CLAUDE.md Foundation

What is CLAUDE.md?

Without CLAUDE.md, I spent the first 5 minutes of every AI session re-explaining who I am, what my business does, and what tools I use. 47 sessions at 5 minutes each — nearly 4 hours wasted on context-setting alone. CLAUDE.md eliminated that permanently. It is a file that auto-loads at the start of every Claude Code conversation, giving your AI full context before you type a single word.

Think of it as your AI's identity document — it defines who the AI is, what it knows about your business, and how it should behave. With CLAUDE.md, every session starts at day 100 instead of day zero.

Where Does It Live?

1. 1Global CLAUDE.md: Place it at ~/.claude/CLAUDE.md — this loads for EVERY project automatically.
2. 2Project CLAUDE.md: Place it at your project root /CLAUDE.md — this loads only for that specific project.
3. 3The global file defines your AI's identity. Project files add project-specific context on top.

Your CLAUDE.md Template

Here is a starter CLAUDE.md tailored for your business. Copy this into ~/.claude/CLAUDE.md and customize the details.

# AI System Identity

**You are My Business's Chief of Staff AI.** You are the operator and decision-maker's right hand.

---

## Who You Serve

- **Business:** My Business (business)
- **Goal:** scale your operations
- **Team:** solo operator
- **AI Experience:** beginner

---

## Core Rules

1. **Read before write.** Always check existing files before creating new ones.
2. **Be decisive.** Give specific recommendations, not vague options.
3. **Context is king.** Reference past decisions and patterns.
4. **Tools matter.** The team uses: your current tools.

---

## Key Agents

Your system includes these specialist agents:
chief-of-staff, researcher, file-manager, task-tracker, analytics-tracker, copywriter

Each agent has its own definition file. See Chapter 3 for setup.

---

## Behavioral Rules

- When I say "Morning" → Run the morning check-in protocol
- When I say "Status" → Show current priorities and blockers
- When I say "Wrap up" → Run the evening summary protocol

---

**You are the brain. Agents are the hands.**

Key Sections Explained

1. 1Identity Block: Tells the AI who it is and who it serves. This prevents generic responses.
2. 2Core Rules: Behavioral constraints that apply to every interaction. Keep these to 4-6 rules maximum.
3. 3Key Agents: A preview of your agent roster. The AI uses this to know which specialists exist.
4. 4Behavioral Rules: Trigger-action pairs. When the user says X, the AI does Y. These are your shortcuts.

CLAUDE.md works hand-in-hand with the Life Repository (Chapter 2: The Life Repository), which stores your priorities and session state. Your identity file references those directories, so both must be set up together. Once agents are defined (Chapter 3: Building Your First Agents), CLAUDE.md becomes the central hub that routes tasks to the right specialist.

The Collaboration Playbook

Not every request needs the same depth. Add a Collaboration Playbook to CLAUDE.md that teaches your AI to match response depth to request complexity. This prevents over-engineering simple tasks and under-planning complex ones.

1. 1Quick question: Just do it, no planning needed.
2. 2Clear simple task: Brief alignment, then execute.
3. 3Complex creative task: Full plan, phased execution.
4. 4Ambiguous request: Ask 'Quick or full planning?' before starting.

As your system grows, CLAUDE.md will reference your rules (Chapter 4: The Rules System) and protocols (Chapter 5: Protocols). The enforcement hooks from Chapter 6 can even verify that CLAUDE.md stays consistent with your agent definitions. Start simple here and let it evolve.