Itential Puts AI Agents and MCP to Work

Itential is taking concrete steps toward a world of AI-driven network operations. This week, the network orchestration and automation provider (and Futuriom 50 member) announced an integration with startup Selector to provide closed-loop automation for spotting and remediating network issues. More broadly, it also announced its first Model Context Protocol (MCP) server, enabling further integrations with AI agents.

This should sound familiar. In April, we interviewed Itential CTO Chris Wade, who bemoaned the barriers facing true network automation. Traditional infrastructure isn't ready for the NetDevOps world where network and security operations are merged alongside application development, he maintained.

Itential's integration with Selector suits those goals. Selector's platform spots anomalous events and determines root causes. And yes, it's AI-driven, with agents digging through logs and telemetry data. What's new is that Selector can now hand off to the Itential platform to execute remediation—in a secure, policy-driven way.

The companies will demonstrate their integrated solution at Cisco Live in San Diego in June.

This automated triage-and-response is an omnipresent goal for both security and networking. It addresses the swivel-chair issue, where ops teams have to jump between alerts, configuration dashboards, and management dashboards. Moreover, operations teams are faced with a growing number of alarms, due not only to security breaches but also to the depth of modern observability tools; that can be a recipe for decision paralysis and alarm fatigue.

Some of the use cases Itential envisions include automated port resets, compliance remediation, and automated fixing of service degradation.

Step Two (and Beyond)

The bigger dream, though, is to extend this automation to arbitrary infrastructure gear. As Wade told us in April, "I want Cisco [or any other vendor] to create an agent with an interface/API I can use."

In the weeks since that interview, the discussion of AI agent capabilities has accelerated, driven by adoption of the Model Context Protocol (MCP). It's a step beyond APIs. MCP provides the vocabulary to let agents talk with various data sources and applications, thus avoiding the step of having to configure APIs for every entity involved. If it develops as its backers hope, MCP could become a universal connector; people have taken to calling it a USB-C port for AI.

Yes, security is an issue. Widespread enthusiasm for building MCP clients and servers has mostly been confined to internal use cases, where security concerns are more finite. To create a hyper-automated world of agents gathering data and invoking tools, the industry will need more rigor around MCP authentication, authorization, and identity.

Itential's MCP Server

Having said all that, MCP experimentation is still valuable. Every vendor seems to be building MCP servers, and Itential is no exception. The company made its first announcement of an MCP server at the Network Automation Forum’s AutoCon 3 in Prague.

The MCP server will make it easer for the Itential Platform to work with large language models (LLMs). But specific to the Selector example, it will also ease the way for AI agents to trigger remediation actions. This happens securely, of course; Itential's own guardrails remain in place, along with the customer's defined policy. Itential is also touting further MCP possibilities such as multi-agent collaboration, natural-language infrastructure management, and configuration drift prevention.

Futuriom Take: Agentic AI can enable true automation, although it's going to take teamwork. Itential is taking steps to get there through its Selector partnership and its embrace of MCP.