🧠 How to Set Up Claude Code(if you're a complete beginner)

TL;DR — Too Long Didn’t Read

• Claude Code is a version of Claude that takes action on your computer — it creates files, runs commands, and builds things, rather than just giving you text suggestions.

• You need a paid Claude subscription (Pro at $20/month works fine to start).

• There are four ways to use it: the Desktop App (easiest), inside VS Code or Cursor (most popular), in the terminal, or directly in a browser with no install at all.

• A file called CLAUDE.md is how you give Claude persistent instructions. Without it, every session starts from scratch.

• Plan Mode, Ask Mode, and Code Mode control how much Claude does on its own versus asking your permission first.

• Git is your undo button. Set it up before you start anything serious.

1. What Is Claude Code?

When you use Claude at claude.ai, it gives you text. Good text, usually. But it doesn't touch your files, doesn't install anything, doesn't create folders, and doesn't run programs. It talks.

Claude Code is different. It acts.

Open it in a folder on your computer and it reads everything in that folder — your files, your structure, your code. Then when you give it an instruction, it writes files, edits them, runs commands, installs packages, and produces working output. Not suggestions. Actual output.

Here's what that looks like in practice: You type "build me a personal portfolio website with a dark theme." Thirty seconds later, there's an HTML file sitting in your folder that you can open in a browser. You didn't write a line of code.

A few other things that make Claude Code meaningfully different from the browser version:

• It understands your whole project. Drop it into any folder and it immediately understands the context of what's there — the file structure, the tech stack, the naming conventions.

• It runs in your environment. When Claude Code executes a command, it runs on your machine. It can install software, use your existing tools, and access local files.

• It connects to your other tools. Through something called MCP (Model Context Protocol), Claude Code can read from Notion, write to Google Sheets, create GitHub pull requests, send Slack messages, and more — from a single prompt.

2. What You Need Before You Start

A paid Claude subscription

Claude Code doesn't work on the free plan. Your options:

• Pro ($20/month): Enough to learn and do real work. Usage limits apply but most beginners won't hit them regularly.

• Max ($100–$200/month): For heavy daily use where you don't want to get cut off mid-project.

• Team/Enterprise: For organizations with multiple people using it.

• API key via Anthropic Console: Pay-per-use pricing, which suits developers who want programmatic access.

Sign up at claude.ai if you don't have an account, then go to Settings → Billing to upgrade.

A compatible computer

• macOS 13.0 or newer (Intel and Apple Silicon)

• Windows 10 or newer (Windows also requires Git for Windows — more on this below)

• Linux — Ubuntu 20.04+, Debian 10+, and most major distribution

3. How to Install Claude Code?

Option 1: The Desktop App (Best Starting Point)

If you've never used a terminal and want to skip all of that, start here. The Desktop App gives you a visual interface — no command lines required.

How to install:

1. Go to claude.com/download and download the installer for your operating system (macOS or Windows).

2. Run the installer like you would any normal application.

3. Open the app and sign in with your Claude account.

4. Click the Code tab at the top of the app.

5. Select a folder on your computer where you want to work, or create a new empty folder.

That's it. You're in.

The Desktop App shows you every file Claude creates or changes, lets you run multiple sessions side by side, includes a visual diff view so you can see exactly what changed, and even supports scheduled tasks that run on a timer.

One honest note from people who've used both: the Desktop App is a great way to start, but if you want to do serious recurring work, learning to use Claude Code inside an editor like VS Code will serve you better over time.

Official guide: code.claude.com/docs/en/desktop-quickstart

Here’s a quick demo:

Option 2: VS Code or Cursor (Most Recommended for Regular Use)

VS Code is a free code editor from Microsoft that has become the standard development environment for millions of people — including many who aren't professional developers. Think of it as a more powerful version of a text editor, where you can see all your files in a sidebar and Claude Code works alongside them.

Cursor is a fork of VS Code with extra AI features built in. Either one works.

Step 1: Download VS Code or Cursor

• VS Code: code.visualstudio.com

• Cursor: cursor.com

Install it like any normal application.

Step 2: Install the Claude Code extension

1. Open VS Code or Cursor.

2. Click the Extensions icon in the left sidebar (four small squares).

3. Search for "Claude Code."

4. Click Install on the official extension by Anthropic.

5. After installing, an orange Claude icon appears in the top-right corner of the editor.

Step 3: Open a project folder

Go to File → Open Folder, navigate to the folder where you want to work (or create a new empty folder), and open it. This tells Claude Code where to operate.

Step 4: Launch Claude Code

Click the orange Claude icon. A panel opens on the side, and you can start typing prompts immediately. Every file Claude creates or changes will be visible in the sidebar on the left.

Why this setup matters: being able to see the files while Claude works gives you visibility and confidence. Even if you never plan to read code yourself, knowing what's being created keeps you in control.

