Skip to content
FREE ACCESS
Verified 30 Sep 2026
Silent Overseer
ClaudeGuide

How to Set Up Claude Code (the Claude CLI) on Your Computer: Step by Step

Published 30 Sep 2026 Updated 30 Sep 2026 Last verified 30 Sep 2026 12 min read
The short version

Claude Code is Claude working on your computer, in your own folders. Instead of copying and pasting between a chat window and your files, you tell it what you want and it reads, edits and creates files for you, asking permission along the way. Install takes one command; you need a paid Claude plan or Console account. It started as a tool for programmers but is useful to anyone who works with files.

What is Claude Code (and what is “the CLI”)?

“CLI” means command-line interface: a text window (the terminal) where you type instructions. Claude Code’s terminal version is the full-featured one. You open a terminal in a folder, type claude, and chat with Claude in plain English while it works on the files in that folder. Anthropic describes it as an agentic coding tool that reads your codebase, edits files, runs commands and connects to your other tools. It also runs in VS Code, JetBrains, a desktop app and the browser.

Plain-English jargon

Terminal: the text window where you type commands. Folder / directory: where your files live. Agent: an AI that takes actions, not just chat. Claude Code is an agent, so read the safety section below. More terms in our glossary.

Why use it? The benefits

  • No copy-paste. It reads your files directly and saves its work back to them.
  • It does multi-step jobs. “Find every document mentioning X, summarise each, and put the results in a spreadsheet.”
  • It checks its own work. It can run things and read the results, then fix what went wrong.
  • It remembers your rules. A small file called CLAUDE.md in a project holds your standing instructions, which Claude Code reads at the start of every session.
  • It’s scriptable. Advanced users can pipe text into it, schedule recurring tasks and run it in automation.
  • You stay in control. By default it asks before changing things.

What you need first

  • A computer: macOS 13+, Windows 10 (1809+) or 11, or a recent Linux (Ubuntu 20.04+, Debian 10+), with 4 GB+ RAM and an internet connection.
  • A paid account. Anthropic’s docs say Claude Code needs a Pro, Max, Team or Enterprise plan or a Console account. The free claude.ai plan does not include Claude Code. It can also run through third-party providers such as Amazon Bedrock or Google Cloud. Check Anthropic’s pricing page for current plans.
  • Nothing else. You don’t need to know how to code. Git for Windows is optional.
Heads up on cost

Because it does real work, it can use your plan’s allowance faster than ordinary chatting. Start with small tasks and see how it goes before committing to anything bigger.

Step-by-step install

Step 1: Open a terminal

  • Windows: press the Windows key, type PowerShell, and open Windows PowerShell (or Terminal). Your prompt starts with PS C:.
  • Mac: press Cmd + Space, type Terminal, press Enter.
  • Linux / WSL: open your terminal app.

Step 2: Run the install command

These are the recommended “native install” commands from Anthropic’s documentation. Copy the one for your system and press Enter:

Windows (PowerShell)
irm https://claude.ai/install.ps1 | iex
Mac, Linux or WSL
curl -fsSL https://claude.ai/install.sh | bash

Prefer a package manager? On Mac, brew install --cask claude-code. On Windows, winget install Anthropic.ClaudeCode. (Those two don’t auto-update.) Windows users on classic Command Prompt (CMD) use a different command, listed in Anthropic’s setup page.

A sensible habit

Only run install commands copied from the official source. This one comes from Anthropic’s own documentation at code.claude.com. Never run one from a random video comment or forum post.

Step 3: Check it worked

Close the terminal, open a new one, and run:

Any system
claude --version

You should see a version number. If it says “command not found” or “not recognized”, the install folder isn’t on your PATH yet: see Anthropic’s install troubleshooting. For a health check, run claude doctor.

Step 4: Go to a folder and start it

Make a practice folder first, for example one called ai-practice containing a few copies of files you don’t mind experimenting with. Then in the terminal:

Example (Windows or Mac)
cd path/to/ai-practice
claude

Step 5: Log in

The first time, Claude Code opens your browser so you can log in with your Claude account. Follow the prompts. (If you use an API key instead, it asks you to approve it once.) That’s it: you’re in.

