Getting Started

Go from zero to productive with Claude Code in 10 minutes.

---

Step 0: Create a Claude Account

If you don’t have one yet, sign up at claude.ai and choose a plan. See the Account Setup guide for plan comparisons and recommendations.

---

Step 1: Install the Playbook (2 minutes)

Option A — Interactive installer:

curl -sL https://raw.githubusercontent.com/ao92265/claude-code-playbook/main/scripts/install.sh | bash

Option B — Manual setup:

git clone https://github.com/ao92265/claude-code-playbook.git
cd claude-code-playbook

# Copy a template to your project
cp templates/CLAUDE.md /path/to/your/project/CLAUDE.md

# Install skills globally
cp -r skills/check-env ~/.claude/skills/
cp -r skills/deploy ~/.claude/skills/
cp -r skills/handoff ~/.claude/skills/

# Install hooks
mkdir -p ~/.claude/hooks
cp hooks/ts-check.sh ~/.claude/hooks/
chmod +x ~/.claude/hooks/ts-check.sh

Step 2: Customize Your CLAUDE.md (3 minutes)

Open the CLAUDE.md you just copied and edit each section:

1. Project Basics — Set your language, framework, and module system
2. Testing — Specify your test runner and testing conventions
3. Verification — Define what “done” means (which commands to run)
4. Git & GitHub — Keep the safety rules as-is

Delete sections that don’t apply. Every line should prevent a real problem.

Step 3: Configure Hooks (2 minutes)

Add to ~/.claude/settings.json:

{
  "hooks": {
    "PostToolUse": [
      {
        "matcher": "Edit|Write",
        "hooks": [
          {
            "type": "command",
            "command": "~/.claude/hooks/ts-check.sh",
            "timeout": 30,
            "statusMessage": "Checking TypeScript types..."
          }
        ]
      }
    ]
  }
}

For the complete 7-hook setup, see config/hooks-example.json.

Step 4: Your First Session (3 minutes)

Start Claude Code and try:

> /check-env

This validates your environment — ports, Docker, credentials, Node.js. Fix any issues it finds.

Then try a real task:

> Fix the failing test in src/auth/login.test.ts.
  The error is: "Expected 200, received 401".
  Only modify files in src/auth/. Don't change tests.

Notice the pattern: specific error + scope constraint + clear goal.

---

What to Learn Next

Goal	Resource
Master the core workflow	Power User Guide
Improve your prompts	Prompt Patterns
Avoid common mistakes	Anti-Patterns
Choose the right skill	Workflow Decision Tree
Pick the right model	Model Comparison
Set up MCP servers	MCP Servers Guide
See real examples	Example Sessions

---

Working in Large Codebases

Three small habits make Claude Code dramatically more useful in monorepos and large polyglot repos. None require new tooling.

1. Start Claude in a subdirectory, not the repo root.

Claude Code walks up the directory tree automatically to find CLAUDE.md and project context. Starting at the root pulls every nested CLAUDE.md, every skills folder, and every irrelevant doc into your initial context. Starting inside the area you actually intend to change scopes the context to what is relevant.

# Bad — root of a 40-package monorepo
cd ~/repos/big-monorepo && claude

# Good — the package you're actually working in
cd ~/repos/big-monorepo/packages/payments && claude

2. Add a .claudeignore file at the repo root.

Same role as .gitignore, but for what Claude reads when exploring the codebase. Exclude generated code, build artifacts, vendored dependencies, fixtures, and snapshots. A minimal example:

node_modules/
dist/
build/
.next/
coverage/
**/__snapshots__/
**/generated/
vendor/

The smaller the scannable surface, the more accurate Grep/Glob answers will be — and the less likely Claude is to confidently quote a stale generated file.

3. Prefer LSP over grep for symbol navigation.

If you have LSP tools (any MCP server exposing goto_definition / find_references / workspace_symbols), use them for symbol lookups in typed codebases. grep "handleAuth" returns every comment, log line, and unrelated function with the same name. lsp_find_references returns only the actual call sites — which is what you almost always want.

---

The 5 Rules That Matter Most

If you remember nothing else from this playbook:

1. One task per session. Multi-task sessions degrade quality.
2. Paste real errors. Don’t describe — paste the stack trace.
3. Scope-lock your prompts. “Only modify src/auth/” prevents drift.
4. Verify with real tests. “Looks correct” ≠ “passes tests”.
5. Start fresh above 80% context. Use /handoff to preserve context for the next session.