Guides

Claude Code for non-developers: how to start

The short answer: You can use Claude Code without knowing how to code. You need a Claude plan, a terminal to paste one install line into, and a folder to work in; then you describe what you want in plain English.

Key points

  • You don't need to know how to code: you describe what you want in plain English and Claude writes the code.
  • You need a paid Claude plan or Console credit, a terminal and a folder to work in.
  • Install with one line from the quickstart, then check it worked with claude --version.
  • Esc stops Claude while it works; /rewind goes back to an earlier point.
On this page
  1. Can I use Claude Code if I can’t code?
  2. What do I need before I start?
  3. What is a terminal?
  4. How do I install Claude Code?
  5. What goes wrong most often?
  6. How do I start my first session?
  7. What should I ask for next?
  8. Will it change things without asking?
  9. How do I stop it or undo a change?
  10. What should my first hour look like?
  11. A short checklist

Can I use Claude Code if I can’t code?

Yes. Anthropic’s own terminal guide says you can use Claude Code even if you have never used a terminal, and that you don’t need to know how to code: you describe what you want in plain English and Claude writes the code (terminal guide).

That does not mean it builds a good app from one sentence. You will still have to decide what the app is for, look at what it made and say what is wrong. This guide gets you to the first working session. Later guides cover planning and checking.

What do I need before I start?

The quickstart lists the basics (quickstart):

  • A terminal (it is already on your computer).
  • A folder to work in. A new empty folder is fine.
  • A Claude subscription (the docs list Pro, Max, Ultra, Team and Enterprise), or a Claude Console account, or access through a supported cloud provider.

So yes, there is a cost: a paid Claude plan or Console credit. Check Anthropic’s current pricing before you decide, because plans change.

Claude Code needs macOS 13.0 or later, or Windows 10 version 1809 or later, according to the terminal guide. It is also not available in every country; the guide links to the list of supported countries.

If the terminal puts you off, the guide mentions a Claude Code desktop app for macOS and Windows that lets you skip the terminal. This guide uses the terminal because it is the version most documentation assumes.

What is a terminal?

It is a window where you type instructions instead of clicking. The terminal guide describes it as a window with a blinking cursor, where you type commands.

  • On a Mac: press Cmd + Space, type Terminal and press Enter.
  • On Windows: press Win + X and choose Windows PowerShell (or Terminal).
  • On Linux: press Ctrl + Alt + T, or search for Terminal.

You can’t click inside it. You move with the arrow keys and press Enter to send.

How do I install Claude Code?

Copy one line, paste it in the terminal and press Enter. On a Mac or Linux the line is:

Terminal
curl -fsSL https://claude.ai/install.sh | bash

On Windows PowerShell it is:

PowerShell
irm https://claude.ai/install.ps1 | iex

Watch out

Always copy the current line from the quickstart, in case it changed.

The install shows no progress while it downloads, so leave the window open. When it finishes it says “Claude Code successfully installed!” Then open a new terminal window and type claude --version. A version number means it worked.

What goes wrong most often?

Install is where most people get stuck, and the docs cover the common cases:

  • “command not found: claude” (Mac, Linux) or “‘claude’ is not recognized” (Windows). The place where claude was installed isn’t on your PATH yet. The terminal guide shows the exact fix for each system, and says to open a new terminal window afterwards.
  • “irm is not recognized” on Windows. You are in CMD, not PowerShell. Open PowerShell instead, or use the CMD line from the quickstart.
  • A web page of code instead of an install. Run the line again. If it keeps happening, the guide suggests installing with Homebrew on a Mac.

Tip

If you hit an error not on this list, paste the error text into a normal Claude chat and ask what it means. That is allowed, and it works.

How do I start my first session?

Make a folder for your project, open the terminal in it (or use cd to go there), and type:

Terminal
claude

The first time, it asks you to log in with your Claude account and opens a browser window for that. After you sign in, you will see a prompt where you can type.

The terminal guide suggests a first message like this:

Ask Claude

make me a simple webpage that says hello world

If Claude asks permission before creating or changing files, you choose Yes. When it is done, double-click the HTML file it made to open it in your browser.

What should I ask for next?

Be specific. The quickstart gives the example: instead of “fix the bug”, say what you see, for example that users see a blank screen after entering wrong credentials. The more you say about what you want and what “done” looks like, the fewer corrections you need (best practices).

Two habits that help from the first day:

  1. Ask it to explain. “What does this project do?” and “explain the folder structure” are good questions the quickstart lists. They teach you what it built.
  2. Not sure what you want yet? The best-practices page suggests letting Claude interview you: tell it the idea, ask it to ask you questions until it could write a spec, and have it save that to a file.

Will it change things without asking?

That depends on the permission mode. The docs describe modes from Manual, where it asks before anything but reading, to ones where it works with fewer questions. Press Shift+Tab to cycle through them. In the quickstart, Anthropic says the mode sets which actions Claude can take without asking you first. Which mode you start in depends on your version and plan, so look at what the screen shows.

If you want to see every step, choose the mode that asks. You can read what it plans to do before you approve it. The security page says you are responsible for reviewing proposed code and commands before approving them.

How do I stop it or undo a change?

  • Press Esc to stop Claude while it is working. Your progress up to that point is kept.
  • Type /rewind (or press Esc twice on an empty prompt) to go back to an earlier point. You can restore the code, the conversation, or both (checkpointing).
  • Know the limit: checkpoints cover files Claude changed with its own editing tools. A file changed only by a command it ran is not restored. The docs say checkpoints work alongside version control, such as Git, and don’t replace it.

You can ask Claude to set up Git for you and to save your work with a message. The quickstart shows prompts like “commit my changes with a descriptive message”.

What should my first hour look like?

  1. Install, log in, make the hello-world page.
  2. Ask Claude what it made and how to open it.
  3. Change one thing: a colour, a word, a button.
  4. Ask it to save your work with Git.
  5. Write down, in one paragraph, what you actually want to build.

That paragraph is the start of everything else. The tool is not the hard part. Knowing what you want, and being able to tell when it is right, is.

A short checklist

  • A paid Claude plan or Console credit.
  • A terminal open in an empty folder.
  • The current install line from the docs, pasted and run.
  • claude --version shows a number.
  • First session started, logged in, first page made.
  • Esc and /rewind known.
  • One paragraph on what you want to build.