September 2026
Intermediate
322 pages
7h 22m
English
A shell script can start an agent process. That does not make the agent operable. Platform teams need to know which model it uses, which tools it can call, which identity it runs as, what changed between revisions, whether it is ready, and where its traces went. kagent brings those concerns into Kubernetes by representing an agent, its model configuration, and remote MCP servers as declarative resources that controllers can reconcile.
In this chapter, we will install and inspect the kagent APIs, declare a platform troubleshooting agent, and connect its ModelConfig to the in-cluster OpenAI-compatible route. We will expose a controlled MCP tool set through a RemoteMCPServer and bind only named tools to the ...
Read now
Unlock full access