PLATFORM

The gateway, in depth.

Everything below is a verified capability of the shipped product — no roadmap slides mixed into feature lists. What we don't do yet lives on the Company page, in the open.

Reference architecture

Customer apps Internal copilots Agents / MCP tools TruCert Gateway — YOUR estate one API · budgets · rate limits · virtual keys failover · load balancing · caching guardrails · logging → your SIEM MCP gateway · SSO at your ingress K8s or VMs · HA · air-gap capable Frontier providers EU endpoints Your self-hosted models Optional outbound-only TruCert connection for license/update checks — disableable; air-gap mode uses offline bundles

Capabilities

Verified feature set of the underlying engine plus TruCert packaging.

Unified API

One OpenAI-compatible endpoint

100+ providers behind one API. Swap or mix models without touching application code; native-format passthrough where needed.

Cost control

Metering, hard budgets, rate limits

Per key, team and application. Virtual keys with full lifecycle: create, scope, rotate, revoke — centrally.

Resilience

Failover & balancing

Automatic retries, provider fallbacks, load balancing, response caching.

Governance

Guardrails & logging

PII masking, prompt-injection detection, secret detection hooks — policies configured per deployment. Logs stay under your retention rules.

Agents

MCP gateway

Agent and tool traffic governed with per-key/team permissioning — the traffic class auditors will ask about next.

TruCert packaging

The hardened distribution

Quarantine Channel releases: verified, re-signed, SBOM included, license-compliance automated in the pipeline, enterprise-directory-free build, hardening defaults (no external telemetry, TLS required, admin surface off by default). Runbook library shipped and version-tested — runbooks that aren't tested on a version aren't shipped.

Deployment models

Standard

Helm/K8s or VM install, PostgreSQL, your ingress + IdP in front, logs to your SIEM, allow-listed egress to chosen providers.

Air-gap

Identical, with internal model endpoints as providers and offline update bundles. No external calls of any kind — guardrails run in-cluster.

Sovereign cloud

Infrastructure-agnostic: the same pattern on EU sovereign-cloud landing zones (Proximus, KPN/STACKIT, hyperscaler EU sovereign regions).

Support, plainly

Your team operates the gateway 24/7 — it's your infrastructure. We make that safe: hardened releases, tested runbooks, expert escalation in business hours (Mon–Fri 09:00–17:00 CET). Managed operations ("Operate") is on the roadmap — waitlist open.

SeverityDefinitionResponse target
S1Production gateway down or all AI traffic blocked, no workaround4 business hours
S2Major degradation with workaround8 business hours
S3Minor defect2 business days
S4Question / how-to3 business days

Targets are response, not resolution. Versions supported: current + previous quarantined release (N, N–1). Full SLA schedule in the subscription agreement.

See it run in 15 minutes.

Live demo: budget hard-stop, 8-second failover, one-click revocation, and an evidence export from the demo's own traffic. Nothing faked.

Book the demo Run the readiness check first