Networking • November 13, 2026 • 7 min read

AI Agents: The Central Control You Need

AI agents are gaining capabilities fast, but without a central control plane, they pose significant risks. Red Hat's OpenShift 5 release offers critical new tools for AI agent management, including an MCP server and gateway.

agent control panel, networked

Deploying AI agents without a central control plane is a dangerous gamble, risking significant harm to your enterprise systems. Red Hat’s recent OpenShift 5 announcement addresses this head-on, delivering new AI agent management tools that provide the necessary guardrails and oversight.

We’re seeing enterprises adopt AI agents across every major platform – AWS, Azure, Google Cloud, Salesforce, SAP, OpenAI, Anthropic. Many are using multiple builders, and the coordination challenge is immense. The failure we see most often is a fragmented approach to agent deployment, where each team spins up agents in isolation, creating security blind spots and operational chaos. This isn’t just about efficiency; it’s about preventing rogue agents from impacting production systems or exposing sensitive data.

Red Hat’s new Model Context Protocol (MCP) server and MCP gateway for OpenShift 5 are designed to act as that critical “air traffic controller.” The MCP gateway, for instance, sits between your AI agents and the tools they interact with, filtering access based on credentials. It integrates Nvidia’s NeMo Guardrails, which are essential for blocking prompt injections and protecting data integrity. More importantly, it can enforce human approval for any system-altering actions an agent proposes. This human-in-the-loop approach is non-negotiable when you’re dealing with autonomous systems capable of making real changes to your infrastructure.

The MCP server for OpenShift ties agent access directly into your existing identity systems and Kubernetes role-based permissions, complete with audit trails. This means you can track exactly what an agent did and when, which is invaluable for compliance and troubleshooting. And for those experimental phases, a new agent sandbox in technology preview allows you to isolate untrusted agent code in virtual machines, preventing potential fallout from untested deployments. We’ve been advocating for this level of isolation for years; it’s a fundamental security practice that applies equally to AI agents as it does to any new application.

Why AI Agent Management Can’t Be an Afterthought

Here’s what nobody is talking about enough: the sheer volume of APIs being exposed to AI agents. Every major platform is pushing API-driven capabilities, and while that’s powerful, it also means a vast attack surface. If an agent with elevated privileges gets compromised or misconfigured, the blast radius is enormous. The Model Context Protocol is table stakes for any modern platform because it provides a standardized way to govern those API interactions. Without it, you’re relying on ad-hoc controls that often fail to scale or provide comprehensive visibility. For a deeper dive into the technical specifications and goals of the Model Context Protocol, refer to the official Red Hat documentation on MCP.

Red Hat isn’t just focusing on OpenShift. They’re extending MCP support to RHEL and Developer Hub 2.1. This is a crucial move. Imagine AI agents diagnosing RHEL issues without unrestricted access, or developers and agents having governed access to your software catalog, operational data, and documentation. This isn’t about letting AI run wild; it’s about giving it surgical access under strict supervision. We at CTS have always emphasized the need for granular access controls, whether it’s for human users or automated systems. The principles remain the same, but the implementation for AI agents is rapidly evolving.

Another significant development in OpenShift 5 is the promotion of OpenShift LightSpeed from a conversational chatbot to a full-fledged AI agent. It can diagnose cluster problems, propose fixes, and assess update risks. In technology preview, it’s even taking on more complex troubleshooting, still with human approval for changes. This evolution from reactive assistants to proactive, governed agents is the direction the industry is heading.

Actionable Steps for Your AI Agent Strategy This Week

If you’re deploying or even just considering AI agents, don’t wait for a crisis. Here are immediate steps:

  • Assess Your Control Plane: Do you have a centralized platform that can manage agents across different hyperscalers and enterprise applications? If not, investigate solutions like Red Hat OpenShift 5’s MCP capabilities or Microsoft Azure Agent 365. You need a single pane of glass, not a dozen disparate dashboards.
  • Implement Guardrails: Integrate prompt injection protection and data sensitivity controls from day one. Nvidia’s NeMo Guardrails is an open-source option worth exploring. The cost of a data breach far outweighs the effort of proactive security.
  • Enforce Human Approval: For any agent action that modifies production systems, require human sign-off. This isn’t about slowing down AI; it’s about preventing unintended consequences and maintaining accountability.
  • Leverage Sandboxing: For any new or untrusted agent code, deploy it within an isolated sandbox environment. This prevents experimental agents from impacting your live infrastructure.
  • Audit & Identity Integration: Ensure every agent’s activity is logged and tied to your existing identity management and role-based access control (RBAC) systems. If you can’t audit it, you can’t trust it.

The race for AI agent control is just beginning, and the stakes are high. Get ahead of it. Need help planning your secure AI deployment? Explore our AI solutions.

Frequently asked questions

What is AI agent management?

AI agent management refers to the tools and processes used to secure, monitor, and control the behavior of autonomous AI agents within an enterprise environment, ensuring they operate within defined parameters and don't pose risks.

Why is a central control plane important for AI agents?

A central control plane is crucial for coordinating AI agents built on different platforms, enforcing consistent security policies, managing access permissions, and providing a unified audit trail to prevent fragmentation and potential harm.

What is the Model Context Protocol (MCP)?

MCP, or Model Context Protocol, is an open standard that enables AI agents to connect securely to data, tools, and other systems via APIs, providing a governed framework for their interactions with enterprise resources.

Can AI agents make changes to my IT systems?

Yes, AI agents are designed to take actions, including diagnosing issues, proposing fixes, and triggering workflows. Robust management tools like Red Hat's MCP gateway allow for human approval before agents make system-altering changes.

Related reading

Ready to upgrade your technology?

Complete Tech Solutions designs, installs, and supports IT, cabling, security, and network infrastructure for businesses across Grand Rapids, West Michigan, and nationwide. Schedule a free site assessment and we’ll map out the right solution for your space and budget.

Learn more about our Consulting services.

Ryan Whitaker

Complete Tech Solutions

Back to Blog

Get the Latest Tech News Delivered

Weekly curated tech news, industry trends, cybersecurity updates, and AI insights — straight to your inbox. No spam, unsubscribe anytime.

Join 500+ IT professionals. Powered by the latest industry RSS feeds and AI-curated content.