 percent of any documented saving versus the requested model’s rate card. If the provider call fails, or measured usage is missing, the reservation is released and nothing is kept for that call.

If the reserved amount exceeds your balance the call is rejected with HTTP 402 and no upstream request is made.

6. No guaranteed savings, cache or uptime

We do not promise a savings percentage, a cache hit rate, a latency figure or an availability SLA. Cache figures are taken from the provider response when the provider reports them; otherwise they are zero. Older site copy that stated fixed savings, cache or uptime numbers was marketing text, not a measurement, and is not part of this contract.

7. Your content

Prompts and model outputs are sent to the backend that serves the requested alias so the model can answer. Most OpenAI aliases and DeepSeek V4-Flash stay in the EU data zone. gpt-6.astra uses Azure OpenAI GlobalStandard and may leave the EU. Grok, Gemini, Llama, Kimi and DeepSeek V4 Pro are served globally and may leave the EU. We do not store prompt or completion text in our application database. You warrant that you have the right to submit the content and that it does not violate law or these terms.

8. Acceptable use

You may not use the service for unlawful content, to attack or overload the gateway or any provider, to probe other tenants, to resell access without a written agreement, or to circumvent prepaid billing. We may suspend a key immediately if we reasonably believe this section is breached.

9. Changes

We may change models, rate cards, the performance fee and these terms. The current versions are published on this site. Material changes will be announced on the site or by email to the address on the account. Continued paid use after that notice is acceptance of the new terms.

10. Liability

The service and model outputs are provided as is. We are not responsible for decisions you make from model output. Except for death or personal injury caused by negligence, fraud, or liability that Danish law does not allow us to limit, our aggregate liability in any twelve-month period is limited to the unused prepaid balance on the account plus the amounts you paid us in that period.

11. Termination

Either party may close the account by written notice to the other. Unused prepaid balance is handled under the Refund Policy. We may also close an account that has had a zero balance and no calls for twelve months.

12. Governing law

These terms are governed by Danish law. The courts of Copenhagen have jurisdiction, unless mandatory law gives you a right to sue elsewhere.