ROUTE
autoLet the gateway choose.
Capability-aware selection with failover. Send model=auto and get a capable model without maintaining routing rules.

eggsplain API
The sovereign gateway for every model. Keep your OpenAI SDK and decide, per request, whether AI runs in the cloud, inside the EU, or on hardware you own.
# Same SDK. Same call. One string decides where it runs.
from openai import OpenAI
client = OpenAI(
base_url="https://api.eggsplain.com/v1",
api_key="sk-egg-…",
)
client.chat.completions.create(
model="sovereign/gpt-6-sol:eu!",
messages=[{"role": "user", "content": "Summarise this contract."}],
)Pinned to the EU data zone. The trailing ! fails closed.
One endpoint. 10+ model families. On your premise.
01 One string
ROUTE
autoCapability-aware selection with failover. Send model=auto and get a capable model without maintaining routing rules.
PIN
sovereign/…:eu!Add :eu to run inside the EU data zone. Add ! and the request fails closed instead of drifting elsewhere.
OWN
onprem/your-nodeRegister an OpenAI-compatible endpoint on your Tailscale network. It stays isolated to your organization and never falls back to cloud.
02 Sovereign by design
Every response tells you where it ran. Pin a request to the EU and the gateway returns what was requested and what was served. If it can't be honoured, a strict request fails instead of quietly going elsewhere.
:eu :se :no with ! to fail closed
x-egg-requested-modelsovereign/gpt-6-sol:eu!
x-egg-residency-requestedeu
x-egg-residency-servedeu
status200 OK
03 Control
Every team gets its own boundary for models, keys and analytics.
Owner, Admin and Member. Admins manage policy; Members use the gateway.
Personal keys for people, service keys for integrations. Revocable, attributable.
Immutable audit events and per-request token analytics, kept in your environment.
04 Deployment
HOSTED
Start in minutes. We run the gateway, console and routing, behind a sovereign boundary.
Open consoleON-PREMISE
Gateway, console and routing run on your premises with your own models and nodes. No cloud dependency at runtime.
Talk to usPer-token prices for every model are public. See pricing

One endpoint. Your rules. Start building on infrastructure that stays yours.
Get started