Setting Up Your CLAUDE.md

The most important setup step. CLAUDE.md is the instruction file Claude reads at the start of every session · it's how you build persistent context that compounds over time.

What CLAUDE.md is

CLAUDE.md is a markdown file Claude reads automatically at the start of every session. It's your permanent instruction layer · the context that's always present without you re-explaining it every time.

Without it, Claude starts every session knowing nothing about you, your business, your stack, or how you want to work. With it, every session picks up with full context already loaded.

Two types of CLAUDE.md

Global (~/.claude/CLAUDE.md)

Applies to every Claude Code session on your machine, regardless of which folder you're in. Use this for your identity, preferences, and anything true across all your work.

Project-level (/your-project/CLAUDE.md)

Applies only when you run claude from that folder. Use this for project-specific context, goals, tech stack, and anything relevant to that one engagement.

You'll end up with both · a global one for your core identity and preferences, and a project-level one for each build.

What to put in vs leave out

CLAUDE.md has a limit. Research shows Claude can follow roughly 150–200 instructions before compliance starts to drop · and Claude Code's own system already uses about 50 of those. That leaves you roughly 100–150 slots.

Include:

• Who you are and what your business does
• Active projects and their current status
• Your tech stack and tools
• How you want Claude to communicate (tone, format, length)
• Key contacts and context it needs to know
• Any instructions Claude consistently gets wrong without being told

Leave out:

• Anything Claude can figure out from reading your files
• Standard conventions it already knows
• Detailed documentation (link to it instead)
• Anything that changes frequently (update it when it changes, don't pre-empt it)

The rule of thumb: for every line in your CLAUDE.md, ask "would Claude make a mistake without this?" If not, delete it.

The quickest way to start: /init

With a project folder open in Claude Code, run:

/init

Claude analyzes your project, detects your stack and structure, and generates a starter CLAUDE.md. The output is usually too long · review it and prune aggressively. Delete anything Claude would know anyway. Keep what's specific to how you work.

Starter template

For a fresh project with no existing code, create a CLAUDE.md manually and paste in the template below. Fill in what you know today, add to it as you work. The goal is a living document, not a perfect one.

# About Me
Name: [Your name]
Role: [What you do · e.g. founder, operator, agency owner]
Business: [Company name and one sentence on what it does]

# Current Projects
- [Project 1 · what it is, who it's for, current status]
- [Project 2 · what it is, who it's for, current status]

# How I Work
- Always draft emails and messages before sending · I review first
- Prefer bullet points for planning, prose for client-facing content
- Default tone: direct, no fluff, no unnecessary preamble
- When I ask for options, give me 2-3 max, not a long list

# My Stack
- [List the tools you use · Notion, HubSpot, Gmail, Stripe, etc.]

# Key Contacts
- [Client or team member names and relevant context]

# Important Notes
- [Anything Claude consistently gets wrong that you want it to remember]

Global CLAUDE.md setup

Create your global file:

mkdir -p ~/.claude
touch ~/.claude/CLAUDE.md

Then open it in any text editor and paste your personal details. This file applies to every Claude session on your machine, so keep it focused on who you are and how you like to work · not project-specific detail.

This file is also the seed of your brain. In the next module you'll turn this base context into a full second brain that every skill on the SkillTree map reads from automatically.