Step 2 — Getting started with Claude Code

Five minutes if you're CLI-fluent. Otherwise do getting-started steps 1–2 first.

Install

curl -fsSL https://claude.ai/install.sh | sh
# or
npm install -g @anthropic-ai/claude-code

claude --version to verify.

First run

cd my-project
claude

Talk in natural language:

> show project structure
> add a usage section to README.md
> replace console.log with logger.info in all .ts files

CLAUDE.md — project rules

A CLAUDE.md at the repo root is auto-loaded each session.

# Project rules

## Never
- console.log in new code — use logger.info
- AI-credit commits (Co-Authored-By, 🤖, AI-assisted)

## Package managers
- frontend: pnpm
- python: uv
- java: gradle wrapper

## Language
- Korean for chat, docs, commits

One file replaces a thousand reminders.

Slash commands

/help    — help
/init    — generate CLAUDE.md
/cost    — token usage
/loop    — looped runs
/commit  — commit guide

Memory — persisted across sessions

Drop notes into memory/ as markdown — auto-read next session.

Try it

In a small project, run claude and type "list .md files in this folder". It auto-calls the LS tool.

Going deeper

Step 3 introduces MCP — the bridge to external systems.