Setting up a local AI coding assistant transforms how you approach software development. By running directly on your machine, this tool reads your local project files and executes commands autonomously. Before you begin the Claude code installation process, you need to ensure your system meets the basic requirements.
Users must have an active paid subscription to Anthropic’s services. The tool does not currently operate on the free tier. Therefore, you will need a Pro or Max plan to access the coding interface. If you are unsure which tier fits your workflow, you can review our breakdown of Claude Code pricing and subscription plans.
You also need a compatible operating system. The software supports macOS, Linux, and Windows. In addition, certain terminal environments require standard developer tools. For example, Windows users will need a compatible bash environment to run the setup scripts correctly.
How to Install Claude Code on Windows
Windows systems require a few specific steps to get the command-line interface running smoothly. The native installation relies on specific bash commands. Therefore, your first step is ensuring you have the right terminal tools downloaded.
First, you need to download Git Bash. You can find the installer on the official Git for Windows website. Download the standard 64-bit installer and open the executable file. You can leave all the default options checked during the setup wizard. Once the installation finishes, you will have the necessary environment to proceed.
Next, you will run the primary installation script. Open your Windows PowerShell application. You can find this by typing “PowerShell” into your Windows search bar. Paste the official installation command provided in the Anthropic documentation and press Enter. The system will download the required packages directly to your local drive.
Finally, you must add the application to your system’s PATH. The PowerShell window will display a specific file path upon completion. Copy that folder location. Open your Windows “System Properties” and navigate to “Environment Variables”. Select the “Path” variable, click edit, and paste your copied location into a new line. Save your changes and restart your terminal. You can now type claude into your command prompt to launch the tool.
Quick recap: Windows users must install Git Bash first, run the designated PowerShell script, and update their system environment variables to complete the setup.
How to Install Claude Code on Mac
Mac users generally experience a faster setup process because macOS includes native bash and zsh terminal support. You do not need to install third-party terminal emulators to begin.
First, open your Mac Terminal using the Spotlight search tool. Paste the official curl command provided by Anthropic into the window and hit enter. The script will automatically fetch the software and place it in your local directory.
After the download finishes, the terminal will instruct you to update your shell configuration. You will need to copy the provided export command and run it. This action adds the application to your ~/.zshrc file. Consequently, your Mac will recognise the software commands globally.
Restart your terminal window to refresh the environment variables. Type claude and press enter. The interface will prompt you to authenticate your account via a web browser. Once authorised, your setup is finished.
Using the Claude Code Desktop App
Many beginners find terminal environments intimidating. Fortunately, Anthropic recently released a dedicated desktop application that removes the need for command-line navigation.
You can download the desktop client directly from the official website. Once installed, log in to your account. You will notice three distinct icons at the top of the sidebar. The first is for standard web chats. The second is for automated cloud tasks. The third icon, resembling code brackets, is your dedicated local coding workspace.
Click that third icon to begin. The interface allows you to select a local folder on your computer visually. Any tasks you request will only affect the files inside that specific folder. This sandbox approach provides a helpful safety boundary for new users. If you are debating between different visual development environments, you might find our Cursor vs Claude Code comparison guide helpful for understanding IDE layouts.
First Steps: How to Use Claude Code
Now that your software is running, you need to understand the basic workflow. The system operates entirely based on the context of the folder you have open.
When you initiate a session, the software asks you to choose a working directory. Create an empty folder on your desktop for your first project. Select that folder within the application. You can now type a natural language prompt asking the system to build a basic tool, such as a calculator or a simple game.
You will encounter two primary operating modes. The default mode is often “Accept Edits”, which allows the system to write and alter files automatically. However, beginners should always switch to “Plan Mode” for new requests.
Think of Plan Mode like an architect drafting a blueprint before the construction crew starts hammering. You get to review the entire design, ask questions, and approve the structure before any permanent changes occur. Once you approve the plan, the system transitions to building the requested files.
Setting Up GitHub and MCP Connectors
Advanced workflows rely on connecting external tools to your coding assistant. You achieve this using the Model Context Protocol (MCP). This standard allows the AI to interact with your other software platforms securely.
Connecting a GitHub repository is a highly recommended first step. By linking GitHub, your coding assistant can automatically commit changes, push updates, and manage version control on your behalf. You will need to generate a Personal Access Token from your GitHub developer settings. Paste this token into your local environment configuration when prompted.
You can also connect productivity tools like Google Drive or Slack. These integrations allow the software to read project documentation or pull task lists directly into your coding environment. Exploring these integrations is a great way to advance your abilities. You can learn more about maximising these connections in our comprehensive guide to advanced Claude Code skills and automation.
Quick recap: After installation, start in a safe local folder using Plan Mode. Then, connect your GitHub account via MCP to secure your version control workflow.
Coverage Highlights and Practical Value
A successful local installation completely changes the dynamic of AI programming. When you use a web-based chat tool, you constantly copy and paste code snippets back and forth. You have to manually track which files changed and where bugs originated.
Running the system locally eliminates that friction. The software reads your entire file directory simultaneously. It writes the code, saves the files, and even runs internal tests to verify its own work. This autonomous capability saves hours of manual debugging. Furthermore, it creates a persistent memory of your project architecture. By generating a claude.md file in your root folder, the system permanently records your stylistic preferences, tech stack, and core rules for all future sessions.
Frequently Asked Questions
Is the software free to download?
The application itself is free to download. However, you must have an active paid subscription (Pro or Max tier) to authenticate and actually use the coding features.
Do I have to use the terminal?
No. While many developers prefer the terminal for speed, the official desktop application provides a fully visual interface that requires zero command-line experience.
Can it read my entire computer?
No. The system operates within a sandboxed environment. It only accesses and modifies files within the specific project folder you explicitly permit it to read.
Does it work without an internet connection?
No. The application runs locally on your machine, but it requires a constant internet connection to communicate with the cloud-based AI models that process your requests.