One endpoint for every European model your team will ever need.
Euronium Gateway speaks the OpenAI protocol on one side and every major EU inference provider on the other. Add models without rewriting clients, swap providers without redeploying, and route by cost, latency or jurisdiction.
- EU-only egress · hard 403, not a warning
- PII masked pre-egress · Presidio-aligned
- Ed25519-signed, hash-chained receipts
- Open source · Apache-2.0 · self-host or Cloud
Built like a real platform, not a wrapper.
OpenAI-compatible
Drop-in /v1 surface. Keep your SDK, keep your prompts.
100+ EU model routes
Mistral, Aleph Alpha, vLLM, OVH, Scaleway, IONOS — out of the box.
Receipts your DPO can read
Hash-chained Ed25519 ledger you can verify offline.
Per-team virtual keys
Budgets, rate limits, env scopes — no shared provider creds.
Native OpenTelemetry
OTLP exporter to Datadog, Honeycomb, Grafana.
Self-host or Cloud
docker compose on day one. Frankfurt / Paris when you scale.
Answers, not hedges.
Do I need to rewrite my OpenAI client?+
No. Change base_url to https://gateway.euronium.eu/v1 and your existing OpenAI / LangChain / Vercel AI / Agno code works as-is.
What happens if a request would leave the EU?+
The gateway returns HTTP 403 with policy=schrems-ii before bytes hit the wire. There is no silent fallback. Every blocked call is recorded with a signed receipt.
Is the core really open source?+
Yes. Gateway, Protocol, and Dashboard are Apache-2.0. Paid tiers add support, managed infrastructure, and SSO/SCIM.
Related pages
Deploy Euronium in 2 minutes.
Open source. Apache-2.0. Dev is free forever. One line in your OpenAI client and your AI calls become sovereign, masked, and signed.