How to set up Claude Code on Windows

You will configure Claude Code on Windows by installing Node.js, authenticating with your Anthropic account, and launching the agent in your terminal. The process takes about fifteen minutes from start to finish.

Prerequisites

Claude Code by Anthropic is an AI Coding Agent that runs in the terminal, IDE, Slack, or web.

Our read: Do not skip checking your terminal environment before you begin. You need a reliable command line interface on Windows to execute the setup commands smoothly. Take two minutes to open your terminal and verify your user permissions. If you lack admin rights, the package installation will fail halfway through, forcing you to restart the process. Sort out your permissions now so the actual installation flows without interruptions.

• Verify your Windows machine has an active terminal application ready for command line inputs
• Confirm you have administrator access to install software packages on the target operating system
• Keep your Anthropic account login credentials accessible for the upcoming browser authentication step
• Ensure your internet connection remains stable during the command line package downloads

Install and connect

Open your terminal application and run the global installation command for Claude Code. Watch the progress indicators to ensure all dependencies download without throwing errors. Once the package finishes installing, type the startup command to launch the agent. Your default web browser will automatically open a login page. Enter your Anthropic account credentials to authorize the terminal tool. Accept the connection prompt in the browser window to link your account to the local machine.

1. Open your preferred terminal application on Windows
2. Run the command to install Claude Code globally on your system
3. Type the launch command to start the agent for the first time
4. Complete the browser authentication prompt using your Anthropic credentials

Settings that matter and why

Most default settings work fine for standard projects, but you should explicitly verify your working directory boundaries. A common mistake is launching the agent in your root drive, which exposes unnecessary files to the coding assistant. Point the terminal directly at your specific project folder before starting any session. This keeps the token usage focused and prevents the agent from indexing unrelated system directories.

• Review default working directory permissions to ensure the agent only reads relevant project files
• Configure output preferences to keep terminal logs readable during large codebase refactoring tasks
• Check network timeout values if you work behind a corporate proxy or strict firewall

Verify it worked

Do not assume the setup worked just because the installation script finished without crashing. Run a simple test query inside your project directory to see if the agent reads your files. If the tool responds with context about your codebase, your connection is fully operational. If you see authentication errors, close the terminal, clear your browser cookies for the auth domain, and run the connection step again.

• Check that the terminal interface displays the active Claude Code prompt without error codes
• Type a basic diagnostic query to confirm the agent processes inputs correctly
• Verify that responses reference your local project files accurately

How to undo it

If you decide to remove Claude Code from your Windows machine, a standard uninstall command clears the primary binaries. Pay attention to leftover configuration files stored in hidden user directories, as these often persist after a package removal. Delete those folders manually to ensure a completely clean system state. Restart your terminal session afterward to clear any cached environment paths.

• Open your terminal with administrative privileges to modify global packages
• Execute the global uninstallation command for the Claude Code package
• Delete any leftover configuration folders from your user profile directory