Week 1: Introduction to Claude Code

Date: 2026-03-31 Format: Concepts + live demo

What is Claude Code?

Claude Code is Anthropic's official CLI tool that brings Claude — the AI — directly into your terminal and IDE. Instead of switching between your editor and a chat window, Claude lives where you work.

You talk to it in plain English. It reads your files, writes code, runs commands, fixes bugs, and explains what it's doing — all without leaving the terminal.

# Start Claude Code
claude
 
# Or point it at a specific task
claude "fix the failing tests in pkg/payments"

Why Not Just Use the Chat UI?

Claude.ai (chat)	Claude Code (CLI)
You paste code manually	Claude reads your files directly
You describe errors	Claude sees the actual stack trace
One-shot answers	Multi-step execution with tool use
No access to your project	Full access to your codebase
Manual copy-paste loop	Changes applied in-place

The chat UI is great for questions. Claude Code is for doing work.

What Can It Actually Do?

Claude Code has access to tools it can use autonomously:

Read — reads any file in your project
Edit / Write — makes targeted changes or creates new files
Bash — runs shell commands (tests, builds, git, etc.)
Grep / Glob — searches your codebase for patterns or files
WebSearch / WebFetch — looks things up when needed

You can watch it work in real time — every tool call is shown before it runs, and you approve or deny it.

Getting Started

Install

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

Authenticate

claude
# Follow the login prompt — uses your Anthropic account

First run

cd your-project
claude

Claude reads your project and is ready to help. No configuration needed.

The Basics — How to Talk to It

Claude Code works best when you're specific:

# Too vague
"fix the bug"

# Better
"the payment webhook handler in app/Http/Controllers/WebhookController.php
is returning 500 on Stripe events — check the signature validation"

You can also give it multi-step tasks:

"read the failing test in tests/Feature/PaymentTest.php,
find the root cause, fix it, and run the test to confirm"

Modes

Interactive mode — default, you chat back and forth

claude

One-shot mode — run a single task and exit

claude "add PHPDoc to all public methods in app/Services/PaymentService.php"

Continue mode — resume your last conversation

claude --continue

Permissions

Claude Code asks before taking actions it's unsure about. You control what it can do:

Allow once — run this one action
Allow always — trust this type of action for the session
Deny — block it and Claude adjusts its approach

You can also set default permission levels per project.

What Makes It Different from Copilot / Cursor?

Tool	What it does
GitHub Copilot	Autocomplete — suggests the next line as you type
Cursor	AI-enhanced editor — edits files in a GUI
Claude Code	Autonomous agent — reads, reasons, acts across your whole project

Copilot fills in the blank. Claude Code does the task.

Live Demo — What We Showed Today

Installed Claude Code and logged in
Opened an existing project and asked Claude to explain the architecture
Asked it to find and fix a bug
Watched it read files, identify the issue, and apply the fix
Asked it to write a test for the fix
Ran the tests — passed

All in one conversation, without touching the editor.

Key Takeaways

Claude Code is a CLI agent, not just a chatbot
It has real tools — read, write, run, search
You stay in control — every action is visible and approvable
It works best with specific, contextual instructions
The more your project has a CLAUDE.md, the better it performs (we'll cover this next week)

Live Demo

During this class we solved a real ISO 8583 engineering task end-to-end using Claude Code — 7 files changed across 2 parallel agents in under 3 minutes.

View the live demo walkthrough →

(Class PIN required)

Next Week

We'll go deeper into the building blocks that make Claude Code project-aware and extensible: CLAUDE.md, Skills, Commands, Agents, Hooks, and Rules — and we'll build each one from scratch.