Private beta · Q4 2026 Azure Spain Central, Madrid

The OpenAI- and Anthropic-compatible AI API that keeps your data where it belongs.

Designed for Azure Spain Central (Madrid). Each project sets its residency tier — Spain, EU or global — and the gateway enforces it on every request: if no allowed route is available in that tier, the request fails instead of leaving it. The catalogue shows which models have a verified route in each tier.

Situra console dashboard: spend, requests, tokens, latency and the traffic split by residency tier. Situra console dashboard: spend, requests, tokens, latency and the traffic split by residency tier.
Real console with demo data.
  • OpenAI and Anthropic compatibleOnly the base URL and the key change
  • Provider list pricesNo token markup
  • Metadata-only by defaultContent is not stored unless turned on
  • Designed for ENS AltaENS Media + ISO 27001 audit targeted Q2–Q3 2027

ES ⊂ EU ⊂ Global. One tier per project, enforced on every request.

A project on the EU tier may use EU and Spain routes, never global ones. If every allowed route fails, the gateway returns an error instead of looking for an alternative outside. Pick a tier and simulate a failure.

Project residency tier
Model family
Simulate a provider failure
Route diagram for a project in the EU tier and the Claude family.GlobalEUESSituraMadridAnthropicBedrock EUVertex EUVertex / Foundry ES
Conceptual diagram, not a geographic map. Each ring contains the inner ones: ES ⊂ EU ⊂ Global. The gateway sits at the centre, in Madrid.

Served by Amazon Bedrock (EU) on the first attempt.

Gateway attempts

  1. Attempt 1EUAmazon Bedrock200 OK
  2. —Anthropic API (Global) is never tried: it is outside the project's EU tier.

Fallbacks only go to routes in the same or a stricter tier. Never to a looser one. Retries happen only before the first byte is streamed to the client.

Response (example)

HTTP/1.1 200 OK
x-situra-residency: eu
x-situra-route-region: eu
x-situra-attempts: 1

Routes for this family

  • GlobalAnthropic APIapi.anthropic.com · global processingBlocked · looser tier
  • EUAmazon BedrockEU regional endpoint · list price +10%Serves the request
  • EUGoogle Vertex AIEU regional endpoint · list price +10%Allowed · standby
  • ESVertex Madrid / Foundry Spain Centralto verifyAvailability in Spain to be verifiedAllowed · standby

Planned catalogue, for illustration. The live catalogue and per-region availability are shown in the console and by GET /v1/models.

Two separate questions

The platform

Azure Spain Central, Madrid

Gateway, control plane, console, metrics and credit ledger run in the Madrid region. This is the ENS scope.

The model

Whichever endpoint the tier allows

The project’s residency tier decides: a regional endpoint in Madrid, in the EU, or global. The model provider is a sub-processor and is listed as one.

Residency and compliance in detail

A console for the people accountable for AI use.

Projects, keys, budgets, usage and billing in one place, with the residency tier visible in every view.

Project residency — Each project sets its tier — Spain, EU or global — and the console shows which models and routes remain available at that tier. Project residency — Each project sets its tier — Spain, EU or global — and the console shows which models and routes remain available at that tier.
Real console with demo data.

Explore the platform

The provider’s list price. No token markup.

Each route costs what its provider publishes, converted to euros. Situra charges a transparent fee on credit top-ups.

  • 0%markup on the per-token price
  • 5.5%launch fee on credit top-ups
  • €0with sandbox test keys, CI included
Euros per million tokens (input / output) for the route each tier uses. Illustrative prices: live prices per route are shown in the console and by GET /v1/models. * Route still to verify.
Model ESSpainEUEuropean UnionGlobal
InputOutputInputOutputInputOutput
Claude Sonnet 4.6Anthropic 3.10* 15.48 3.10 15.48 2.82 14.08
Claude Haiku 4.5Anthropic 1.03* 5.16 1.03 5.16 0.94 4.69
GPT-5.2OpenAI 1.81* 14.45 1.81 14.45 1.64 13.14
GPT-5 miniOpenAI 0.26* 2.06 0.26 2.06 0.23 1.88
Gemini 2.5 FlashGoogle 0.31 2.58 0.31 2.58 0.28 2.35
Mistral Small 3.2Mistral AI 0.10* 0.31 0.09 0.28 0.09 0.28
DeepSeek V3.2DeepSeek 0.28* 1.13 0.28* 1.13 0.28* 1.13

Change the base URL and the key. That’s it.

The official OpenAI and Anthropic SDKs, curl and Claude Code work as they are. The project decides residency, not your code.

  1. Create a project Pick its residency tier: Spain, EU or global.
  2. Generate a key With an expiry and a model allowlist. situ_test_ keys go to the sandbox, free.
  3. Point your client Change the base URL and the key in the SDK you already use.
app.py
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.situra.ai/v1",
    api_key=os.environ["SITURA_API_KEY"],  # situ_live_…
)

response = client.chat.completions.create(
    model="gpt-5-mini",
    messages=[{"role": "user", "content": "Summarise this case file in three sentences."}],
)
print(response.choices[0].message.content)

Security you can check.

Each control is described in enough detail to verify it. The trust center brings together the architecture, the real compliance status, sub-processors and documents.

Staff never see content
No internal API returns prompts or responses. Content logging, if enabled, is encrypted under the organisation’s key.
Hash-chained audit log
Append-only; every row carries the hash of the previous one. The chain can be verified and exported.
Four-eyes approval
Suspensions, credit adjustments, provider credentials and second-factor resets need a second approver.
Residency attestation
Every response carries signed evidence of where it was processed, verifiable with our public keys.

Frequently asked questions

More pricing questions
Is Situra ENS certified?

Not yet. Situra is designed for ENS category Alta, and the ENS category Media + ISO/IEC 27001 certification audit is targeted for Q2–Q3 2027.

Azure services in the Spain Central region are certified at ENS Alto by Microsoft. That certification belongs to the cloud provider, not to Situra.

Where is data processed?

The platform (gateway, control plane, console and metrics) runs in Azure Spain Central, Madrid. The model runs on whichever endpoint the project’s tier allows: Spain, the EU or global.

What happens if no route is available in the project’s tier?

The request fails: 503 when no allowed route exists and 502 when every allowed route has failed. The gateway never falls back to a route in a looser tier.

Are prompts and responses stored?

Not by default: only metadata is recorded (tokens, cost, latency, route). Content logging is opt-in per project and encrypted under the organisation’s key; Situra staff cannot access it.

Do we need to change our code?

Only the base URL and the key. Situra is compatible with the OpenAI and Anthropic SDKs, curl and Claude Code.

How much does it cost?

The provider’s list price for each route, with no token markup, plus a 5.5% launch fee on credit top-ups. Public administrations can pay by invoice.

Who can join the beta?

Three to five Spanish public administrations and regulated organisations, from Q4 2026. Requests go through the contact page.

Private beta · Q4 2026

We are looking for three to five organisations for the beta.

Spanish public administrations and regulated organisations that want to use AI models with data residency under control. We work with each organisation on ENS categorisation, the DPA and the integration.