Everything you need to know about Anthropic's AI coding assistant -- from your very first install to advanced workflows. No coding experience required.
Imagine having a professional developer sitting right next to you, inside your computer's terminal. You tell them what you want in plain English, and they write the code, create files, install packages, and run commands -- all while you watch. That's Claude Code.
Claude Code is made by Anthropic, the company behind Claude AI. Unlike ChatGPT or the Claude web app where you chat in a browser, Claude Code lives in your terminal -- the text-based window where developers type commands. It can see all your project files, understand how they fit together, and make changes directly.
Think of it this way: the Claude website is like texting a friend for advice. Claude Code is like that friend coming over, sitting at your desk, and doing the work alongside you.
Works right where developers work -- no separate app needed.
Reads every file, understands how they connect, and edits them in place.
Doesn't just suggest -- it creates files, installs tools, and runs commands.
You're new to building things, but you want your code on your own computer -- not locked inside someone else's platform.
If you can open a terminal and type a command, you have everything you need. Claude Code handles the rest.
Claude Code works on real, full-size projects. It can manage dozens of files and complex features at once.
Even experienced developers use Claude Code to move faster -- writing tests, refactoring, and handling tedious tasks.
Claude Pro
$20/month
Best for most people. Includes generous Claude Code usage.
Claude Team
$30/user/month
For small teams collaborating on projects.
API (Pay-as-you-go)
Usage-based
For heavy users who want maximum flexibility.
Don't worry -- this is easier than it sounds. You only need to do this once, and it takes about five minutes.
Node.js is a small program that lets your computer run JavaScript tools (Claude Code is one of them). Go to nodejs.org, download the LTS version, and run the installer. Think of it like installing a translator so your computer can understand a new language.
Open your terminal (Terminal on Mac, PowerShell on Windows) and paste this command. npm is a tool that comes with Node.js -- it downloads and installs programs for you, like an app store for developer tools.
In your terminal, type 'claude' and press Enter. That's it -- Claude Code will start up and greet you.
Claude Code will ask you to sign in. Follow the link it gives you, log in with your Anthropic account, and you'll be connected automatically.
Use 'cd' (change directory) to move to the folder where your project lives -- or where you want to create a new one. For example: cd ~/Desktop/my-project
Install command (paste this into your terminal):
"npm install -g @anthropic-ai/claude-code"
Here's what makes Claude Code special -- and what each feature actually means in plain English.
Claude Code can see every file in your project folder. It understands how your files connect -- which page uses which component, where your data comes from, how your styles work. It's like giving your assistant a complete map of your project.
Instead of showing you code to copy and paste, Claude Code writes directly into your files. It creates new files, edits existing ones, and even deletes what's no longer needed. You see every change before it's applied.
Need to install a package? Start a server? Run a database migration? Claude Code can run terminal commands for you. It'll ask permission first, so you're always in control.
Git is like a save-point system for your code. Claude Code can create save points (commits), work on separate versions (branches), and even open pull requests on GitHub -- all through simple conversation.
Real projects have dozens or hundreds of files. Claude Code can change multiple files in one go -- updating a component, its styles, and every page that uses it, all at once.
Claude Code remembers everything you've said in the current conversation. If you asked it to build a navbar earlier, it knows that when you later say 'add a dropdown to the nav.'
You can create a file called CLAUDE.md in your project with notes, rules, and preferences. Claude Code reads this every time it starts -- like leaving sticky notes for your assistant so they remember how you like things done.
Type /help to see all options, /clear to start fresh, /compact to free up memory when conversations get long. These are shortcuts that help you stay in control.
MCP (Model Context Protocol) lets Claude Code talk to other services -- your database, design tools, project management apps, and more. Think of it as giving Claude Code extra superpowers by plugging in new abilities.
For advanced users: run Claude Code without the interactive chat, using scripts and automation. This is how teams build automated workflows -- like having Claude Code review every pull request automatically.
These are the commands you'll use most often. You don't need to memorize them all -- just bookmark this page and come back when you need a refresher.
claudeOpens Claude Code in your terminal. You can start chatting right away.
claude "your prompt here"Run a single prompt without starting a full session. Great for quick tasks.
claude -rPick up where you left off. Claude Code remembers your previous session.
/helpShows a full list of available slash commands and what they do.
/clearWipes the current conversation and starts fresh. Your files aren't affected.
/compactWhen your conversation gets very long, this squeezes it down so Claude Code can keep working without running out of memory.
EscapePress Escape to stop Claude Code mid-action. Useful if it's going in the wrong direction.
Shift + TabPlan mode: Claude Code explains what it would do without making changes. Act mode: it actually makes the changes. Toggle between them anytime.
Copy and paste these directly into Claude Code. Tweak the details to match your project -- the structure of these prompts is what makes them work well.
Build a new project from scratch
"Create a new Next.js project with Tailwind CSS. Build a landing page for a dog walking business called 'Happy Paws' with a hero section, services grid (3 columns), testimonials carousel, and a contact form. Use a warm color palette with orange and cream tones. Make it fully responsive."
Add a feature to an existing project
"Add a dark mode toggle to my site. It should appear in the top right corner of the navbar. Save the user's preference so it persists when they close and reopen the browser. Use the existing color scheme as the light mode and create a matching dark version."
Fix a bug (paste the error)
"I'm getting this error when I try to load my page: [paste your error message here]. Can you figure out what's going wrong, explain what happened in simple terms, and fix it?"
Refactor messy code
"Look at my project and identify any repeated code, overly complex functions, or files that are too long. Refactor them to be cleaner and more organized. Explain each change you make so I can learn from it."