Step 6: Give it a first task

A safe first prompt
Look at the files in this folder and tell me what's in them. Don't change anything yet.

Then try something small, like “create a file called notes.md with a summary of these documents.” Claude Code shows you what it wants to do and asks permission first.

Step 7: Useful commands

  • Type /help inside Claude Code to see what’s available.
  • claude update updates it (native installs also update automatically).
  • Ctrl+C stops what it’s doing. To leave, type /exit.

Prefer clicking to typing? The Claude desktop app has a Code tab that includes Claude Code, with no separate CLI install. A paid plan is still required.

Stay safe: the golden rules

  1. Practise in a copy. Work on copies of files until you trust the workflow.
  2. Read before you approve. It asks permission to edit files or run commands. Read what it’s about to do. Don’t just press Enter.
  3. Keep backups. Use cloud sync or version history for anything important.
  4. Keep secrets out. Don’t point it at folders full of passwords, bank details or other people’s private data. See AI privacy basics.
  5. Verify results. Like all AI, it can be confidently wrong. See how to check an AI answer.

What can you do with it? Examples from different walks of life

Anthropic’s own description focuses on coding, but the underlying skill is “work with the files on my computer using plain English”. These are illustrative ideas for the kinds of jobs it suits; results will depend on your files and your review.

Who you are Things you could ask it
Developer “Write tests for the login module, run them, and fix any failures.” “Explain how this project is organised.” “Commit my changes with a clear message.” (These match Anthropic’s documented uses: features, bug fixes, tests, git commits and pull requests.)
Small-business owner “Turn this folder of invoices into a summary of who owes what.” “Draft consistent product descriptions from this spreadsheet.” “Tidy and rename my customer files by date.”
Freelancer / creative “Make a simple portfolio website from the project descriptions in this folder.” “Write a proposal from my notes.” “Create a template contract checklist.” (Have a professional review legal documents.)
Student “Summarise each lecture-note file and build a revision timetable.” “Turn my essay outline into a reading list checklist.” Use it to understand and organise, not to submit work that isn’t yours.
Teacher “Create differentiated worksheets from this lesson plan.” “Sort these resources into a folder per topic.” “Make a marking checklist from this rubric.”
Parent / household “Sort our scanned school letters by child and date and list upcoming deadlines.” “Build a weekly meal plan and shopping list from these recipes.” Avoid giving it sensitive documents.
Job seeker “Tailor my CV to each job advert in this folder, without inventing anything.” “Track my applications in a spreadsheet.”
Writer / blogger “Check all my drafts for consistent style.” “Build an index of my posts.” “Turn my notes into a chapter outline.”
Researcher / analyst “Read these PDFs and extract the key figures into a table with page references.” “Clean this messy CSV and tell me what you changed.”
Hobbyist / tinkerer “Make a small game or a simple app.” “Automate renaming my photo files.” “Explain this error and fix it.”

These are examples of tasks people commonly hand to file-aware AI assistants, not guarantees. Start small and check the output.

Common mistakes

  • Trying to use it on the free Claude plan. It isn’t included.
  • Running it in a folder full of important, un-backed-up files on day one.
  • Approving every prompt without reading.
  • Asking for a huge vague job. Break it into steps.
  • Copying install commands from unofficial sources.

Limitations

  • It needs an internet connection and a paid account or approved provider.
  • Like every AI, it can misunderstand or make mistakes, so review its work.
  • The terminal isn’t for everyone. If you’d rather click, use the desktop app’s Code tab.

What to try next

Sources

Silent Overseer uses AI tools to help research and draft, and every article is reviewed before publishing. How we work. Spotted a mistake? We correct and date changes.

The Overseer Brief

Important AI developments, practical tools and useful freebies without the noise.

One useful email at a time, unsubscribe any time. We store your email to send the Brief; see the privacy policy.