Official guide: code.claude.com/docs/en/vs-code

Option 3: The Terminal (Full Control, More Setup)

The terminal gives you access to every Claude Code feature and is the most powerful way to use it. If this feels unfamiliar, start with Option 1 or 2 and come back.

On Mac or Linux:

Open your Terminal app (on Mac, press Cmd + Space and search "Terminal") and paste this:

curl -fsSL https://claude.ai/install.sh | bash

On Windows (PowerShell):

Open PowerShell and paste:

irm https://claude.ai/install.ps1 | iex

On Windows (Command Prompt):

curl -fsSL https://claude.ai/install.cmd -o install.cmd && install.cmd && del install.cmd

Alternative install methods:

• Homebrew (Mac): brew install --cask claude-code

• WinGet (Windows): winget install Anthropic.ClaudeCode

After installing:

Type claude --version to confirm it worked. Then type claude and press Enter. On first launch, it will ask you to sign in through your browser. Follow the prompts.

If something goes wrong at any point, type claude doctor — this command diagnoses your installation and tells you what needs fixing.

Official setup guide: code.claude.com/docs/en/setup

Option 4: Use It in the Browser (Zero Install)

If you don't want to install anything, go to claude.ai/code and start working. Your code runs on Anthropic's cloud infrastructure, which means you can start long-running tasks and check back later, work from any device including a tablet or phone, and run multiple parallel tasks without using your own computer's resources.

The trade-off: you don't have direct access to files on your local machine. For exploring Claude Code, creating things from scratch, or working on cloud-hosted repositories, it's a solid option. For work that requires access to files already on your computer, you'll want one of the other options.

4. What to Do When You Open Claude Code?

No matter which option you chose, the first thing to do is understand what you're working with.

Trust the folder: When Claude Code opens in a directory for the first time, it asks you to "trust this folder." This is a safety check. Confirm it. Claude Code needs permission to read and write files in that folder.

Start by asking what it can do: Literally type this:

What can you do?

Claude will explain its own capabilities. Get in the habit of asking any AI tool to explain itself at the start — it orients you and sometimes reveals features you didn't know existed.

Understand how permissions work: By default, Claude Code asks your permission before doing anything that changes your system — writing a file, running a command, installing a package. You'll see prompts like "Allow write to index.html?" with Yes/No options. This is Ask Mode. It's slow but safe, and it's where you should start.

As you get more comfortable with a task, you can switch to auto-approval or set specific permissions. More on modes below.

CLAUDE.md: How to Give Claude a Persistent Memory

Here's something that trips up almost every new user: Claude Code starts fresh every single session. It doesn't remember what you told it last week, what rules you set, or what decisions you made. Every session, you're starting over.

The fix is a file called CLAUDE.md. Place it in the root of your project folder, and Claude Code reads it at the start of every session. Think of it as an instruction sheet for a new assistant who shows up every morning.

What to put in your CLAUDE.md

Keep it under 200 lines. Longer files cause Claude to start losing track of earlier instructions. Only include things Claude can't figure out by reading your project files:

• Your tech stack — what languages, frameworks, or tools the project uses

• Code style preferences — naming conventions, formatting rules

• Workflow rules — "always create a Git branch before making changes," "always run tests after editing code"

• Commands Claude needs — how to build, test, or run your project

• Things to avoid — specific patterns, libraries, or approaches you don't want

What not to put in your CLAUDE.md

Skip anything Claude can already figure out from your code. Skip standard conventions it already knows. Skip long explanations or tutorials. If Claude is already doing something correctly without the instruction, don't add it.

A simple CLAUDE.md that actually works:

# Code style
- Use ES modules (import/export) syntax, not CommonJS (require)
- Destructure imports when possible

# Workflow
- Run npm test after making code changes
- Commit with a descriptive message after each working milestone

# Stack
- React frontend, Node.js backend, PostgreSQL database

How to create one

Option 1: Type /init in any Claude Code session. Claude will analyze your project and generate a starter file based on what it finds.

Option 2: Ask Claude to create one: "Analyze this project and create a CLAUDE.md that covers the tech stack, project structure, how to run the project, and any conventions you can detect."

The second approach often produces more thorough results because it's a custom prompt rather than an automated scan.

Check your CLAUDE.md into Git so it travels with the project. Review it when Claude starts making mistakes. Prune it when it gets too long. Treat it like code.

Official memory guide: code.claude.com/docs/en/memory

5. Plan Mode, Ask Mode, and Code Mode: What Each One Does

Claude Code has three working modes. Understanding when to use each one changes how useful it is.

Plan Mode

Claude reads your files and creates a detailed plan — but writes nothing and changes nothing. Use this for anything non-trivial: new features, significant refactors, anything that touches multiple files. It forces Claude to think before acting and gives you a chance to catch problems before they happen.

Toggle Plan Mode with Shift + Tab in the terminal, or use the mode selector in VS Code or the Desktop App.

