Most AI coding tools are simple wrappers. They take your prompt, send it to an API, and print the response. Claude Code is different. It's an agentic CLI tool that lives inside your terminal, designed to be an active participant in your development workflow. Instead of just suggesting code, it can read your project files, understand their relationships, run tests, and even commit changes. It operates with a degree of autonomy that simple prompt-response tools lack.

Getting started requires Node.js and npm. The tool is installed globally, making the claude command available from any directory.

The first time you run claude in your terminal, it will initiate an OAuth authentication flow. Your web browser will open, asking you to log into your Anthropic account and authorize the CLI tool. This process securely links the terminal application to your account without you having to manually manage API keys.

Once authorized, Claude Code stores a token locally. You won't need to repeat this process unless the token expires or is revoked. This single sign-on makes subsequent sessions seamless.

After authenticating, you'll be greeted by a TUI (Terminal User Interface). This is a text-based interface within your terminal that's more dynamic than a simple command prompt, often featuring panes, menus, and real-time updates.

Because Claude Code can read and write files, security is paramount. It doesn't get unrestricted access to your entire file system. Instead, it operates within a sandboxed environment defined by a configuration file. You explicitly grant it permission to access specific directories using an allowlist in your project's configuration.