Ahead of KubeCon + CloudNativeCon North America in Salt Lake City, VMblog sat down with Neil Cresswell, CEO and co-founder of Portainer, to talk about one of the biggest challenges facing platform engineers today: how to let AI agents help manage Kubernetes without letting them break things.

Cresswell explains how Portainer has evolved over 10 years from a Docker UI into a multicluster operator control plane that lets teams govern access to, and manage, fleets of Docker and Kubernetes environments from one central place. He describes the company’s approach as the “antidote to complexity” in a cloud native ecosystem he compares to an ultimate Lego shop, where assembling 60 to 70 discrete tooling choices can leave organizations with platforms nobody wants to touch.

In this interview, Cresswell discusses:

• Portainer’s turnkey approach to Kubernetes: Why a single, integrated product can take teams from nothing to running a large fleet of clusters in about an afternoon
• A real-world customer story: An organization running roughly 700 Kubernetes clusters managed through Portainer, after struggling with outdated tooling, CVE exposure, and multiple overlapping GitOps tools
• Portainer Command: An MCP server that acts as a “permission vending machine” for AI agents, issuing temporary, read-only credentials, turning proposed changes into Git pull requests, and requiring human approval before anything is reconciled through GitOps
• Safety controls for CISOs: Visibility into agent conversations, one-click rollback of agent changes, and a lockdown button that instantly cuts off agent access
• New AI add-ons built on Portainer’s API: Portainer Run (a self-service portal for safely deploying vibe-coded apps into isolated sandboxes), AI Grid (pooling idle fleet compute for RAG and OCR workloads), and AI Secure (agentic AI red teaming), all at no additional charge on top of a Portainer license
• A strategic pivot to Kubernetes: Why Cresswell says workloads in the AI era must be assumed hostile, and why Portainer is moving away from Docker to focus on Kubernetes policy controls
• KubeSolo: Portainer’s single-node Kubernetes distribution, optimized for AI inferencing at the far edge
• KubeSchool: The free, plain-language Kubernetes training course Cresswell wrote, which he is presenting through CNCF University at the show

The video also includes a walkthrough of Portainer Command from an administrator’s perspective, presented by Chief Product Officer Nathan Peck, showing per-cluster agent permissions, isolated AI workspaces, pull request-based changes, session timelines, undo and rollback, token usage monitoring, and instant agent shutdown.

Cresswell’s key takeaway: You could build this yourself, but it will probably take a year and a thousand wrong turns. Portainer offers something safe, secure, and ready to turn on tomorrow.

Learn more:

Portainer.io for the container and Kubernetes management platform
Portainer.ai for the new AI products

KubeSchool: https://kubeschool.portainer.io/

Back to: KubeCon + CloudNativeCon 2026