Technology · Ebook

kagent: Cloud-Native Agentic AI for Kubernetes

by Shriira Press

kagent is a CNCF sandbox project, created by Solo.io, that brings AI agents into Kubernetes as first-class citizens. Instead of running an agent as a laptop script, you declare it the way you declare a Deployment: a manifest, version-controlled in Git, reviewed in a pull request, and reconciled by a controller. Each agent is backed by a language model, connected to tools through the Model Context Protocol, and able to investigate incidents and operate the cluster. This book walks the whole system: the controller and engine, the Agent, ModelConfig, and ToolServer resources, MCP tools, multi-agent A2A workflows, observability, security, human-in-the-loop gating, and how to put it all into practice on a real cluster.

Contents

1. 1Preface
2. 2Chapter 1 — Agents That Live in the Cluster
3. 3Chapter 2 — Architecture
4. 4Chapter 3 — The Agent Resource
5. 5Chapter 4 — Models and ModelConfig
6. 6Chapter 5 — Tools and the Model Context Protocol
7. 7Chapter 6 — Multi-Agent Workflows and A2A
8. 8Chapter 7 — Observability, Security, and Human-in-the-Loop
9. 9Chapter 8 — kagent in Practice