Claude Code: Up and Running

Chapter 8. Subagents, Agent Teams, and Plugins

A Note for Early Release Readers

With Early Release ebooks, you get books in their earliest form—the author’s raw and unedited content as they write—so you can take advantage of these technologies long before the official release of these titles.

This will be the 8th chapter of the final book. The GitHub repo is active.

If you’d like to be actively involved in reviewing and commenting on this draft, please reach out to the editor at [e-mail].

Chapters 6 and 7 taught the agent you already have some new tricks. Skills gave it repeatable workflows, and MCP servers gave it live systems to work with. Both of those still assume one worker in one session. Some tasks want more than that. A large review benefits from several distinct perspectives at once, and a single session cannot hold them all without its attention spreading thin.

One of my custom subagents is a Java test quality expert. It reviews a feature branch in the background, uses the SonarCloud MCP server from Chapter 7, and reports whether one person can reasonably maintain the result. I created it through a conversation with Claude Code’s /agents command, which is used both to configure your existing agents and to create new ones. Among the rules I added for the testing agent are “do not modify existing code until the tests have run”. The unexpected part was that I had to tell it to run the tests. You’d think telling it to write the tests would be enough, but that ...