Reference Agent Sandbox

See Orienta change an AI action before it reaches a customer.

A simulated customer-support workflow showing how an external AI agent can call Orienta before responding. The agent drafts a reply, Orienta reviews it, and unsafe actions are rewritten or held for human review.

1. DraftAgent prepares a reply
2. AuditOrienta checks the action
3. ActApprove or rewrite

Simulated Business Scenario

This is a fictional integration sandbox, not Orienta customer support. The backend uses Gemini when available, and every completed run is saved to MongoDB as an agent audit record.

Agent Execution

READY
Enter a scenario and run the agent.

Original Agent Draft

Waiting for the agent to draft a reply.

Final Governed Reply

Waiting for Orienta to approve or rewrite the action.
Action takenNot run
Risk score--
Risk flags--
Governance direction
--

Audit Record

The MongoDB run ID and collection will appear here.