Ask any NOC or SOC engineer what actually eats their day, and it’s the twenty minutes before taking action in a console: logging into the SASE console to check a path, tabbing to the switch manager to check a port, opening a third tool to check the Access Point, and a fourth to check the endpoint. Versa’s own field data puts a number on it: customers report spending up to 70% of incident response time simply gathering information from different systems before they can act on any of it.
That fragmentation tax is what the industry is now racing to eliminate, and enterprises are converging on the Model Context Protocol (MCP) as the way to do it: a common interface for connecting AI agents to every product in the stack, rather than a one-off integration per vendor per tool.
At a high level, the flow looks the same regardless of which vendors sit behind it: a request in plain language flows through an AI assistant layer to an enterprise routing agent, which fans out to a set of governed MCP servers (Versa’s included) that each speak for one piece of the infrastructure, all wrapped in a single zero-trust governance layer.
Unifying the Technology Stack
In 2026, no single vendor is betting on MCP alone. The entire enterprise infrastructure stack is converging on it independently, each from its own angle:
Network platform vendors are shipping MCP servers that expose telemetry to helpdesk assistants, cross-vendor ops copilots, and custom agents. They are often paired with new governance tooling (e.g. AI bills-of-materials) to control which MCP servers and agents are even allowed to run. Security platform vendors are folding MCP support directly into their SOC platforms, wiring in thousands of playbooks and integrations, while also selling identity controls purpose-built for MCP data access, arguing that MCP servers are “vital bridges to enterprise data” but dangerously inconsistent on authentication. Cloud-managed networking vendors have extended their AIOps platforms with MCP support so third-party agents can plug into wireless and campus data through no-code connectors. And security vendors have gone a step further, launching AI brokers built specifically to sit inline on MCP and agent-to-agent traffic.
Versa’s own move came earlier than most: the Versa MCP Server shipped in 2025 as an open-source utility letting Claude, ChatGPT, Gemini, Copilot, or an internal copilot query the VersaONE platform directly (appliance health, alarms, routing tables, templates, bandwidth) through one API layer instead of five consoles. Versa’s engineering team has reported a 45% reduction in mean-time-to-resolution among customers using it.
Put those data points together and the theme is clear: nearly every major vendor in networking and security, Versa included, is building an MCP front door into its product. Enterprises buying across this landscape are about to have a lot of front doors.
Empowering the Operator
A wall of vendor-specific MCP servers, each one a clean doorway into its own product, solves the “one tool per vendor” problem but not the “one enterprise” problem. An engineer troubleshooting Wi-Fi doesn’t experience the network as a stack of vendor logos. They experience it as their network: this branch, this SSID, this user, this SLA. The MCP servers know their own product. They don’t know each other, and they don’t know the enterprise’s topology, entitlements, or history.
That gap is exactly where the next layer of value is emerging: an enterprise routing and orchestration layer. It sits between the user’s prompt and the fleet of vendor MCP servers, and it gets smarter about a specific enterprise’s infrastructure the longer it runs.
Here is what that flow looks like end to end. A NOC or SOC analyst types something as plain as “MS Teams calls keep freezing, can you find out why” into whichever LLM their organization has standardized on: Claude, OpenAI, Gemini, or Copilot. The prompt lands on an enterprise routing agent, which decomposes the question, decides which domains it touches, and fans it out to the right MCP servers. Versa MCP handles the WAN and security path. A network infrastructure MCP covers the switch it rides on. A wireless/access MCP covers the access point. An endpoint MCP covers the client itself.
Each server queries its own system of record and hands structured context back. The routing agent correlates the results and returns one answer, in plain language, instead of five dashboards’ worth of raw telemetry.
That correlation layer is the “local intelligence”: not a chatbot bolted onto one product, but an operational brain that understands this organization’s specific mix of vendors well enough to make every one of them faster:
- Time to provision. New branch, new SSID, new segmentation policy. Instead of sequencing changes by hand across a SASE console, a switch manager, and a wireless controller, one governed request propagates the intent to each relevant MCP server. The routing layer then confirms consistency across all three.
- Time to operate and troubleshoot. Root cause increasingly lives at the seam between vendors: a WAN path degrading because of an upstream switch buffer, a wireless client roaming into a segment with the wrong firewall policy. A routing layer with visibility across MCP servers can correlate WAN, LAN, and security telemetry in one pass, instead of asking a human to hold five systems in their head.
- Kill chain mapping and threat hunting. Frameworks like MITRE ATT&CK already give defenders a shared language for adversary behavior, but mapping a real kill chain still means stitching together firewall logs, identity events, endpoint telemetry, and network flow data by hand. An MCP-connected routing layer can pull ATP, UEBA, and firewall context from Versa alongside endpoint and switch data from other vendors. A hunt that starts with “show me lateral movement from this host in the last 24 hours” no longer requires five separate queries and a spreadsheet to reconcile them.
Where This Leaves SASE Architecture
For enterprises already moving SD-WAN, SSE, and SD-LAN onto a unified SASE platform, this is just the next step. VersaONE was already built as a converged data plane, so network and security context doesn’t sit in silos. Versa’s Zero Trust MCP Server extends that same governance to AI agents, so every request is authenticated, scoped, and audited.
No single vendor’s MCP server solves this alone. The enterprises that get ahead will treat the routing and governance layer above all their MCP servers as core infrastructure. Do that, and MCP becomes the operating model for enterprise infrastructure, not just an integration trick.
Learn more about how Versa is enabling secure, governed AI operations with the Zero Trust MCP Server and VersaONE.
Ponnaiyan leads solutions engineering at Versa Networks, working with enterprise customers on SASE, SD-WAN, and Zero Trust architecture.