CLAUDE.md Is Not Optional

Here's a hot take: if you're using Claude Code without a CLAUDE.md file, you're not really using Claude Code. You're using a fancy autocomplete with a terminal UI.

What CLAUDE.md Actually Does

Think of it as the onboarding doc for your AI pair programmer. It tells Claude Code:

• What stack you're using (so it stops suggesting Express when you use Hono)
• Your coding conventions (tabs vs spaces, naming patterns, file structure)
• How you want responses formatted (direct, no fluff, code-first)
• What to read at session start and what to write at session end

Without this file, Claude Code starts every single session from zero. No memory. No preferences. No context. Every. Single. Time.

The Five Layers

CLAUDE.md isn't just one file. It's a five-layer architecture:

1. Enterprise — org-wide rules everyone follows
2. Project root — your repo's CLAUDE.md
3. Feature-scoped — nested CLAUDE.md in subdirectories
4. User-level — your personal ~/.claude/CLAUDE.md
5. Session-level — memory and handoffs that persist

The One-Minute Version

If you do nothing else today, run claude /init in your project. It auto-generates a starter CLAUDE.md by scanning your codebase. Then spend 60 seconds adding your preferences.

That single file is the difference between "Claude Code is meh" and "Claude Code is my operating system."

Not optional. Foundational.