The AI control plane
your org was missing.
The cloud bill. The compliance audit. The breach notification. That's how most companies find out they needed this. Aqta is what you install before any of those happen.
Every request. Governed. Before it becomes a problem.
A 4-stage enforcement pipeline runs on every request across your org. PII in a prompt, a budget-blind agent, a loop running all weekend, all caught before a single token is wasted or a policy is violated. Gateway overhead is measured and returned in every response.
Incoming prompt
Choose a scenario above, then hit Run demo →
One line. No refactoring. Full control.
You shouldn't have to rewrite your AI stack to get visibility into it. Change base_url to api.aqta.ai/v1 and you're done.
import openai
# Drop-in: change base_url only
client = openai.OpenAI(
api_key="sk-aqta-...",
base_url="https://api.aqta.ai/v1",
)
response = client.chat.completions.create(
model="gpt-4o",
messages=[{"role": "user", "content": "Hello"}],
)
# Standard OpenAI response + aqta metadata object
print(response.aqta["cost_eur"]) # → 0.0024
print(response.aqta["trace_id"]) # → "tr_abc123"
print(response.aqta["latency_added_ms"]) # → 8.3The response your code already expects, plus everything missing
Standard OpenAI-format response, unchanged. Plus an aqta object with the cost, trace, and policy data you never had before.
{
"id": "chatcmpl-9xKp2mQw",
"object": "chat.completion",
"model": "gpt-4o",
"choices": [{
"message": {
"role": "assistant",
"content": "Here are the top 3 growth opportunities…"
},
"finish_reason": "stop"
}],
"usage": { "prompt_tokens": 42, "completion_tokens": 270, "total_tokens": 312 },
// Added by Aqta, zero changes to your existing code needed
"aqta": {
"trace_id": "tr_9xKp2mQw",
"cost_eur": 0.0024,
"provider": "openai",
"model": "gpt-4o",
"cache_hit": false,
"latency_added_ms": 8.3,
"loop_score": 0.04,
"policy_applied": ["pii_scan", "loop_guard", "budget_guard"],
"status": "passed",
"analytics": {
"response_time_ms": 847.2,
"tokens_per_second": 368.6
}
}
}Three moments every AI org hits. Usually too late.
Each one is preventable. Each one has already happened to a company like yours.
The surprise bill
A five-figure AI invoice hit the cloud bill. Nobody owned it, nobody saw it coming. Per-team, per-model cost attribution, so it never happens to you.
The compliance audit
Legal asks: what AI touched what data, under what policy, and when? Most teams scramble for two weeks. Aqta has the tamper-proof answer ready before they ask.
The data incident
Patient records, financial data, privileged documents, sent to a third-party model. Aqta scans and blocks at the gateway before a single token crosses your perimeter.
The EU AI Act is law. Your regulator won't wait for you to get ready.
Aqta maps every enforcement action to the relevant article automatically. When a regulator asks for evidence, you export a conformity package, not a spreadsheet you built the night before. Built in Europe, for European orgs, by people who read the actual regulation.
EU AI Act Compliance Coverage
First European AI gateway built for EU AI Act. Generates conformity evidence automatically.
Budget enforcement, runaway detection, and cost anomaly alerts, continuous policy-linked risk controls on every request.
HITL routing and dual-approval workflows for high-stakes decisions. Available in Health and FinServ packs.
PII scanning, adversarial input detection, jailbreak prevention, and cross-deployment threat pattern sharing.
Tamper-evident audit trail with SHA-256 hash chain on every proxied request. Queryable and exportable.
Auto-generated conformity evidence packages exportable as PDF for regulatory submission.
Conformity evidence packages exportable as PDF or JSON for regulatory submission · Check your AI Act status →
Your industry has specific rules. We've already mapped them.
Healthcare CISOs, bank compliance teams, and legal ops leads all have different nightmares. Here are the ones we've already solved.
Health Pack
Early AccessPHI scanning, HITL routing, and 7-year clinical audit trails.
- PHI / PII scanning on every prompt
- HITL routing for high-risk queries
- 7-year tamper-evident receipts
- Bias detection across demographics
FinServ Pack
Early AccessDual-approval workflows and immutable audit trails for financial AI.
- Dual-approval for high-risk operations
- Transaction limit enforcement
- AML/KYC compliance automation
- Immutable financial audit trails
Legal Pack
Coming SoonPrivilege classification, conflict detection, and immutable legal audit trails.
- Attorney-client privilege scanning
- Conflict-of-interest detection
- Jurisdictional compliance flags
- Immutable legal audit trails
Industry packs start at €49/mo (Starter). Free tier includes basic cost tracking and 7-day traces.
Works with your entire AI stack
GPT-4o · GPT-4 · o3-mini
Claude Sonnet 4.6 · Opus 4
Gemini 1.5 Pro · 2.0 Flash
Sonar Pro · Sonar
More providers: Mistral, Llama (AWS Bedrock), Cohere, and any OpenAI-compatible endpoint
Ship fast. Stay compliant.
The two aren't mutually exclusive.
Your engineers won't notice it's there. Your CISO will. Live in 60 seconds, starts free.
No credit card · Free evaluation · GDPR-native · Hosted in EU