PARTNERS

Add selected Workbench capabilities through bounded OEM and partner integrations

Model Gateway Technical PreviewPolicy Router
by AI Security LLC › preview and evidence trail
Fixture-drivenPolicy review

Preview surface

One interface, many executors, with policy first.

The demo shows how the gateway routes requests across local tools and hosted APIs while preserving a reviewable evidence trail. It is a preview, not a promise of a specific provider or deployment.

Routing targets

Where requests can go.

Claude Code CLI

Codex CLI

Gemini CLI

Local model

Hosted API

Fallback chain

Policy modes

AllowRedact then allowReview firstBlockLog only

Evidence capture

Prompt and response logsRouting decisionsApproval eventsBlocked requestsRedaction eventsAudit export packages

What it is not

This preview does not imply that every executor is enabled or that a specific model provider is approved. It exists to show how routing, policy, and evidence fit together.

Evidence-Grounded Steps in Attack Path Analysis — from gateway policy findings

1 attack paths1 tools in this fixture4 evidence-graph nodes2 evidence-graph edges

Evidence from this surface, normalized and joined into the shared evidence graph. Every step traces to a specific tool’s output; grounded steps were observed or deterministically derived.

Adding a mandatory approval/control at model-gateway:target_action_executor blocks 1 of 1 analyzed paths.

  • Groundedattack· conf 90%11model-gateway
    Untrusted inputaction_executor
    owasp_llm:LLM06
LegendGroundedInferredSpeculativeOpen Evidence Graph