Skip to main content
🎤 Luca Berton is speaking at Red Hat Summit & KubeCon EU 2026!Learn more →
intermediate

kagent 1.0: Run AI Agents on Kubernetes

Install kagent 1.0, build a Kubernetes-native AI agent, give it MCP tools, contain it with egress policies and trace everything it does.

7 lessons40mUpdated November 2026

What you'll learn

  • Install kagent 1.0 and Agent Substrate on a local Kubernetes cluster
  • Define an agent with a Harness, an AgentTemplate and an Agent, and talk to it through a Session
  • Bind MCP tools to an agent and read what the model can call
  • Control what an agent can reach with egress policies
  • Trace agent requests with OpenTelemetry
  • Suspend, resume and debug agents when something goes wrong

Who this is for

  • • DevOps engineers deepening their expertise
  • • SREs building production-grade systems
  • • Backend developers expanding into infrastructure

Prerequisites

  • • Comfortable with kubectl and YAML
  • • Docker or Podman installed
  • • An API key for a model provider (OpenAI or Anthropic)

Course content

2 modules, 7 lessons, 40m of video

Module 1

3 lessons, 17 min
1

6 min

Create a kind cluster on Kubernetes 1.37, install the kagent CLI and the kagent 1.0 release candidate, check every component that the installer creates, and look at the result in the kagent web interface.

2
Your first agent

6 min

Apply a Harness, a ModelConfig, an AgentTemplate and an Agent, wait for the Agent to become ready, then open a Session and talk to it in the terminal and in the web interface. The video points the agent at an OpenAI-compatible endpoint with my-model-config. If you installed kagent with an OpenAI key, the installer already created default-model-config: name that one in the template instead.

3
Give your agent tools with MCP

6 min

List the tools of the built-in MCP server, bind two of them to your AgentTemplate and watch the agent answer from live cluster data, in the terminal and in the web interface, where each tool call and its result is a card in the chat.

Module 2

4 lessons, 23 min
4
Keep agents in a box with egress policies

5 min

See what an agent may reach by default, grant one destination with an EgressPolicy, and check that kagent admitted it.

5
Trace your agents with OpenTelemetry

7 min

Deploy the lightweight OTel stack, turn on tracing in kagent and read the spans of one agent request.

6
Suspend, resume and debug

6 min

Understand what happens to an agent between turns, see the actors on the Substrate page of the web interface, read the Agent conditions to find why one is not ready, and plan upgrades.

7
Run your kagent agents on a local model with Ollama

7 min

The learner configures kagent to use a local model served by Ollama on a Mac, demonstrating how to connect the cluster to the local service. They learn to switch an agent's model configuration to use the local Ollama model, allowing the agent to run with no per-token cost.

Get started

Start Free Lesson

First lesson is always free — no signup required

This course includes

7 hands-on lessons
Intermediate level
40m total
Final assessment

Want all courses?

See Pro plan →

Your instructor

Luca Berton

Founder and CEO, Open Empower. About the author

More in AI Agents

All courses in this track