Guided Narrative Demo

The Golden Path:
Intent → Action → Outcome

See how FuseGov closes the loop on agentic governance by enforcing intent at the boundary and verifying outcomes with human accountability.

1

1. Setup: Registry & Policies

The agent is registered in the service catalog with its owner. A versioned policy bundle is deployed to the Gateway PEP, defining that 'Write' operations to production databases require step-up approval.

2

2. Action: Risky Tool Call

The agent attempts to execute a tool call: `delete_customer_record(id='cust_99')`. This action is intercepted by the FuseGov PEP at the network boundary.

3

3. Decision: Escalate for Approval

The PEP identifies the action as high-risk based on the registry tier. Instead of allowing it, it triggers an escalation. The Agent Owner receives a notification to approve or reject.

4

4. Outcome: Execution & Verification

The owner approves the one-time action. The PEP allows the tool call to proceed to the database. FuseGov observes the state change to verify it matches the intent.

Enforcement Sequence

The sub-millisecond handshake between agent, policy, and human approver.

Demo Result

Outcome aligned / not aligned
Human owner notified
Evidence pack generated

"By closing the loop with evidence and verification, we allow agents to solve complex problems without creating unmanageable risk."