Versa extends zero‑trust to MCP
- Versa said on May 21 it added a patent-pending zero-trust architecture for Model Context Protocol actions inside VersaONE to validate AI-agent steps before execution. - Versa said every agent action is checked against user identity, role-based access controls and system policies, with human approval when administrators require it. - Versa described the controls in a May 21 news release and an April 30 product blog post on its Zero Trust MCP Server.
Versa said on May 21 that it had added zero-trust controls for Model Context Protocol, or MCP, actions inside its VersaONE platform, expanding the company’s security model from user and device access to the individual steps taken by AI agents. The company said the new architecture is patent-pending and is designed to validate each agent-generated action before it is executed. Versa positioned the feature as a response to a growing enterprise problem: AI agents can trigger multiple actions across security and networking systems from a single prompt. The announcement was published in a company news release and follows an April 30 product blog post describing Versa’s Zero Trust MCP Server. ### What exactly did Versa say it is securing? Versa said the new controls apply to MCP, a protocol used to connect AI applications and agents to external tools and systems. In its March 18 blog post, founder and chief development officer Kumar Mehta described MCP as a standard that lets AI applications connect to tools, and said the company viewed tool access as a separate control point from model access. (versa-networks.com) The May 21 release said no AI action is implicitly trusted under the new design. Instead, each step is validated against user identity, role-based access controls and system policies before execution, with explicit human validation when administrator-defined policies require it. ### Where does this sit inside Versa’s product stack? Versa said the capability is integrated with Versa Verbo, its AI-powered operations co-pilot, and with the VersaONE Universal SASE platform. (versa-networks.com) The company’s product pages describe VersaAI as the AI layer inside VersaONE and say the broader platform combines networking, security, AI-powered operations and edge compute. (versa-networks.com) Versa’s documentation published in May said its agentic AI capability uses MCP to let network engineers, managed service providers and customers interact with Versa SD-WAN, SSE and SASE control planes. The same documentation said a ToolCaller component uses MCP servers and connectors for systems including Director and Concerto. (versa-networks.com) ### How is the company describing the control model? Rajesh Kari, a director of product marketing at Versa, wrote on April 30 that the Zero Trust MCP Server is intended to secure “every AI interaction,” enforce governance and support agentic AI adoption at scale. The post said the system applies identity verification, policy enforcement and governance checks to AI-driven actions rather than trusting a single user prompt as sufficient authorization. (docs.versa-networks.com) The May 21 release said the architecture validates not only who initiated a request, but whether the requested task is permitted under enterprise policy. Versa said those checks are tied to orchestration and approval workflows, creating an audit trail around agent actions. ### Why is Versa focusing on per-task validation? Versa said a single AI prompt can produce multiple downstream actions across critical network and security systems, which creates a trust gap if only the original user session is authenticated. (versa-networks.com) The company said AI systems can also misinterpret intent or generate unintended actions, making step-by-step validation necessary in operational environments. (versa-networks.com) Kevin Sheu, Versa’s vice president of product strategy and solutions, told industry coverage that the capability is embedded within Versa Verbo. That report said the company is extending zero-trust controls to the actions taken through MCP workflows, not just to the user who launched the request. ### What comes next from here? Versa’s next public reference point is its existing MCP product line. (versa-networks.com) The company launched an MCP Server in April 2025 and published connection documentation for Versa MCP Server desktop clients in mid-May 2026. The May 21 release and April 30 blog post now serve as the company’s main public documents for the zero-trust extension, with Versa Verbo and VersaONE named as the products carrying the controls forward. (versa-networks.com) (newsbreak.com)