Managing Infrastructure with AI Agents
AI agents that detect a malfunction in a Kubernetes cluster, identify the cause in the system logs, and propose a fix on their own: SUSE is integrating this capability across its entire portfolio. In this TechTalk, Hugo de Vries of SUSE demonstrates how the Model Context Protocol (MCP) connects AI agents to Linux servers and Kubernetes clusters—and how to set limits.
How does agentic AI work for infrastructure management?
An AI agent does not provide advice; it performs actions. To do so, it needs access to your systems. MCP is the open standard that governs that access: an MCP server provides the agent with a defined set of actions, such as retrieving pending patches or restarting a service. SUSE has integrated MCP into SUSE Linux Enterprise Server 16, SUSE Multi-Linux Manager, SUSE Rancher Prime, and SUSE AI. You choose which language model to use.
What is a police officer allowed to do?
Technically, it works. The difficult choice lies in setting the limits. An agent that detects a critical vulnerability on five servers and asks for permission to apply a patch is different from an agent that restarts immediately. Hugo discusses where to build in approval workflows, how to limit agent costs, and which components are still in technical preview.
Meet our speaker
Hugo de Vries
Solution Architect, SUSE
Hugo is a Solution Architect at SUSE and advises organizations in the Benelux on the entire SUSE portfolio, from SUSE Linux to Rancher Prime. Prior to that, he worked for six years at StackState, the observability platform that became part of SUSE in 2024. As a Solutions Engineer and later as a Customer Success Lead, he helped enterprises trace outages back to the changes that caused them.
What We Treat
- How MCP Servers Grant AI Agents Access to Linux Servers and Kubernetes Clusters
- Identify an issue, correlate it with system logs, and resolve it with a pull request
- Approval Process: Which Actions an Officer Performs Themselves and Which They Do Not
- Choose your own language model, without being locked into a single AI provider
- Limiting the Costs of AI Agents
- What's already available and what's still in technical preview
Sign up for this TechTalk at the SUE office, located at De Ooyen 9 in Geldermalsen. Afterward, there will be dinner and drinks so we can continue the conversation.
