Requesty
Back|JUN '26BEST PRACTICES / REQUESTY FEATURES
3 MIN READ|

OpenRouter in Europe: Why EU Teams Are Switching to an EU-Hosted AI Gateway

Thibault Jaigu
Thibault Jaigu
CEO & Co-Founder
Published

If you've searched for "OpenRouter EU" or "OpenRouter Europe", you've probably hit the same wall as a lot of European engineering teams: unified LLM APIs are built on US infrastructure, and your data protection officer keeps asking where the prompts go.

This post covers what EU data residency means for an LLM gateway, where US-hosted routers fall short for European teams, and how to run Claude, GPT, Gemini and 400+ other models with the full request path inside the EU.

Why data residency matters for LLM traffic

Every request through an LLM gateway carries your prompts — which in production means customer messages, internal documents, source code, sometimes personal data. Under GDPR, where that data is processed matters:

  • Processing location. If the gateway runs on US infrastructure, your prompts transit and may be processed under US jurisdiction, which puts you in Schrems II / transfer-mechanism territory.
  • Retention. Logs that include prompt content are personal data the moment a user pastes personal data into a chat box. A gateway that retains request bodies becomes a data processor you need to audit.
  • The model endpoint itself. Routing through an EU gateway helps little if the model endpoint is us-east-1. True residency needs EU-region model deployments end to end.

For a fintech, healthcare or public-sector team, these aren't preferences — they decide whether the project clears review.

Where US-hosted routers fall short for EU teams

OpenRouter is a solid product, and for teams without residency requirements it works well. But it routes through US infrastructure and inherits whatever compliance posture each upstream provider has. There's no EU-hosted routing tier, no residency guarantee, and retention behavior depends on the provider behind each model.

That leaves European teams doing one of three things: negotiating DPAs provider by provider, building their own thin proxy in eu-central-1, or finding a gateway that's European by design.

What an EU-first gateway looks like

Requesty's EU gateway was built for exactly this case:

  • Hosted in Frankfurt. The routing layer runs in Germany, so prompts are processed under EU jurisdiction.
  • Zero data retention. Request and response bodies aren't stored. Observability works on metadata — tokens, latency, cost — not content.
  • EU-region models end to end. 77 EU-region deployments are routable today: Claude Opus, Sonnet and Haiku on AWS Bedrock eu-central-1, eu-west-1 and eu-north-1; Gemini 2.5 on europe-west; GPT-5 on Azure France Central. The full request path — gateway and model — stays in Europe.
  • GDPR-ready governance. SSO, role-based access control, audit logs and per-key spend limits, which is what your security review will ask about right after residency.

Over 1,200 European companies route their LLM traffic this way.

Migrating is a two-line change

Both OpenRouter and Requesty expose an OpenAI-compatible API, so moving doesn't touch your application logic:

Python
from openai import OpenAI
 
client = OpenAI(
    base_url="https://router.requesty.ai/v1",  # was openrouter.ai/api/v1
    api_key="YOUR_REQUESTY_KEY",
)
 
response = client.chat.completions.create(
    model="bedrock/claude-opus-4-8@eu-central-1",  # EU-region Claude
    messages=[{"role": "user", "content": "Hello from Frankfurt"}],
)

Model IDs keep the familiar provider/model format. To pin a request to an EU region, pick an EU-region deployment from the model catalog — anything tagged @eu-central-1, @eu-west-1 or europe-west.

Checklist for evaluating any EU LLM gateway

Whatever you choose, these are the questions that decide a European security review:

  1. Where does the routing layer physically run? "EU customers supported" is not the same as "hosted in the EU".
  2. Are request bodies retained, and for how long? Zero retention simplifies your DPA dramatically.
  3. Can you pin models to EU regions? Gateway residency without model residency is half a solution.
  4. Is there a DPA and GDPR documentation you can hand to legal?
  5. Does governance come built in? SSO, RBAC and audit logs are usually the next blocker after residency.

If you want to see the EU gateway in detail — including the Frankfurt architecture and the list of EU-region models — start at requesty.ai/eu, or compare Requesty and OpenRouter feature by feature.

Frequently asked questions

Does OpenRouter offer EU data residency?
OpenRouter routes through US infrastructure and inherits the compliance posture of each upstream provider. If your workload requires guaranteed EU data residency, you need a gateway hosted in Europe — Requesty's EU gateway runs in Frankfurt with zero data retention.
What is an EU AI gateway?
An EU AI gateway is an LLM router whose infrastructure runs inside the European Union, so prompts and completions are processed under EU jurisdiction. Combined with EU-region models (for example Claude on AWS Bedrock eu-central-1 or Gemini on europe-west), the full request path stays in Europe.
Which AI models are available with EU data residency?
Through Requesty's EU gateway you can route to 77 EU-region model deployments, including Claude Opus and Sonnet on Bedrock eu-central-1/eu-west-1, Gemini 2.5 on europe-west, and GPT-5 on Azure France Central — all through one OpenAI-compatible API.
Is switching from OpenRouter to an EU gateway hard?
No. Both expose an OpenAI-compatible API, so the migration is a base-URL and API-key change. Your model IDs keep the same provider/model format.
Related reading