Versa applies zero-trust controls to AI agent actions with new MCP architecture

Secure access service edge firm Versa Networks Inc. today introduced a zero-trust architecture for the Model Context Protocol that validates every action an artificial intelligence agent takes inside its network operations co-pilot before that action executes.

The design sits within Versa Verbo, the company’s AI-powered operations co-pilot and is integrated with the VersaONE Universal SASE Platform. Versa is pitching it as a fix for what it calls a trust gap in enterprise agentic AI, where a single user prompt can quietly trigger multiple downstream steps across network and security systems with no real visibility into what the agent actually did.

Under the architecture, no AI-generated action is implicitly trusted. Each step is checked against user identity, role-based access controls and system policies before it runs, with explicit human approval required when administrator-defined policies demand it. Administrators set those policies in advance to determine which agent actions execute automatically, which require sign-off and which are blocked outright, based on identity, role, system context, action type and risk level. Every approved action is logged with full attribution.