Getting started with Claude Code

Claude Code is Anthropic's coding assistant for the terminal. It reads the files in your project, writes and changes code, runs commands and handles git, all from requests written in plain language. It also comes as a web version (claude.ai/code), a desktop app, extensions for VS Code and JetBrains, and inside Slack. This guide covers the terminal, which is the most complete way to use it.

What you need first

1. A paid Claude plan: Pro, Max, Team or Enterprise. The free plan does not include Claude Code. The alternative is a Claude Console account, which is API access with prepaid credits.
2. A computer running macOS 13 or later, Windows 10 (version 1809) or later, or Linux (Ubuntu 20.04+, Debian 10+, Alpine 3.19+), with 4 GB of RAM and an x64 or ARM64 processor.
3. An internet connection and a terminal window.

Step 1: install

On macOS, Linux or WSL, open a terminal and paste:

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

On Windows, open PowerShell and paste:

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

If you use the Windows Command Prompt (CMD) rather than PowerShell, the command is different:

> curl -fsSL https://claude.ai/install.cmd -o install.cmd && install.cmd && del install.cmd

On Windows, install Git for Windows as well. It is optional, but without it Claude runs commands through PowerShell instead of Bash.

If you prefer a package manager:

$ brew install --cask claude-code        # Homebrew, macOS
$ winget install Anthropic.ClaudeCode    # WinGet, Windows

To confirm it worked:

$ claude --version

You should see a version number followed by "(Claude Code)". If something looks wrong, claude doctor checks the installation without starting a session.

The native install updates itself in the background. Homebrew and WinGet do not: run brew upgrade claude-code or winget upgrade Anthropic.ClaudeCode from time to time. On Homebrew, the claude-code package follows the stable channel, which runs about a week behind and skips releases with serious regressions; install claude-code@latest if you want every release the day it ships. Debian, Ubuntu, Fedora, RHEL and Alpine also have official apt, dnf and apk repositories (see the links at the end).

Step 2: sign in

Go to your project folder and start Claude Code:

$ cd /path/to/your/project
$ claude

The first time, your browser opens so you can sign in. After that your credentials are stored. To switch accounts later, type /login inside the session.

Step 3: your first conversation

Start by letting Claude get to know the project:

> what does this project do?
> explain the folder structure
> where is the main entry point?

Then ask for a small change:

> add a hello world function to the main file

Claude finds the file and shows you the change. If it asks before saving, choose Yes.

Who approves what it does

This matters before you go further. On Pro, Max and Team plans, sessions start in auto mode. In auto mode you do not approve each step: a second model reviews the actions, and Claude changes most files and runs most commands without asking you. Some things stay blocked by default, such as deploying to production or force-pushing in git, but everything else goes ahead.

On other plans, and sometimes in the very first session after installing, it starts in Manual mode, which asks for your permission before it changes files or runs commands.

Press Shift+Tab at any time to switch modes. If you are new to this and would rather see every change before it happens, choose Manual mode. To start that way every time, add this to ~/.claude/settings.json:

{
  "permissions": {
    "defaultMode": "manual"
  }
}

Claude Code may ask you once whether you want to switch to auto mode. If you say no, your setting stays.

Everyday commands

In your terminal:

• claude starts a session
• claude "task" starts a session with a first request already typed
• claude -c continues the most recent conversation in this folder
• claude -r lets you pick an earlier conversation to resume
• claude -p "question" answers once and exits, without opening a session

Inside a session:

• /help lists the commands
• /init creates a CLAUDE.md file with notes about the project, which Claude reads in every session
• /model chooses the model
• /clear clears the conversation
• /exit (or Ctrl+D twice) leaves

Type / to see every command available to you. Tab completes, and the up arrow brings back earlier requests.

Five habits that pay off

1. Be specific. Instead of "fix the bug", say "fix the bug where login shows a blank screen after a wrong password".
2. Break big tasks into numbered steps.
3. Let it explore first ("analyze the database schema") before you ask for changes.
4. Use git. You see exactly what changed and can go back. In auto mode this matters even more, because many changes happen without asking you. Claude will make commits if you ask it to.
5. Create CLAUDE.md with /init and keep adding your project's rules to it. It saves you repeating yourself every session. The CLAUDE.md guide explains how.

Sources

• Quickstart, official documentation, read on 16 September 2026.
• Advanced setup: requirements, installers, apt, dnf and apk, release channels.
• Choose a permission mode: auto mode, Manual mode, defaultMode.
• Troubleshoot installation