A ready-to-use
.claude/folder +CLAUDE.mdfor your project. Clone, copy, edit.
Part of agent-anatomy — boilerplates for every AI coding agent.
git clone https://github.com/agent-anatomy/claude.git .claude-boilerplate
cp .claude-boilerplate/CLAUDE.md ./CLAUDE.md
cp .claude-boilerplate/CLAUDE.local.md.example ./CLAUDE.local.md
cp -r .claude-boilerplate/.claude ./.claude
rm -rf .claude-boilerplateThen edit CLAUDE.md and .claude/ to match your project.
.claude/
├── settings.json ← default permissions (edit as needed)
├── settings.local.json.example → copy → settings.local.json (gitignored)
├── commands/
│ ├── review.md → /project:review
│ ├── fix-issue.md → /project:fix-issue
│ └── deploy.md → /project:deploy
├── rules/
│ ├── code-style.md ← coding conventions
│ ├── testing.md ← test rules
│ └── api-conventions.md ← API patterns
└── agents/
├── code-reviewer.md ← strict review persona
└── security-auditor.md ← security audit persona
CLAUDE.md ← main instruction file (edit this first)
CLAUDE.local.md.example → copy → CLAUDE.local.md (gitignored)
CLAUDE.md is the primary instruction file for Claude Code. Claude reads it automatically at the start of every session. It tells Claude about your project — the stack, commands, conventions, and rules it must follow.
Think of it as a README for Claude, not for humans.
What to put in CLAUDE.md:
- Project overview (2-3 sentences)
- Build, test, lint commands
- Coding conventions
- What Claude should never touch
- Domain-specific context not obvious from the code
The .claude/ folder is the control center for Claude Code in your project. It contains:
| Path | Purpose |
|---|---|
settings.json |
Which tools Claude can run without asking |
commands/ |
Custom /project: slash commands |
rules/ |
Modular instruction files |
agents/ |
Subagent personas for specialized tasks |
| File | Commit? |
|---|---|
CLAUDE.md |
✅ Yes |
CLAUDE.local.md |
❌ No — personal |
.claude/settings.json |
✅ Yes |
.claude/settings.local.json |
❌ No — personal |
.claude/commands/ |
✅ Yes |
.claude/rules/ |
✅ Yes |
.claude/agents/ |
✅ Yes |
Add to your .gitignore:
CLAUDE.local.md
.claude/settings.local.json
Does Claude Code read CLAUDE.md automatically? Yes. Every session, without you doing anything.
Can I have multiple CLAUDE.md files? Yes. One at the root, plus one in any subdirectory. Claude merges them — subdirectory rules apply when working in that subtree.
What's the difference between CLAUDE.md and .claude/rules/?
CLAUDE.md is the main file — start here. rules/ is for splitting large instruction sets into topic files that Claude also loads.
Should I commit .claude/settings.json? Yes. It defines which tools Claude can run in your project. Commit it so the whole team gets the same defaults.
What are Claude Code slash commands?
Commands you invoke with /project:name. Each .md file in .claude/commands/ becomes one. The file content is the prompt Claude receives.
| Agent | Boilerplate |
|---|---|
| Cursor | agent-anatomy/cursor |
| OpenAI Codex | agent-anatomy/codex |
| Windsurf | agent-anatomy/windsurf |
| GitHub Copilot | agent-anatomy/copilot |
| Aider | agent-anatomy/aider |
| Gemini CLI | agent-anatomy/gemini |