Architecture

One gateway. Every model.
Policy applied at the wire.

Shield sits between your users and every model provider — inspecting, routing, and logging every request against your governance policy.

Shield architecture: dispatcher, model registry, ZDR enforcement, and surface banners

Flow

The dispatcher classifies each request and applies policy. The registry picks the best-fit model — respecting cost, latency, residency, and provider health. The ZDR layer strips retention and grounds the prompt before it ever reaches a provider. Surface banners give users clear signal on the outbound channel.

Feature grid

What ships in the box.

Model Governance Matrix

Every provider, side by side — ZDR posture, residency, model version, cost.

Provider Health

Live signal on latency, error rate, and rate-limit headroom, per provider.

Weekly Governance Digest

One PDF for the board: usage, exceptions, policy violations, remediation.

Paste Sentinel extension

Browser-level guardrail. Intercepts sensitive fields before they leave the surface.

Surface Banners

In-context signal so employees know when they're on-policy — and when they're not.

Model Registry

Approved providers, versions, and defaults — governed like any other production dependency.

Model Governance

The Model Governance Matrix — provider posture at a glance.

/insights/ai-command-center/models
Provider
ZDR
Residency
Health
OpenAI · GPT-4.1
Enforced
US / EU
Nominal
Anthropic · Claude 4
Enforced
US
Nominal
Google · Gemini 2.5
Enforced
US / EU
Degraded
Azure OpenAI
Enforced
EU
Nominal

Screenshot preview from the Shield Model Governance page. Live data in the app.