Extreme Agent ONE™ Coworker Is Now Generally Available

772299-Extreme-Agent-ONE-Coworker-Blog-Images_v1_960x300.jpg

We made you a promise. Today we’re keeping it.

At Extreme Connect 2026 this past May, I made you three promises about Extreme Agent ONE™. Let me start with the third, because it’s the one that is hardest to make and easiest to break… a release date.

This industry has a habit of announcing a vision from a conference stage and delivering it two years later, if at all. We promised we would not do that, and here we are. Extreme Agent ONE Coworker is live, across our entire global Extreme Platform ONE installed base, already working alongside our customers to run their networks. No beta, no early-access list, nothing to request, simply delivered as promised.

For those of you unfamiliar with Extreme Agent ONE, it is not AI bolted onto software that was designed without it. Extreme Agent ONE sits on top of a whole stack, and every layer of it is purpose-built and already running in production.

The colleague we all want

Extreme Agent ONE Coworker is the colleague we all want. Someone who works alongside us, flags the things that matter, catches problems while they are still small, and knows which decisions are theirs to make and which ones need our sign-off.

That framing changes who carries the work. A tool waits for you to know when and how to use it, but a colleague already knows what matters on your network, notices when you are stuck, and when needed, speaks up. We call this the Nudge.

When a critical issue arises at 2 a.m., Coworker is already digging in before the page ever reaches your phone, so you can wake to a set of findings instead of an alarm. That's the difference between escalation and action, enabling teams using Extreme Agent ONE Coworker to resolve issues up to 15x faster.

Our AI stack

There is no magic here, only architecture: four layers, each one depending on the one beneath it.

  • Reasoning. The foundation is frontier AI models from the leading cloud providers and AI labs. As those models improve, everything built on top of them improves too, without a migration project. No one waits two years for the latest model to reach the product.
  • Context. Reasoning without context produces generic answers. We spent years building AI Core, the intelligence layer that knows your specific network. It does not just take in data from the network; it interprets it. Every entity in your Extreme Networks environment is defined, understood, and connected through a knowledge graph that maps the relationships: who is connected to what, what policies govern whom, and how every change cascades through your environment. Anyone can license the models. The context is the part that has to be earned.
  • Procedural knowledge. The Skills layer is where the stack becomes specific. This is where we encode the actual SOPs and the actual patterns that separate a well-run network from a poorly run one. A frontier model can talk through a troubleshooting problem in the abstract all day. What it cannot know is your maintenance window, your guardrails, and the operational limits your team works within. The Skills layer supplies that knowledge in a form the agents can execute, rather than a runbook somebody has to read first.
  • Execution. The agentic layer is where it all comes to life. Multiple specialized agents work in parallel, decomposing tasks, executing within their domains, and validating the results, all orchestrated by a supervisor agent and wrapped by a harness. The harness runs the full agent lifecycle, holds governance in place at every step, retains persistent memory, and leaves behind an auditable record of what was done and why. Strip out the harness and you are left with a convincing demo. Keep it, and you have something that can survive an audit.

The skills your team needs

Extreme Agent ONE Coworker includes six incredible skills. Four are available now and are faster and smarter than the versions you have been using:

  • Talk to Data—Ask your network a question in plain English and get the answer back from live telemetry, with no query syntax and no dashboard to hunt through first.
  • Talk to Knowledge—GTAC articles, product documentation, and your specific CVE exposure in a single conversation, with the source cited. It slashes the time your team spends onboarding new engineers by up to 50%.
  • Talk to Support—Uses live network context to recommend fixes and, when needed, automatically escalates to our GTAC team with all the evidence Extreme Agent ONE Coworker has already gathered. Track your cases without leaving Extreme Agent ONE.
  • Canvas—Customizable visualizations like tables, charts, and graphs to monitor your network, your way.

Two are brand new:

  • Nudge—That 2 a.m. critical issue arrives as a “nudge” in the command bar, and because Extreme Agent ONE Coworker has already started investigating, the Nudge carries findings rather than just an alarm. Informational nudges work the same way at lower volume: relevant to the page you are on, gentle, and easy to dismiss.
  • Talk to RRM—24/7 Skill to continuously tune Wi-Fi radio settings (channel and power) so the network automatically adapts to interference, load, and client behavior for better performance.

Our other two promises

The third promise is delivered. The first two hold exactly as they did the day we made them.

The technology your network demands. Frontier-model reasoning, context drawn from across your environment, procedural knowledge built over decades of enterprise networking, and execution that runs at machine speed around the clock. None of that complexity should ever become your problem. What you should notice is a network that finally moves as fast as the AI it carries.

The relationship. We built the entire stack for your network. But we built Extreme Agent ONE for you! The network engineer who hasn't had an uninterrupted Monday in years. The operations leader holding together a team that is stretched too thin. The practitioner who will always know their environment better than any AI can. You deserve technology that adapts to how you work, not technology that forces you to work around it.

What comes next

Extreme Networks has never been comfortable waiting to see what the market does first. Extreme Agent ONE is the bet we are making on how AI helps you run your networks, minimizing your effort and maximizing your value.

Our customers are using Extreme Agent ONE Coworker right now, and they’re already telling us what they want us to add next. That feedback is invaluable, and it is shaping the roadmap in real time. We’re going to keep moving quickly, so stay tuned. Explore Extreme Agent ONE Coworker today.

About the Author
Nabil Bukhari headshot 2.png
Nabil Bukhari
President of AI Platforms and EVP, Chief Technology Officer

Nabil serves as the Chief Product and Technology Officer, and GM of Extreme’s subscription business. In his role, Nabil leads the vision, strategy, execution and delivery of Extreme’s suite of differentiated products and technologies, driving an ever-improving experience for customers.

Full Bio