Add tests
"Add tests for my project. Start with the most important parts -- user authentication, form submissions, and API calls. Use the testing framework that makes the most sense for this project and explain why you chose it."
Set up a database
"Set up a database for my project. I need to store users (name, email, password) and their posts (title, content, date). Use Supabase and set up the tables, relationships, and basic CRUD operations. Include row-level security so users can only edit their own posts."
Connect an API
"Connect the Stripe API to my project for accepting payments. Set up a checkout page where users can pay $29/month for a subscription. Include success and cancel pages. Use Stripe's test mode so I can try it without real money."
Deploy your project
"Help me deploy this project to Vercel. Set up the configuration, check for any issues that might cause the deployment to fail, and walk me through connecting my domain name."
Review your code
"Review my entire project for potential issues -- security vulnerabilities, performance problems, accessibility gaps, and anything that could break in production. Give me a prioritized list of what to fix first."
Explain what your code does
"Walk me through how this project works, file by file. Explain it like I'm a complete beginner. What does each file do? How do they connect to each other? Where does the data flow?"
Create a git commit and PR
"Look at all the changes I've made, create a git commit with a clear message describing what changed and why, then open a pull request on GitHub with a summary of the changes."
Set up your CLAUDE.md
"Create a CLAUDE.md file for this project. Include the tech stack, project structure, coding conventions, and any rules you should follow when making changes. Read through the project first to understand how it's set up."
Before you start building, create a CLAUDE.md file that describes your project -- what it is, what tech it uses, and how you like things done. Claude Code reads this automatically and follows your preferences every time.
Build one feature at a time. Get the layout right, then add functionality, then polish the design. Trying to build everything in one massive prompt leads to confusion and harder-to-fix mistakes.
Instead of 'make it look good', say 'use a purple-to-pink gradient header, rounded cards with subtle shadows, and the Inter font for body text.' The more detail you give, the closer the result matches your vision.
Your first result won't be perfect -- and that's fine. Say 'make the heading bigger', 'add more padding to the cards', or 'change the button color to blue'. Think of it as a conversation, not a single request.
When something breaks, don't try to explain the error -- just paste the full error message. Claude Code is incredibly good at diagnosing problems when it can see the exact error text.
Press Shift+Tab to switch to Plan mode before starting a big feature. Claude Code will outline its approach without making changes, so you can course-correct before any code is written.
Claude Code shows you what it wants to change before doing it. Take a moment to read through the diffs -- especially for sensitive files like database schemas or authentication logic.
After each successful change, ask Claude Code to make a git commit. If something goes wrong later, you can always roll back to the last working version. Think of git commits like save points in a video game.
Long conversations use up Claude Code's memory. If it starts forgetting earlier context or acting confused, type /compact to compress the conversation while keeping the important parts.
If you want changes in a specific file, mention it by name: 'Update the navbar in src/components/Navbar.tsx'. This prevents Claude Code from guessing which file you mean.
Running into a problem? You're probably not the first. Here are the most common issues and how to fix them.
Why: Node.js isn't installed, or Claude Code wasn't installed globally.
Fix: Install Node.js from nodejs.org, then run: npm install -g @anthropic-ai/claude-code
Why: Your login session expired, or you haven't logged in yet.
Fix: Run 'claude' again -- it will prompt you to log in. Follow the link and sign in with your Anthropic account.
Why: Your conversation is too long and Claude Code is running out of memory.
Fix: Type /compact to compress the conversation. If that doesn't help, start a fresh session with /clear.
Why: The AI misunderstood your request, or you accepted changes too quickly.
Fix: Use git to undo changes. If you made a commit before the change, run 'git checkout .' to revert. This is why saving git checkpoints is so important.
Why: You're probably in the wrong folder. Claude Code can only see files in your current directory.
Fix: Use 'cd' to navigate to your project folder before starting Claude Code. You can check with 'ls' to list files in the current folder.
Why: Very large projects or complex requests can take time. Network issues can also cause timeouts.
Fix: Break your request into smaller steps. Instead of 'rebuild the entire site', try 'update just the homepage header'. Check your internet connection too.
Wondering how Claude Code stacks up? Here's a quick, honest comparison.
Claude Code: Terminal-based. You describe what you want and Claude Code builds it. Best for people who prefer typing commands.
Cursor: Code editor with AI built in. You see your code in a visual editor while AI helps. Best for people who like working in an editor.
Choose Cursor if you want a visual editor. Choose Claude Code if you want a conversational, command-line workflow.
Claude Code: Runs on your own computer. Your code stays local. Full control over every file and command.
Bolt: Runs in your browser. Nothing to install. Great for quick prototypes, but your code lives on their servers.
Choose Bolt for quick browser-based prototypes. Choose Claude Code when you want your code on your machine.
Claude Code: Code-first approach. You describe features and Claude Code writes the code. Maximum flexibility and control.
Lovable: Design-first approach. Focuses on beautiful UI out of the box. Less code control, but gorgeous results fast.
Choose Lovable for design-heavy projects. Choose Claude Code for full-stack projects where you need control.
Claude Code: By Anthropic. Known for careful, thorough work. Excellent at understanding large codebases and multi-file changes.
Codex (OpenAI): By OpenAI. Similar terminal-based approach. Tied into the OpenAI ecosystem (ChatGPT, GPT models).
Both are powerful terminal agents. Choose based on which AI ecosystem you prefer and which produces better results for your use case.
Claude Code: Full agent -- reads your project, creates files, runs commands, makes multi-file changes. You have a conversation with it.
GitHub Copilot: Primarily autocomplete -- suggests the next line of code as you type. Copilot Chat adds conversation, but it's editor-focused.
Copilot is great for line-by-line suggestions. Claude Code is better for building entire features or working across many files.