A guide for Design + Friends
Not a chatbot, an agent
You already know Claude as a chat window: you type a question, it types an answer. Useful, but it stops at words on a screen. You still do the doing.
Claude Code (and Claude Cowork) is the same Claude, but it can do the work, not just describe it. It still gets things wrong, but we have a great sense of what is good or not. That is our job.
The difference
It tells you how, then stops at words on a screen. You still do the doing.
An agent: it goes and does the job on your real, local files. You meet it as Claude Code, scoped to the folder you pick, or Cowork, which runs longer jobs on its own in the background. Plain English the whole way, no syntax to learn.
Claude has no taste. It produces, you direct. Looking at a draft and saying exactly what is wrong is the muscle you already have. The better a designer you are, the better the work you pull out of it.
You and Claude Code
You open the Code tab and pick the folder for the job. Claude works inside that folder and only reaches elsewhere on your Mac if you allow it.
You write what you want in plain English, the way you would brief a colleague. "Summarise this PDF." "Rename these files." Then hit Enter.
Claude does the work and shows you exactly what it changed. You approve it, or ask for a tweak, and it has another go.
A quick walkthrough
Step through the window below at your own pace. Move with the arrows, the dots, or your keyboard. Each step lights up the part it is describing.
Be specific. "Find the latest sales report in Drive and summarise the key numbers" works far better than "help me with sales."
You are not learning software. You are briefing a coworker who happens to be very fast.
Well beyond a chat window
Tap any one and Claude opens with the question already typed, ready to show you how you'd use it. Don't memorise use cases, ask for them.
The creative-heavy ones lean on tools the team has already wired up. If you are not sure where to start, ask Claude how, or drop a question in #claude-crew.