Deployment Model

Blackridge is currently delivered as a guided assessment or customer deployment. Public self-serve installation is not available.

In a deployment, Blackridge sits inline for model traffic, writes body-free economic evidence, and exposes that evidence through the GraphQL investigation API: attribution, usage, cost, routing, lineage, evidence grades, and findings. The first assessment is about validating evidence quality against real traffic, not asking your team to build from source.

1. Runtime evidence

Blackridge's runtime path is the inline token economics control point for observed model traffic. For a first assessment, the key idea is simple: route existing model clients through Blackridge, then confirm body-free economics evidence is flowing.

2. Zero-code attribution

Before instrumenting anything, configure the attribution resolver ladder against identity you already have — IdP tokens (EntraID, Okta, Keycloak), trusted proxy headers, Kubernetes workload identity, or API-key mapping. Most environments reach useful ownership coverage without touching application code.

3. Assessment surfaces

The gateway component is enough to start collecting runtime evidence. Add the investigation API and production deployment components when you are ready to review findings, expose role-based UI views, or automate assessment exports.