Claude Code in Practice · Module 01
Getting started – From chat box to teammate
How we went from copy-pasting answers out of a chat window to letting Claude Code work inside the DevOps Peach repository: the change of role, a clean setup on a Windows laptop, a CLAUDE.md that grew out of real mistakes, and the loop we use to build a feature.
What you will learn
- Explain the difference between a chat assistant and an agent that edits files and runs commands
- Install Claude Code on Windows and fix the PATH and execution-policy errors most people hit
- Write a short CLAUDE.md from real review comments and keep it current
- Build a feature with the plan → build → look → refine loop instead of one giant prompt
Lessons
- From answering to doing: what changes with Claude Code
A chat assistant gives advice you carry out yourself. Claude Code opens the files, runs the commands and shows you the result, so your job moves from typing code to giving clear work and checking it.
- A clean setup on a Windows laptop
Install Node.js and Claude Code, get past the two errors almost everyone hits on Windows, choose an IDE, and learn the three things on the first screen that matter: the working folder, the permission mode and /context.
- CLAUDE.md: the rules we learned the hard way
Claude starts every session knowing nothing about your project. Our CLAUDE.md is 74 lines, and almost every line is there because something went wrong once.
- Building a feature: plan, build, look, refine
The loop we use for every visible change on DevOps Peach: options or a plan first, a build in small steps, a screenshot to look at, then one round of focused corrections.