Claude Code: Up and Running

Preface

This book is about Claude Code, but it is not a feature catalog. The official docs do that, and they are updated faster than any printed book could be. What you are holding is a book about how the work itself changes when you have an agent in your terminal: what kind of work becomes possible, what kind of judgment it asks of you in return, and why getting good at the loop matters more than memorizing every flag.

Claude Code is the tool we will use. The practice layer around it is agentic development: you let an agent drive the implementation while you drive the direction, success conditions, verification, and judgment.

That distinction matters. Claude Code can write code very quickly. It can also inspect a repository, explain unfamiliar syntax, run tests, update documentation, review a pull request, and call out through integrations to other systems. The danger is that a tool that can do all of that can make every problem look like a prompt away from done. This book takes the opposite view: the more capable the agent becomes, the more important the developer’s judgment becomes.

I do not work for Anthropic, OpenAI, Google, Cursor, or any other AI-tooling vendor. I use these tools because they are useful, not because I have taken a sacred oath over a terminal window. Claude Code is the center of this book because it is the tool I most often reach for right now, and because its terminal-first workflow makes the agentic loop visible enough to teach. Many of the principles ...