Start AI Coding with One Command: claude (Beginner's First Step)

Lots of people install Claude Code to try AI coding, only to get stuck wondering, "So how do I actually start it?" The answer is surprisingly anticlimactic: just type claude in the terminal. But there's one important rule hidden in that single line, and knowing it keeps you from getting confused from the start.

Definition (What It Is)

claude is the most basic command for launching Claude Code. Just type claude in the terminal (command prompt), and a chat window opens, taking the folder you're currently in as its workspace. The key point here is "current folder = work target." That is, Claude looks at and works on the files inside the very folder where you typed the command. So which folder you launch it in matters a great deal.

How to Use It (By Difficulty)

Before you start: Open the terminal and first move into the project folder you want to work on. You move with the cd folder-path command (cd = change directory). The first time you run it, it may ask you to log in—we'll cover that in "Common Pitfalls" below.

Basic — Move into a folder, then launch

cd ~/projects/my-blog
claude

This flow moves into the blog folder first, then launches claude. ~ is shorthand for your home folder; on Windows that's C:\Users\username\. So on Windows you'd write something like cd C:\Users\username\projects\my-blog. When you run it, the screen switches to conversation mode, an input box appears at the bottom, and you can type commands as if chatting. To finish, type /exit or press Ctrl+C twice.

Applied — Assign the first task as you launch

claude "이 프로젝트 구조를 설명해줘"

If you append a task in quotes after claude, it starts on that task the moment it launches. It's especially good for grasping "what the whole thing looks like" when you first open a new project.

Advanced — Start with a chosen model and extra folder

claude --model opus --add-dir ../shared-lib

--model opus is the option for choosing which AI model to work with, and --add-dir is the option that tells it to look at an additional folder beyond the current one (../shared-lib points to a shared library folder inside the parent folder). It's useful when dealing with a project spanning multiple folders.

Common Pitfalls

The first time you launch claude, it may ask you to log in. When that happens, don't panic—just type claude auth login to log in with your Anthropic account. You usually only have to log in once, after which it launches right away.

One more thing: if you launch it in the wrong folder, Claude ends up looking at the wrong files. It's a good habit to check "am I in the right folder right now?" with the pwd command (which shows the current folder path) before launching.

Real-World Example

There's a case where, opening blog code for the first time on a new laptop, someone moved into the folder and grasped the entire picture of the project in five minutes with the single line claude "Read the README and summarize the project". It's far faster than opening and reading each file one by one yourself.

More Ways to Use It

• Since it operates on a per-folder basis, splitting projects into separate folders keeps conversation history and settings cleanly separated.
• For projects you use often, jot down the folder location and get into the habit of jumping straight there with cd.

How Others Use It

• The official docs lay out the four-stage flow (explore → plan → implement → commit) that Anthropic recommends and the creators use daily: Common workflows — Claude Code Docs
• A write-up of how a real developer uses Claude Code every day is also worth a look: How I Use Claude Code — Boris Tane

Wrap-Up

claude is the start button for AI coding. Just remember the rule that "the current folder is the work target," and you can move into the right folder and kick off your work with the single line claude. Even if a login prompt appears, claude auth login resolves it, so don't be intimidated.

Reference: Claude Code v2.1.154 (2026.05)