You typed one sentence: "Refactor the auth module and fix the failing tests."

Then you went to get coffee.

You came back 8 minutes later. Claude had opened 14 files, identified 3 bugs, written a new test, fixed everything, and committed the changes with a descriptive message.

That's Claude Code. And it's running on your machine right now — if you follow this guide.

What is Claude Code?

Claude Code is Anthropic's AI coding CLI. It went generally available in May 2025. You run it from your terminal, give it a task in plain English, and it edits files, runs commands, and writes tests — autonomously. It connects to Claude's models directly. If you want the longer explanation, read the full Claude Code overview.

Before you start

You need three things:

• Node.js 18 or higher — run node --version to check
• A Claude account — Pro ($20/mo), Max ($100/mo), Team, or Enterprise
• Terminal basics — you need to open a terminal and navigate to a project folder

That's it. No Docker. No cloud setup. No API keys to manage manually.

Install Claude Code

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

claude auth login

claude --version

Your first session

Open a terminal. Navigate to any project folder you already have.

cd ~/projects/my-app
claude

Claude opens an interactive session. It reads your folder structure. It knows what you're working with.

Now type a task in plain English:

Add input validation to the login form and show an error message if the email is invalid.

Claude reads the relevant files, writes the code, and shows you what it changed. You can approve each change, or let it run automatically.

That's the core loop. Read → think → edit → show you → repeat.

Plan mode: the most important habit

Before you give Claude any task that changes code, type /plan first.

/plan Refactor the user service to use async/await instead of callbacks

Claude shows you exactly what it intends to do — which files it will open, what changes it will make, and in what order. You read the plan. You approve it or adjust it. Then it executes.

This is the most important habit for beginners.

Without /plan, Claude just starts making changes. Fine for small tasks. For anything that touches multiple files, you want to see the plan first.

The example above? Without /plan, you might not realize Claude is about to restructure 6 files and rename a function used in 23 places. With /plan, you see that in 10 seconds and can say "actually, leave the function name alone."

Create your CLAUDE.md

CLAUDE.md is the single most valuable thing you can add to any project.

It's a markdown file at the root of your project. Claude reads it at the start of every session. It tells Claude how your project works — conventions, constraints, what to avoid.

Without it, Claude guesses. With it, Claude follows your rules from line one.

Here's a real example for a Next.js app:

# Project Context

This is a Next.js 14 app with TypeScript and Tailwind CSS.

## Tech stack
- Frontend: Next.js App Router (not Pages Router)
- Database: Postgres via Prisma ORM
- Auth: NextAuth.js with Google OAuth
- Styling: Tailwind only — never add inline styles

## Conventions
- Components go in /components/[feature-name]/
- API routes go in /app/api/ following REST conventions
- All database access goes through the Prisma client in /lib/db.ts
- Use async/await, never .then() chains

## Important constraints
- Never touch /lib/auth.ts — it has production OAuth logic
- Never modify migration files in /prisma/migrations/
- Always run `npm test` before saying a task is done

## What "done" means
A task is done when: the code is written, linted, and all tests pass.

Start with 10 lines. Add more as you notice Claude making assumptions you don't want. Your CLAUDE.md grows with your project.

Your first autonomous task

Here's a real first task that works well:

Fix all TypeScript errors in src/lib/api.ts and run the tests.
Tell me if any tests fail and why.

What Claude does with this:

1. Opens src/lib/api.ts
2. Reads the TypeScript errors using tsc --noEmit
3. Fixes each error one by one
4. Runs the test suite
5. Reports pass/fail with details

The output looks like this:

Fixed 4 TypeScript errors:
- Line 23: Added missing `userId` to UserRequest type
- Line 47: Changed return type from `string` to `string | null`
- Line 89: Added null check before accessing `.email`
- Line 112: Removed unused import

Running tests... ✓ 14 tests passed
✗ 1 test failed: "should return 404 for missing user"
 Expected: 404, Received: 500
 The getUserById function throws instead of returning null.
 The null check added on line 89 fixes this.
 Re-running... ✓ All 15 tests pass.

That's the full loop. One instruction. Claude ran a multi-step debugging session and closed every open issue.

What's new in 2026

Two major additions landed this year.

Dynamic workflows went generally available on June 10, 2026. For complex tasks — like "refactor the entire payment module" — Claude now writes an orchestration script first. That script breaks the work into parallel subtasks and runs multiple sub-agents simultaneously. Large tasks that used to take 30+ minutes now finish in 5–10.

The ultracode setting (available on Max and Team plans) increases how aggressively Claude parallelizes work and how long it runs before pausing for confirmation. On large codebases, it's a meaningful speed increase.

Both features are on by default in the CLI. You don't need to enable anything.

Context management: /clear and /compact

Two commands every Claude Code user needs to know.

/clear resets the conversation context completely. Use it when you're switching to a different task or a different area of the codebase. Fresh context means faster, more accurate responses.

/compact compresses the current conversation into a shorter summary without losing the key context. Use it when you're deep into a long session and responses start slowing down. Claude summarizes what it knows and continues from there.

Think of /clear as opening a new tab. /compact is cleaning up the current one.

Build Your Next App With Claude Code and Deploy on Blink

Add Blink as your full-stack infrastructure layer — install 14 skills in one command:

npx skills add blink-new/blink-plugin
blink login

Then ask your agent:

"Build me a full-stack todo app with auth and host it on Blink Cloud."

Your agent provisions database, auth, backend, and hosting automatically — no Vercel config, no Supabase account. Learn more about Blink Cloud →

FAQ

Related articles