HIOP Governed Agent Control Plane

Models propose. HIOP governs the effect.

HIOP is Hood Intelligence's vendor-neutral authority, execution-control, and evidence layer for consequential actions by AI agents, software, and machines. It evaluates whether a specific actor may cause a specific effect on a specific resource under current policy, then preserves evidence of the decision and observed outcome.

What HIOP governs

HIOP is the Hood Intelligence Operations Platform. It is designed to place an independent decision and evidence layer between an AI-enabled requester and a consequential effect. The question is not only whether an actor authenticated successfully. The question is whether that actor may cause this exact effect, on this exact resource, under the policy that applies now.

HIOP is model- and provider-neutral. It can govern actions requested through different models, agents, applications, human operators, and machines without granting any one provider the power to define its own authority.

Decision lifecycle

1. Establish identity

Bind the person, agent, service, organization, or asset requesting action.

2. Evaluate authority

Evaluate the requested effect, target resource, scope, time, approvals, and current policy.

3. Control execution

Allow, deny, constrain, interrupt, or revoke execution through the approved route.

4. Preserve evidence

Record what was requested, decided, executed, denied, and independently observed.

What HIOP does not claim

HIOP does not claim that a model is always correct, that a sandbox cannot fail, or that evidence proves every fact in the world. The public name for that effect-authority boundary is CRUSHIA AUTHORITY. CRUSHIA — HIOP Authority Engine determines what an agent, system, person, service, or asset is authorized to cause. The current HIOP engine sequence is Narwhal → Meerkat → CRUSHIA → Rain → Fossil. The current functional sequence is Identity → Verification → Authority → Execution → Evidence. Compass (former name), superseded by CRUSHIA. It is an operational control system for bounded authority and reviewable outcomes. Availability and deployment scope depend on the applicable plan, integration, and written commercial terms.

Architecture

Identity · Verification · Authority · Execution · Evidence

Authentication identifies the requester; it does not grant unlimited effects. Learning does not expand authority. HIOP does not claim to prove that AI will always be safe or correct.

Start an evaluation Read the Problem Pages