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