Claude Code in Practice: From Your First Session to Own AI Agents

There is a tool that will work through a folder of four hundred files while you are at lunch, and most people never open it because of the second word in its name.Claude Code was built for software engineers. Then Anthropic measured what people actually did with it - about 400,000 sessions from about 235,000 users - and found analysts, accountants, lawyers and managers analyzing data and producing documents. The finding worth repeating: domain expertise mattered more than a coding background.So this is two books in one cover, separated deliberately.PART ONE is for people who do not write code, and uses no programming vocabulary at all. It installs the desktop app, walks a first session end to end, then does the work: handing over a whole folder instead of a file, bulk renaming and converting, reconciling twelve monthly exports that should match and do not, pulling structured data out of two hundred contracts, and writing down how you work so next month takes a sentence.PART TWO is for people who do. Orienting an agent in a codebase you did not write, plan mode and what skipping it costs, the configuration layers a team actually needs, and hooks - because instructions guide and only hooks enforce. Then underneath the product entirely: the Messages API and a tool-use loop written by hand, the Agent SDK that is Claude Code as a library, and the decisions that make an unattended agent affordable rather than alarming.Over 360 figures show the screen at every step, with the thing you need circled.Every command is taken from official documentation. Beta features are labeled. Prices carry a date.