Here’s how this works:

Ask Mode (Default)

Claude does the work but pauses and asks permission before each change. You approve or reject file writes and command executions one at a time. This is the right mode when you're learning or doing something unfamiliar. It's slower, but you stay in control.

Code Mode (Auto Accept)

Claude writes and executes without stopping to ask. Fastest option, but you need to trust what it's doing. Only use this once you've seen how Claude handles a particular kind of task and you're confident in the direction. Having Git set up is essential here, because mistakes can happen quickly.

Recommended workflow for any non-trivial task: Start in Plan Mode. Review the plan. Switch to Ask Mode (or Code Mode if you trust the task) to execute. Commit the result with Git.

The Commands and Shortcuts That Actually Matter

Which model to use

Claude Code lets you switch models. The three main ones:

• Opus: The most capable. Use it for complex tasks, architecture decisions, anything requiring deep reasoning. Consumes more of your usage.

• Sonnet: The everyday model. Good for most tasks. Start here.

• Haiku: Fast and lightweight. Use it for quick questions and simple tasks.

Type /model to switch. For most people, Sonnet is the right default.

6. How to Connect Claude to Your Other Tools (MCP Servers and Plugins)

MCP (Model Context Protocol) is what allows Claude Code to interact with apps you already use. Without MCP, Claude Code only works with files on your computer. With MCP, you can ask it to check your Notion page, update a Google Sheet, create a GitHub issue, or send a Slack message — all from one prompt.

What you can connect

Some commonly used integrations:

• GitHub — create repositories, manage pull requests, review code

• Notion, Google Sheets, Airtable — read and write to documents and databases

• Slack — send messages and read channels

• Figma — pull design specs into your code

• Playwright — automate browser testing

• Context7 — pull up-to-date documentation so Claude doesn't use outdated code examples

How to add an MCP server from the terminal

claude mcp add <server-name> -- <command>

Add --scope user if you want the server available across all your projects, not just the current one.

You can also install MCP servers through the Desktop App by going to Customize → Browse Plugins.

Plugins: One-Click Bundles

Plugins are pre-packaged bundles of MCP servers, skills, and settings that install in one step. Type /plugin in Claude Code to browse the official marketplace.

Three worth knowing about:

• Playwright — browser automation and testing

• Superpowers — adds brainstorming, parallel agents, and debugging workflows

• Context7 — ensures Claude always references current documentation

Official MCP guide: code.claude.com/docs/en/mcp
Plugin marketplace: claude.com/plugins

7. Git: The Undo Button You Need From Day One

Every tutorial, every guide, every experienced Claude Code user says the same thing: set up Git before you do anything else.

Git is a version control system that takes snapshots of your project. If Claude makes a mistake — and it will sometimes — you can roll back to the last working state instead of trying to manually undo a series of changes you can't fully see.

How to get Git set up

If you don't have Git installed, you can ask Claude Code to install it for you. Type: "Can you install git for me?" and it will figure out the right commands for your system. Or download it directly from git-scm.com.

Connecting to GitHub

GitHub stores your code in the cloud, which means it's backed up and you can access it anywhere.

1. Install the GitHub CLI (gh) by downloading it from cli.github.com, or ask Claude Code to do it.

2. Run gh auth login and follow the browser prompts to authenticate.

3. Claude Code can now create repositories, push code, and manage branches on your behalf.

Commit early and often

Every time Claude produces something that works, commit it. Tell Claude to do this automatically by adding a rule to your CLAUDE.md: "Always commit changes to Git with a descriptive commit message after completing each task."

If something goes wrong after you commit, you can always get back to the last good state.

8. Three Prompts to Try When You First Open Claude Code

These are designed to show you what Claude Code can actually do, not just what it can talk about.

Prompt 1 — Build something immediately:

Create a beautiful personal portfolio website with the following sections: a hero section with my name and a short bio, an About section, a Projects section with 3 placeholder cards, and a Contact section with a simple form. Use modern design with a dark theme. Make it a single HTML file I can open in a browser.

You'll have a working website in about 60 seconds. This works because the request is specific about the output, the structure, and the constraints.

Here’s how we did it:

Prompt 2 — Automate a real task:

Research the latest news about [topic you care about] on the web. Write a LinkedIn post about it in a professional but conversational tone, under 200 words, ending with a thought-provoking question. Save it to a file called linkedin-draft.md.

Replace the topic with anything relevant to you. This shows Claude Code can search the web, write content, and save files — all from one prompt.

Prompt 3 — Plan before you build (switch to Plan Mode first with Shift+Tab):

I want to build a simple task manager web app where I can add tasks, mark them complete, and delete them. Data should be saved locally so it persists when I refresh the page. Ask me questions to clarify the requirements and design preferences before you create a plan.

Claude will interview you before doing anything. This "ask me questions first" technique produces significantly better results than vague starting prompts because it forces alignment before any work happens.

9. Key Resources

Everything you need is in Anthropic's official documentation. These are the pages worth bookmarking: