Interactive Tour

See how a corporate
AI gateway works

Follow an AI request from the terminal through an encrypted tunnel to a European data center — and back.

Scroll down

Sending a Request

A developer enters a command in the terminal. Just redirect BASE_URL and all communication goes through the corporate gateway.

Terminal

# Request routed to corporate gateway...

{

"model": "claude-opus-4-6",

"messages": [{

"role": "user",

"content": "Analyze this contract for GDPR compliance..."

}]

}

export ANTHROPIC_BASE_URL="https://gateway.company.com/v1"

One single change — redirect traffic to your corporate gateway

Encrypted Tunnel

The request travels through an encrypted tunnel. DNS policy blocks direct access to public AI services.

BLOCKED

api.openai.com

BLOCKED

api.anthropic.com

ALLOWED

gateway.company.com

The Gatekeeper — Nginx

The entry gateway verifies request security before it reaches the gateway.

TLS Termination

Encrypted connection with valid certificate

TLS 1.3

Rate Limiting

Max 30 requests per minute per IP address

4/30

Security Headers

HSTS, X-Frame-Options, CSP, X-Content-Type-Options

Request Size

10 MB limit per request

2.1 KB

Gate Open

Request forwarded to port 4000 → LiteLLM

The Brain — LiteLLM

Central control layer — verifies user identity, checks budget and allowed models, logs every request.

API Key Validation

Verify user identity

key: sk-dev-team-frontend ✓

Access Profile

Budget Check

Monthly limit for the team

€30.80 / €46.00 ✓

€0 67% used €46

Model Allowlist

gpt-5.4

claude-opus-4-6

gemini-3.1-pro

gpt-image-1

Model Aliasing

"model": "claude-opus-4-6"

"model": "azure/claude-opus-4-6-eu-west"

Developers use a simple model name, the gateway routes to a specific EU instance

Logging & Observability

Structured trace: timestamp, user, model, tokens, latency, cost

trace #4847

Private Connection

The request never leaves the private network. Public internet? Disabled.

Public access: DISABLED

Private IP

10.0.1.4

Private DNS

privatelink.openai.azure.com

Azure AI Foundry

Private Link

Data never leaves the Azure private network. No public access point.

AI Processing — EU Datacenter

The model processes the request in a European datacenter. Data never leaves the EU.

Ireland (EU West) Stockholm (EU North) Belgium (europe-west1)

AI Response

Content filtering: OK Zero data retention: Active Training opt-out: Default

Response Returns

The response returns the same secured path. Every step is recorded, costs update in real time.

Terminal — response

Cost Update

€30.80 €30.82

+€0.02 for this request (1,247 tokens)

Log Record

trace: #4847

user: dev-team-frontend

model: azure/claude-opus-4-6-eu-west

tokens: 312 in / 935 out

latency: 2.3s

cost: €0.02

Overview & Management

Everything in one place — access profiles, usage records, costs, and regulatory compliance.

Dev Sandbox

Active users12

Requests this month4,289

Cost€172.50

Business Docs

Active users8

Requests this month1,847

Cost€85.00

Restricted HR/Legal

Active users3

Requests this month421

Cost€26.10

Usage Records

2026-03-18 14:23:01 dev-sandbox claude-opus-4-6 → azure/claude-opus-4-6-eu-west — 1,247 tokens — €0.02

2026-03-18 14:22:45 business-docs gpt-5.4 → azure/gpt-5.4-eu-west — 2,891 tokens — €0.04

2026-03-18 14:22:12 dev-sandbox gemini-3.1-pro → vertex/gemini-3.1-pro-eu — 987 tokens — €0.01

2026-03-18 14:21:58 BLOCKED gpt-image-1 — model not in allowlist for dev-sandbox

2026-03-18 14:21:30 hr-legal claude-opus-4-6 → azure/claude-opus-4-6-eu-west — 534 tokens — €0.01 — redacted

Monthly Cost by Profile

Dev Sandbox €172.50

Business Docs €85.00

Restricted HR/Legal €26.10

Total: €283.60 / month

AI Act Ready NIS2 Compliant GDPR Aligned SIEM Export

Ready to take control
of your AI?

One-time delivery. No recurring fees. Infrastructure runs on your Azure or AWS.

View packages → Readiness Scan Book a call

← Back to product page

See how a corporate AI gateway works