Get started

Add the models you want to use

Pick from 100+ EU and global model routes. Start with one EU-hosted model (Mistral) and one global model (gpt-4o-mini) — you can always add more.

Apply default policies

Turn on EU-only egress and PII masking. These two cover most GDPR and Schrems-II concerns out of the box. Non-EU calls return a hard 403 with a signed receipt.

Create your first virtual API key

Scope the key to a team, environment, and budget. Rotate or revoke any time. Your existing OpenAI / LangChain / Vercel AI code keeps working — only the base URL changes.

Wire it into your app

Point your OpenAI-compatible client at the Euronium gateway. Two-line change.

from openai import OpenAI

client = OpenAI(
  base_url="https://gateway.euronium.eu/v1",
  api_key="eu_live_xxx", # your virtual key
)

resp = client.chat.completions.create(
  model="mistral-large-latest",
  messages=[{"role": "user", "content": "Hallo Europa"}],
)

Send your first request

Run your app or curl the gateway. The request will route to the model you allowed, in the region you pinned.

See the request land

Watch latency, tokens, and provider attribution appear live in Traffic. Drill in to any call for the full trace.

See the cost

Spend in euros lands in Cost — per model, key, and team. Set a daily cap so you never get surprised.

See the proof receipt

Every request gets an Ed25519-signed, hash-chained receipt your DPO can verify offline. Replay or export to your SIEM.