What is an AI gateway?
An AI gateway is the infrastructure layer between your application and the model providers: one endpoint, authentication, metering, rate limiting and routing to the provider you want. The more models you use, the more work that layer saves — and the more relevant the question becomes: who manages that layer, and under which law?
What a gateway does
A gateway normalises API differences between providers (every provider has its own dialect), manages API keys and credits, logs usage per key or team, and can route requests on price, speed or availability — with failover when a provider goes down. Without a gateway you build all those parts yourself, per provider.
The three kinds of gateways
Open-source proxies (like LiteLLM) you host yourself: maximum control, but also your own maintenance, updates and compliance. In-product gateways (like those of cloud platforms) are convenient but tie you to that platform. And hosted aggregators (like OpenRouter) take all the work out of your hands — but are based outside the EU, with the contract and transfer story that comes with it.
Why the jurisdiction of your gateway matters
Your gateway sees all your prompts and answers. It is therefore a party in your data flow, just like the provider itself. For teams with a DPO or heavy customer contracts, a US-based gateway means an extra transfer that must be justified. An EU-based gateway with a data processing agreement and published sub-processor list removes that discussion.
Gatewaybird as a European gateway
Gatewaybird builds the EU version of the hosted gateway: one OpenAI-compatible endpoint, credits and API keys, usage insight — incorporated in the Netherlands, processed in the EU, invoiced in euros. Phase 1 (now) is the price comparison and waitlist; the working gateway follows for the first members.
Frequently asked questions
Is an AI gateway the same as a reverse proxy?
Partly: under the hood it is, but an AI gateway understands the token model, meters per request, manages keys and credits and normalises differences between providers. A bare reverse proxy does none of that.
Can you build your own AI gateway?
Yes — open-source proxies make that accessible. But you run the maintenance burden, the availability and the compliance justification yourself. A hosted gateway buys that off.
What does an AI gateway cost?
Hosted gateways typically charge a markup on the token price of the provider (5-15%) or a subscription. Building yourself is "free" in licence but not in hours and infrastructure.
Why would I want an EU gateway if the models are American?
Because the gateway is a separate party in your data flow. You choose the model deliberately per use case; you choose the gateway on jurisdiction, contract and invoicing. Those two decisions are independent.
Join the waitlist
One email at launch. No spam.
You are on the waitlist
We will email you as soon as the API goes live.