Claude Code has fundamentally changed how software gets built. This entire website (chiraghasija.cc) was built with it. Features that would have taken months shipped in days.

This isn’t hype. This is a practical guide to getting started, what it actually does well, and how to get the most out of it.

What Is Claude Code?

Claude Code is Anthropic’s agentic coding tool that runs directly in your terminal. Not a browser tab. Not a VS Code sidebar that autocompletes your lines. It’s a full coding agent that can:

  • Read your entire codebase and understand the architecture
  • Write, edit, and create files across your project
  • Run terminal commands (build, test, deploy)
  • Search the web for docs and references
  • Manage git operations (commits, PRs, branches)

Think of it as having a senior engineer sitting next to you who never gets tired, never judges your questions, and can context-switch between your frontend, backend, and infra without blinking.

Installation

Getting started takes about 30 seconds:

npm install -g @anthropic-ai/claude-code

That’s it. You need Node.js 18+ installed. Once it’s global, just navigate to any project directory and run:

claude

It’ll ask you to authenticate with your Anthropic account on the first run. After that, you’re in.

Your First Session

When you launch Claude Code inside a project, it immediately starts understanding your codebase. You don’t need to explain your folder structure or point it to specific files. It just knows.

Try something simple first:

> explain the architecture of this project

It’ll read through your files, understand the relationships, and give you a clear breakdown. This alone is incredibly useful when onboarding onto a new codebase.

Then try something real:

> add a dark mode toggle to the settings page

Watch what happens. It’ll:

  1. Find your existing settings page
  2. Understand your styling approach (CSS modules, Tailwind, whatever)
  3. Create the toggle component
  4. Wire up the state management
  5. Ask you before making anything destructive

It’s not generating a code snippet for you to copy-paste. It’s actually doing the work inside your project.

The CLAUDE.md File — Your Secret Weapon

This is the feature most people miss. Create a CLAUDE.md file in your project root:

# Project Instructions

- This is a Next.js 14 app with TypeScript
- We use Tailwind CSS for styling
- All API routes go in /app/api/
- Run tests with: npm run test
- Never use any as a TypeScript type
- Prefer server components unless state is needed

Claude Code reads this file at the start of every session. It’s like giving your AI partner a team onboarding doc. The more specific you are, the better it performs.

Pro tip: Add your code style preferences, naming conventions, and architectural decisions here. Claude Code will follow them consistently.

Real Workflow: How to Actually Use It

Here’s what a typical session looks like:

1. Start with context

> read the README and understand the project structure

2. Plan before building

> I want to add a blog section. Plan the implementation
> before writing any code.

Claude Code will lay out a step-by-step plan and ask for your approval before touching anything. This is huge - it doesn’t just yolo changes into your codebase.

3. Build iteratively

> implement step 1 of the plan

Review the changes. Give feedback. Continue.

4. Test and fix

> run the tests and fix any failures

It’ll execute your test suite, read the failures, understand the root cause, and fix them. Often in a single pass.

5. Ship it

> commit these changes with a clear message

It writes the commit message, stages the right files, and commits. You can even ask it to create a PR with a proper description.

Claude Code in VS Code

If you’re not a terminal person, Claude Code also has a native VS Code extension — and it’s not an afterthought. It’s the full agent, right inside your editor.

How to Set It Up

  1. Open VS Code
  2. Go to Extensions (Cmd+Shift+X on Mac, Ctrl+Shift+X on Windows/Linux)
  3. Search for “Claude Code”
  4. Install the extension by Anthropic
  5. Authenticate with your Anthropic account when prompted

That’s it. You’ll see a Claude Code panel in your sidebar.

What You Get Inside VS Code

The extension isn’t just a chat window bolted onto your editor. It’s deeply integrated:

  • Inline diffs — When Claude Code edits a file, you see the changes highlighted right in the editor. Accept or reject individual changes, just like a code review.
  • File awareness — It knows which file you have open, what you’ve selected, and what your current cursor position is. Say “refactor this function” and it knows exactly which one you mean.
  • Terminal integration — It can run commands in your VS Code integrated terminal. No switching windows.
  • Status bar — See what Claude Code is doing at a glance. Reading files, running commands, waiting for your input — it’s all visible.
  • One-click accept/reject — Every file change gets a diff view. You stay in control of what actually gets written.

The Best Part: Context from Your Editor

When you highlight code and ask Claude Code a question, it automatically uses your selection as context. This is incredibly useful for:

Select a function → "explain what this does"
Select a block → "refactor this to use async/await"
Select a test → "this test is failing, fix it"

You don’t need to copy-paste code or explain where things are. Your editor selection is the context.

Terminal vs VS Code — Which Should You Use?

Honestly? Both. Here’s how to think about it:

  • VS Code extension — great for focused work on specific files, reviewing diffs visually, and when you want that tight editor integration
  • Terminal — great for project-wide tasks, git workflows, running multi-step operations, and when you want maximum speed

They use the same underlying agent, so the capability is identical. It’s just about which interface fits your workflow at that moment.

What Makes It Different

Compared to Copilot, Cursor, and ChatGPT with copy-paste workflows, here’s why Claude Code hits different:

It’s agentic, not autocomplete. It doesn’t predict your next line. It understands your intent and executes multi-step tasks autonomously. “Add authentication” isn’t a prompt that gets you a code block - it’s a prompt that gets you a working auth system.

It runs in your terminal. No context-switching. No browser tabs. It operates directly on your files, in your environment, with your tools.

It understands your whole project. Not just the file you have open. It greps through your codebase, understands imports and dependencies, and makes changes that are consistent with your existing patterns.

It asks before breaking things. Destructive operations (deleting files, force-pushing, dropping tables) always get a confirmation prompt. It’s opinionated about safety.

Tips for Getting the Most Out of It

  1. Be specific about what you want. “Make the homepage better” is vague. “Add a hero section with a gradient background, a heading, a subtitle, and two CTA buttons” is actionable.

  2. Use CLAUDE.md religiously. The more context you give upfront, the less you need to correct later.

  3. Review changes before committing. Claude Code is good, but you’re still the engineer. Read the diffs.

  4. Break big tasks into smaller ones. Instead of “build me an e-commerce site”, go feature by feature. Add cart. Add checkout. Add payment integration. One at a time.

  5. Let it run your tests. Seriously. The feedback loop of “write code → run tests → fix failures” is where Claude Code shines the most.

  6. Use slash commands. /commit for quick commits, /review-pr for PR reviews, /help when you’re stuck. They’re shortcuts that save time.

What It’s Not Great At (Yet)

Being honest here:

  • Very large refactors across dozens of files can sometimes lose context. Break them into smaller chunks.
  • Highly domain-specific logic (like complex financial calculations) still needs your expertise to verify.
  • UI pixel-perfection — it’ll get the structure right, but you might need to tweak spacing and colors manually.

These aren’t dealbreakers. They’re just areas where you stay in the driver’s seat.

The Bottom Line

Claude Code isn’t replacing developers. It’s making good developers unreasonably productive. The gap between “I have an idea” and “it’s live in production” has never been smaller.

If you write code for a living - or even as a hobby - you owe it to yourself to try this. The setup takes 30 seconds. The productivity gain is permanent.

npm install -g @anthropic-ai/claude-code

Go build something.