No seats, no subscription, no minimum commitment. Every model is priced 5% below the market reference rate, and volume earns a further discount on top of that.
For building, evaluating and everything up to production scale.
A further automatic discount, stacked on the standard 5%, once monthly spend passes $2,000.
For regulated teams and committed-use agreements.
Daily and total ceilings per key, per model.
SSE streaming and tool calling on every supported model.
Per-request fallback chains at no extra charge.
Token counts and cost returned on every response.
EU-compliant invoices with reverse charge where applicable.
Your prompts are never used to train any model.
Disable request logging entirely, per key.
Multiple members on one balance, with roles.
Billing follows the token counts reported by the upstream provider for each request — input tokens, output tokens and, where the provider distinguishes them, cached or reasoning tokens. Those counts are returned to you on the response so you can reconcile independently. Balance is debited when the request completes; failed requests are not charged.
No. Prepaid credit stays on the account until you spend it. If you close your account with an unspent balance from a card or bank top-up, you can request a refund of the remaining amount minus any processing costs already incurred.
Prices are quoted in USD because upstream providers price in USD. You can settle in EUR or USD; for EUR settlement we apply the ECB reference rate on the day of the invoice and show it on the document.
Our published rate already sits 5% under the market reference, so for most teams the answer is no. At very large single-model volume a direct committed-use contract with one provider can still beat any aggregator, and we will tell you when that is the case. EURU wins when you use several models, want to switch between them without re-procuring, or value one EU invoice over ten vendor relationships.
Routing overhead is typically under 30 ms and streaming is passed through without buffering, so time-to-first-token is essentially the upstream's. The gateway runs in EU regions with connections pooled to each provider.
If you supplied a fallback chain, we transparently retry the next model in your list. If you did not, we return the upstream error unchanged so your own retry logic stays in control. Either way you are billed only for a request that produced output.
Yes, on Volume and Enterprise plans. You keep your existing contract and rate with a provider, and route through EURU purely for unified access, controls and analytics. That traffic is charged a flat platform fee instead of per-token pricing.
Send us your monthly token estimate and model mix — you will get a written quote back, not a discovery call.
