Claude Code in the Terminal: Setup Guide for Non-Developers

I use Claude Code from the terminal, and I think you should too, even if you've never written code. Install it with one command, open a terminal in a folder, type claude, trust the folder and log in. That's the whole setup. The reason to bother: in the terminal you have direct control over your files, your CLAUDE.md instructions, memory and skills. That makes you faster.

This is the first video in a series on using Claude Code from the terminal. This post covers the install, your first session, and the handful of commands I use every day.

Why the terminal and not the app

There's nothing wrong with the desktop app. You can do about 90% of things from it. But there are things you can't do from a GUI, and sooner or later you end up in the terminal anyway. So I'd rather you get comfortable there from the start.

What the terminal gives you:

• Your file system. You can create folders, touch files and delete things, and see exactly where Claude is working.
• Your configuration in plain files. CLAUDE.md, settings, memory and skills are all files on disk. In the terminal you can open, edit and version them like anything else.
• Several sessions at once. I run Claude Code inside tmux, so I can have multiple sessions side by side and switch between them. That's a topic for its own video.
• Scripting. You can call Claude Code from a script or a scheduled job, which you can't do from a window.

Maybe I feel this way because I'm a developer. But the terminal isn't as hard as it looks, and you only need a few commands.

What you need before you start

• A Mac, Windows or Linux machine. macOS 13+, Windows 10 1809+, or a recent Ubuntu or Debian.
• A paid Claude plan: Pro, Max, Team or Enterprise, or a Console account with API billing. The free plan doesn't include Claude Code.
• A terminal. On Mac, that's Terminal or iTerm. On Windows, PowerShell or Windows Terminal.

Step 1: Install Claude Code

Go to the Claude Code docs, open the setup page and copy the native install command. It's one line.

macOS, Linux, WSL:

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

Windows PowerShell:

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

If you prefer a package manager, brew install --cask claude-code on Mac and winget install Anthropic.ClaudeCode on Windows also work. Those don't auto-update, though. The native install does.

Open a new terminal window and check it worked:

claude --version

If you get "command not found", the install folder isn't on your PATH yet. Close and reopen the terminal first. If that doesn't help, run the troubleshooting steps from the docs. claude doctor prints a health check of your install.

Step 2: Start your first session

Make a folder and start Claude in it:

mkdir test-claude
cd test-claude
claude

Claude Code asks whether you trust this folder. Press Enter to confirm. Now you have a session running in that exact directory. Claude can read and change files in it, and it asks before it does anything risky.

Step 3: Log in

The most important thing after installing: log in with your subscription. Inside the session, type:

/login

It opens your browser. Choose your account, authorize, and go back to the terminal. You're logged in. On a first run, Claude Code usually walks you through this automatically.

Step 4: Give Claude your project's rules with CLAUDE.md

CLAUDE.md is a plain text file that Claude reads at the start of every session in that folder. Put your rules in it: what the project is, how you like things done, what to never touch.

Let Claude write a first version:

/init

It looks at the folder and drafts a CLAUDE.md. Open it, fix what's wrong, and keep it short. There are two levels:

• ./CLAUDE.md in the project folder: rules for this project. Commit it so your team shares it.
• ~/.claude/CLAUDE.md in your home folder: rules for you, in every project.

Example for an agency reporting project:

# Client reporting scripts

- Python 3.12, run with `uv run`.
- Never print or log access tokens.
- Client IDs live in clients.yaml. Don't hard-code them.
- Reports go to ./out/<client>/<yyyy-mm>.csv

Step 5: The commands I use every day

Inside a session:

From the shell:

claude               # start a session in the current folder
claude -c            # continue the most recent session here
claude -p "Summarise every CSV in ./out into one table"   # one-shot, prints and exits
claude update        # update now

claude -p is the one that opens up automation. You can run it from a cron job or a script, and it prints the answer and exits.

Step 6: Five terminal commands you need

If you're new to the terminal, these get you through 90% of it:

pwd              # where am I?
ls               # what's in this folder?
cd my-project    # go into a folder
cd ..            # go up one level
mkdir reports    # make a folder

Everything else, you can ask Claude to do for you.

Step 7: Decide what Claude may do without asking

By default, Claude Code asks before it edits files or runs commands. That's the right default while you learn. Once you trust a workflow, open /permissions and allow the specific commands you approve every time, like running your report script. Keep anything that deletes files, pushes code or touches client ad accounts on "ask". Those rules are saved in a settings file, so they apply to every session in that project.

Common problems and how to fix them

claude: command not found. Open a new terminal window. If it still fails, the install directory (~/.local/bin on Mac and Linux) isn't on your PATH. The troubleshooting page in the docs shows the fix for each shell.

'irm' is not recognized. You're in CMD, not PowerShell. Use the CMD installer from the docs or open PowerShell.

Login loops or fails. Make sure your plan includes Claude Code. If you have an ANTHROPIC_API_KEY set in your shell, Claude Code may offer to use it instead of your subscription. Unset it if that's not what you want.

Claude is working in the wrong folder. It works where you started it. Exit with /exit or Ctrl+C twice, cd to the right folder and start again.

What's next

Once this is working, the next step is skills. They're reusable instruction packs that teach Claude a workflow once, so you don't repeat yourself. I cover the tool I use to manage them in the best way to manage Claude skills.

How Scalvo can help

We use Claude Code to build reporting tools, ad-ops automations and internal apps for marketing agencies. If you want your team set up properly, with shared CLAUDE.md rules, MCP connections to your ad platforms and a few automations running in week one, look at our AI automation sprint or book a call.