Your First 10 Minutes with Claude Code

by

You’ve installed Claude Code. You’ve logged in. Now you’re staring at a terminal and wondering what to actually type.

This guide walks you through your first real session from start to finish. By the end of it you’ll have run Claude Code on an actual project, seen it in action, and know the basic workflow you’ll use every time.

This is part of our complete Claude Code Beginner’s Guide.

TL;DR

  • You start Claude Code by running claude in your project folder
  • You give it tasks in plain English — no special syntax required
  • Claude Code will show you what it’s doing before (usually) making changes
  • You review, approve or reject, and it continues
  • Your first task could be as simple as “explain this codebase to me”

Before You Start: What You Need

Make sure you’ve completed installation and login first. If you haven’t:

You also need a project folder to work in. This guide assumes you have something on your computer — even a small project with a few files works perfectly. If you truly have nothing, we’ll show you how to create a simple test project.

Step 1: Navigate to Your Project Folder

Open your terminal and navigate to the folder where your project lives. The cd command (change directory) moves you into a folder.


# Navigate to your project folder (replace with your actual path)
cd ~/Documents/my-project

If you’re not sure of the path, you can also drag the folder from Finder (Mac) or Explorer (Windows) into the terminal window and the path will appear automatically.

To confirm you’re in the right place:


# List the files in your current folder
ls

You should see your project’s files listed. On Windows Command Prompt, use dir instead of ls.

Terminal showing the ls command output with a list of project files like index.js, package.json, README.md

Step 2: Start Claude Code

Type this and press Enter:


# Launch an interactive Claude Code session in the current folder
claude

Claude Code starts and you’ll see an interface appear in your terminal. It shows your current folder, some keyboard shortcut hints, and a prompt waiting for your input.

The Claude Code interface open in terminal showing the project path, a welcome message, and a text input prompt at the bottom

You’re now inside a Claude Code session. Everything you type here is a message to Claude Code.

Step 3: Ask Claude Code to Explain Your Project

The best first thing to do in any new Claude Code session is ask it to understand your project. Type this:


What files are in this project and what does each one do?

Press Enter. Watch what happens.

Claude Code will scan your project folder, read through your files, and give you a plain-English summary of what it found. For a typical web project, it might say something like:


I can see this is a Node.js web application. Here's what I found:

- package.json: Defines the project and its dependencies
- index.js: The main entry point — sets up an Express server on port 3000
- routes/: Contains three route files handling users, posts, and authentication
- models/: Database models for User and Post using Mongoose
- .env.example: Shows the environment variables this project needs

This is genuinely useful. You learned something about your own project just by asking. And now Claude Code understands the structure, which means every task you give it from here will be more accurate.

Step 4: Give Claude Code a Real Task

Now try asking Claude Code to do something. Start with something small and low-risk. A few good first tasks:

If you have a project with code:


Add a comment to each function in index.js explaining what it does

If you have a README file:


Update the README.md to include a "Getting Started" section with installation instructions

If you want to see its reasoning:


Look at this codebase and suggest three improvements I could make to the code structure

Type one of these (or your own version) and press Enter.

Step 5: Understand What Claude Code Does Next

After you give Claude Code a task, it doesn’t just silently run. You’ll see it work through a process:

  1. It reads relevant files. You’ll see lines like Reading index.js... or Scanning routes/... — this is Claude Code gathering context.
  1. It proposes what it plans to do. Before making changes, Claude Code typically explains its plan and may show you a diff (a view of what will change) for approval.
  1. It asks for permission on significant actions. Writing to files, running commands, and other real actions often require you to confirm.
  1. It makes the changes.
  1. It tells you what it did and often asks if there’s anything else you want to adjust.

Claude Code terminal showing a file diff with green lines for additions and red lines for deletions, with a “Do you want to apply these changes? [y/n

” prompt]

When you see a [y/n] or similar prompt asking for approval, type y and press Enter to approve, or n to skip. Claude Code won’t make the change until you say yes.

💡 Pro Tip: You can type y to approve individual changes, or you can tell Claude Code to proceed without asking by using a command like /auto — but save that for when you’re more comfortable with how it works.

Step 6: Have a Back-and-Forth Conversation

Claude Code is conversational. After it does something, you can respond to refine or continue:


That looks good. Now add error handling to those same functions.

Or:


Actually, don't use that approach. Use arrow functions instead.

Or:


Can you explain why you added that try/catch block there?

Every message builds on the context from the previous ones. Claude Code remembers your whole conversation within a session, so you don’t need to re-explain the project with every message.

This is where Claude Code becomes genuinely powerful — you’re having a working conversation with an AI that has your full project in view and is actively making changes.

Step 7: Create a Test Project from Scratch (Optional)

If you don’t have a project handy, here’s how to create a simple one to practice with:


# Create a new folder and navigate into it
mkdir claude-test && cd claude-test

# Create a simple JavaScript file with a few functions
cat > app.js << 'EOF'
function addNumbers(a, b) {
  return a + b;
}

function greetUser(name) {
  return "Hello, " + name;
}

function calculateTax(price, rate) {
  return price * rate;
}
EOF

Now start Claude Code:


# Launch Claude Code in your new test project
claude

Try this task:


Add JSDoc comments to each function, then create a README.md that documents what this file does

You’ll see Claude Code read the file, write comments for each function, and create a README. That’s a real multi-step task done in a few seconds.

Useful Things to Know in Your First Session

How to exit

Press Ctrl + C to stop the current operation. To exit Claude Code completely, type /exit and press Enter, or press Ctrl + C twice.

How to see available commands

Type /help to see a list of built-in commands and shortcuts.

How to start a fresh conversation

Type /clear to reset the conversation context without leaving Claude Code. Useful when you’re switching to a completely different task.

How to run it without interactive mode

If you want to give Claude Code a single task and have it complete it without an interactive session, you can pass the task directly:


# Run a single task without entering interactive mode
claude "Add error handling to all functions in app.js"

This is useful for quick, one-off tasks or scripting.

What Not to Do in Your First Session

Don’t give it huge tasks immediately. Ask it to do one specific thing first. Get comfortable with the confirmation flow. Then scale up.

Don’t approve changes you haven’t read. Claude Code is capable but not infallible. Review the diffs it shows you before typing y. This is especially important when it’s touching configuration files or database-related code.

Don’t run it with auto-approve on production code right away. Once you’re comfortable, features like auto mode are genuinely useful. But learn what Claude Code does before you let it run without checkpoints.

Don’t forget to save its output. Claude Code makes real changes to your real files. These are immediately on disk. If you want to undo something, you’ll need git. Always work in a git repository — or at minimum, make a copy of your project before experimenting.

⚠️ Warning: Claude Code edits real files on your machine. If you’re worried about something going wrong, commit your current state to git before a session, or work in a separate branch. That gives you a clean way to revert if needed.

Common Things People Ask in Their First Session

Here are real examples of productive first-session tasks:


Explain the architecture of this project to me


What would be a good first feature to add to this project?


Find any obvious bugs or issues in this codebase


Add TypeScript types to this JavaScript file


Write unit tests for the functions in utils.js


Refactor this function to be more readable

All of these work with a single natural-language message. No special syntax. No configuration. Just type and press Enter.

Key Takeaways

  • Start every session with claude in your project folder
  • Ask Claude Code to explain your project first — it sets up better context for everything that follows
  • Claude Code proposes changes and asks for approval before modifying files
  • Keep sessions conversational — you can refine and redirect as you go
  • Always review diffs before approving, especially early on
  • Use git so you have a safety net when experimenting

Next Steps

Return to the Claude Code Beginner’s Guide for the full learning path.