So Mike, How Was AGNTCon + MCPCon?

This week was Agentic AI Foundations (AAIF) AGNTCon + MCPCon, and it was an absolutely fun time. The people, the presentations, the keynotes, and the conversations were all a major highlight and a good indication on where things are going in the agentic era.

In this quick blog post, I talk about the highlights, conversations, and what can be expected when you go to AGNTCon + MCPCon.

What Was the Most Asked Question?

One of the reasons I really enjoy working the booth is because of the conversations I have. The majority of the convos that I have at conferences when working the booth and just walking around/chatting with people in general are all fun and deep engineering conversations. AGNTCon + MCPCon did not disapoint in that regard.

The primary questions I was asked about were around:

1. Agentic sandboxing (I talked a lot about kagent and Agent Substrate).
2. Local/open-weight Models and the best way to run them.
3. How to secure Agents in terms of what they can and can't interact with (MCP Servers, LLMs, other Agents, etc.)
4. Securing people in the world of agentic and ultimately understanding how to ensure humans using Agents can stay safe.

What I also noticed was the number of students, newcomers, and folks who are just starting down the AI journey path. This was really cool because ultimately, the majority of convos at conferences like Kubecon are people in the cloud-native and agentic world, which makes total sense. What AGNTCon + MCPCon was able to do is open up the doors for everyone else. For example, I had several conversations with hardware networking folks that have been Network Admins/Engineers for years and are now interested in the agentic space. They see that AI is the way the world is going and they want to be prepared. It's a really cool experience to be able to help out.

What Was the Theme Of the Presentations?

If I were to categorize the talks, I would think of the following breakdown for what the mantra of the sessions were:

1. Agentic infrastructure
2. AI security & governance
3. Agent performance and quality output

And ironically enough, these are the topics that even the folks who aren't Platform or DevOps Engineers are asking. If someone is a Data Scientist or in MLOps, they're even asking questions about these topics to get a better understanding of what needs to be implemented within the environment.

Maybe they don't care about k8s, but they care about "how the thing will be secure and optimized" without wanting to know the underlying system.

AI In Production as A Whole

The question "So, what do people care about for AI in production based on your conversations at AGNTCon + MCPCon" has a few answers, but there isn't just one. There's such a wide range of questions.

Some people are thinking "we have Agents running and we need to scale this out to thousands of machines". Other people are thinking "Efficiency, security, and performance is great, but geez, I just need to get an Agent working first before I think about all of that".

The biggest "theme" I saw in the convos I had at the level 300-400 level were around security, efficiency, and scale.

I, of course, brought up a ton around how kagent and Agent Substrate working together is solving that exact problem, and from a security perspective on your proxy (how an Agent is routing to an LLM and MCP Server), the convo went in the agentgateway direction.

When thinking about "AI in Production", this is where we are right now:

• Sandboxing (isolation layer)
• Efficiency and performance
• Securing point A (an Agent) to point B (it's destination)
• Scale (multi-cluster, scaling out Agents, where memory/context/sessions are stored, etc.)

Wrapping Up

I really enjoyed the conference as a whole. If I heard the numbers correctly, I believe there were about 4,000 or 5,000 people. It's really cool because it reminded me of the early days of cloud-native and Kubernetes. People coming together at the early days of Kubecon with specific questions and engineering needs to solve a particular problem. People in the early days of k8s and cloud-native knew and saw the way the tech world was heading, and they wanted to get up to speed. I received similar "vibes" from AGNTCon + MCPCon, which was absolutely amazing.