PART OF HOOKBUS ENTERPRISE

AgentProtect CRE

When there is no human in the loop, CRE is the Agent in the Loop.

The policy enforcement component inside HookBus Enterprise. Every consequential tool call can pass through AgentProtect CRE before execution. The AI cannot bypass, disable, or argue with it.

Patent Pending GB2604445.3 Runs on HookBus™

How it works

Deny. Allow. Ask. Three outcomes. One gate.

HOOKBUS EVENT

Agent tool call intercepted

AGENTKNOWLEDGE

Inject grounded policy context, business rules, and relevant local information before evaluation.

AGENTPROTECT CRE

Policy Evaluation

L1: <10ms deterministic L2: Probabilistic alignment

DENY

Blocked. The agent cannot execute this action. Full audit record.

ALLOW

Permitted. Tool call passes through. Agent continues.

ASK

Human approval required. AgentNotify pushes to user.

APPROVER NOTIFIED

Approver receives a notification link. Opens the approval UI in seconds. Approve or deny from any device. No PINs, no shared secrets. Full audit trail.

PROBABILISTIC GOVERNANCE

L2 evaluates agent alignment: is this tool call coherent with the user's instruction? Or has the agent drifted, substituted goals, or reinterpreted policy? Catches floundering before it causes damage.

Two layers

Deterministic speed. Probabilistic depth.

LAYER 1

Agentic Thinking deterministic patterns

Under 10ms

AgentProtect CRE's enterprise L1 is our deterministic pattern layer: explicit action, resource, scope, risk, and organisation-policy patterns evaluated before consequential actions execute.

LAYER 2

Agentic Thinking probabilistic patterns

Customer-approved model path. Local or private-cloud capable.

AgentProtect CRE's enterprise L2 is our probabilistic pattern layer. It evaluates ambiguous intent, alignment, and policy context using the customer's approved inference path, catching substitutions, shortcuts, and creative reinterpretations that deterministic rules alone miss.

Features

What CRE does that system prompts cannot.

✓

Mechanical enforcement

Operates outside the AI's context window as a mandatory checkpoint. The AI cannot skip, modify, or argue with it.

✓

Intent alignment

L2 evaluates whether the agent's tool call matches the user's explicit instruction and surrounding policy context.

✓

PIN override

Human approval chain for sensitive operations. The user types override 0000, CRE retries, the audit trail records it.

✓

Self-learning rules

L2 decisions automatically promote to L1 patterns with human approval. Faster and more accurate over time.

✓

Anti-evasion

Detects encoded commands, lateral movement, and scripts designed to bypass enforcement.

✓

SOC2-ready, ISO 27001-ready, ISO 42001-ready

Full audit trail, every decision logged. Works with HookBus™ Auditor for compliance evidence.

Free tier