Part 2: Understand Your Workspace
The Problem
Claude Code is powerful — but only if it understands YOUR project.
The difference between generic AI output and genuinely useful work comes down to one thing: context.
This lesson teaches you how Claude Code sees your project and how to make it smarter.
How Claude Code Sees Your Project
When you launch claude inside a project folder, it can:
Read any file in the project
Source code, config files, documentation — everything in your folder is visible.
Understand file relationships
Imports, references, links — it sees how pieces connect.
See your git history
What changed, when, and why — full commit context.
Run commands in the project directory
Build, test, deploy — it can execute real tasks.
Remember project-specific instructions
Via CLAUDE.md — your project's custom instruction manual.
What Claude Code Sees
📁 your-project/
├── 📄 CLAUDE.md ← “Your project's instruction manual”
├── 📁 src/ ← “All your source code”
├── 📁 public/ ← “Images, assets, static files”
├── 📄 package.json ← “Dependencies and scripts”
└── 📄 README.md ← “Project documentation”
CLAUDE.md — Your Project's Brain
This is the single most important concept in this course. CLAUDE.md is a markdown file you place in your project root that tells Claude Code:
What the project is about
Purpose, domain, audience — the big picture.
What tech stack it uses
Frameworks, languages, tools — so it writes compatible code.
Your preferences and conventions
Naming patterns, file structure, style rules.
What to always/never do
Guardrails that prevent mistakes before they happen.
Example CLAUDE.md
# Project: Marketing Website
Stack: Next.js + Tailwind CSS
Domain: marketing.company.com
## Rules
- Never change the header or footer without asking
- Always use our brand colors (#1a1a2e, #e94560, #0f3460)
- Commit messages should be descriptive
- Always run the build before pushing
Think of CLAUDE.md as the brief you'd give a new freelancer. The better the brief, the better the work. Except this freelancer never forgets and never needs to be reminded.
The .claude/ Directory
Every project that uses Claude Code has a .claude/ directory. Here's what's inside:
settings.json
Permissions and preferences — what Claude Code is allowed to do in this project.
Memory files
Things Claude Code learns about your project over time — patterns, preferences, decisions.
You mostly don't need to touch this directory — it manages itself. Think of it as Claude Code's personal notebook for your project.
Context is Everything
See how the same request produces wildly different results depending on whether Claude Code has context:
Without Context
“Make the homepage better”
Generic changes, wrong brand colors, mismatched styles. Claude Code guesses — and guesses wrong.
With CLAUDE.md
Same request → totally different result
CLAUDE.md defines your brand, stack, and conventions → changes match your exact design system.
Without Context
“Add a contact form”
Random placement, default styling, no validation. It looks like it belongs on a different site entirely.
With CLAUDE.md
“Use Tailwind, match existing form patterns”
Consistent with your site — right placement, right styling, proper validation patterns.
Without Context
“Fix the bug on the pricing page”
Might fix one thing and break three others. No safety net, no verification.
With CLAUDE.md
“Always run tests before committing”
Safe, verified fix — Claude Code runs your test suite before committing anything.
What You Just Learned
- You understand how Claude Code reads and navigates your project
- You know what CLAUDE.md is and why it's the single most important file for AI quality
- You can write a basic CLAUDE.md that makes Claude Code 10x more useful
- You understand that context = quality (better instructions → better output)
Next: Now that Claude Code understands your project, let's learn how to give it instructions that actually work.