The Claude Family: Code, Cowork, Design

3 min
Audience
curious developers, beginner to intermediate
Time
15 to 25 min
Module
1/5
Skill goal
pick the right Claude interface for the deliverable in front of you

In one picture

An experienced carpenter does not drive a nail with a sander. Depending on the task, they reach for the saw, the drill or the sander — but it is always the same craftsperson, with the same know-how, simply swapping tools. The Claude family works the same way: Code, Cowork and Design share the same "brain", but each interface offers different "hands", shaped for a specific kind of deliverable. The real professional reflex is therefore not to memorise three separate products, but to recognise, before you start, what done actually looks like for the task in front of you.

How it works

There are three main interfaces in the Claude family, and each one shines on a different kind of work. The diagram below sums them up, and then we detail what each brings to a developer's day-to-day.

  • Claude Code (Claude Code CLI/IDE) lives in your terminal or your editor. It reads your real files, proposes diffs, runs shell commands — it is the companion when the deliverable is executable software: a fixed bug, an added endpoint, a data-import script.
  • Claude Cowork (Cowork workspace) is a shared, persistent workspace. Several people drop in over the weeks, Claude keeps the context of the project, and the space is shaped with plugins tuned to your team.
  • Claude Design (Design canvas) offers a visual canvas to prototype mockups, UI screens or posters. You click and drag instead of writing code.

The right reflex is therefore to qualify the deliverable before opening a tool:

  • "At the end, I have code that passes the tests." → Claude Code
  • "At the end, my team has moved a shared, ongoing project forward." → Claude Cowork
  • "At the end, I have a visual or a UI someone can look at." → Claude Design

Try it now

  1. Take the last task you started in your IDE and ask yourself: is the deliverable runnable code, ongoing collaborative work, or a visual?
  2. Open a terminal in a real project and run claude to start a Claude Code session — even without asking anything, notice where the conversation opens.
  3. Type /help in that session and read the list of available commands: this is your first map of the territory.
  4. Rephrase the same task out loud three ways: "I want code", "I want to collaborate", "I want a visual". Check how the intent changes each time.
  5. Pick one task for the coming week that you will hand to Claude Code rather than to a chatbot in a browser tab.

Key takeaways

  • Same Claude, three "hands": Code for software, Cowork for a shared project, Design for a visual.
  • The choice of tool depends on the deliverable, not on the topic.
  • Picking the wrong tool costs more than not knowing every command of the right one.

Pitfall to avoid

Using Claude Code as a general chatbot ("explain monads to me") and forgetting that it can actually touch your disk — or, the other way around, copy-pasting code into a browser chat when a local claude could have read the file itself. Each interface has a sweet-spot: outside of it, you lose the value that interface brings.

So what? Before you open a tool, say one sentence to yourself: "what does done look like for this task?" — the answer tells you which Claude interface to open.

Getting Started with Claude Code