Contents

What is Claude Code

Claude Code is a CLI tool from Anthropic that runs Claude right in your terminal. It sees the project files, runs commands, reads code and makes changes. It's not a copy-paste chat — the agent works in the context of your repository.

Installation

Native installer (recommended):

# macOS / Linux / WSL
curl -fsSL https://claude.ai/install.sh | bash

# Windows PowerShell
irm https://claude.ai/install.ps1 | iex

# macOS via Homebrew
brew install --cask claude-code

# Windows via WinGet
winget install Anthropic.ClaudeCode

The old method npm install -g @anthropic/claude-code is deprecated and not recommended since 2025.

Verify:

claude --version

Browser-based authentication on first run — just type:

claude

First commands

# Run in the current folder
claude

# Sprint mode: execute a task without dialog
claude -p "add ESLint to the project"

# Continue an interrupted session
claude --continue

CLAUDE.md — instructions for the agent

CLAUDE.md is a file in the project root that Claude reads on every run. Put project context, rules and commands here.

# Project

E-commerce on Next.js + PostgreSQL.

## Stack

- Next.js 15 (App Router)
- TypeScript (strict: true)
- Prisma ORM
- Tailwind CSS

## Commands

- `npm run dev` — start dev server
- `npm run test` — tests (Vitest)
- `npm run lint` — linter

## Rules

- Write in TypeScript, don't use `any`
- Components in `src/components/`, one file = one component
- Commits in Conventional Commits style
- Don't touch files in `src/generated/`

The more precise CLAUDE.md is, the less you need to explain in each session.

Settings (.claude/settings.json)

The .claude/settings.json file in the project root manages permissions:

{
  "permissions": {
    "allow": [
      "Bash(npm run *)",
      "Bash(git *)",
      "Bash(npx *)"
    ],
    "deny": [
      "Bash(rm -rf *)"
    ]
  }
}

Global settings live in ~/.claude/settings.json — they apply to all projects.

Modes of operation

Mode Command What it does
Dialog claude Interactive chat with the agent
Auto mode claude --dangerously-skip-permissions No confirmations (use carefully)
One-shot claude -p "task" Execute and exit
Continue claude --continue Resume the last session
Bare claude --bare -p "task" Fast mode for scripts — skips hooks/skills/MCP/CLAUDE.md

Useful hotkeys in a session

Combo Action
Ctrl+C Interrupt current operation
Ctrl+D Exit Claude Code
/help List of all commands
/clear Clear session context
/compact Compress history to save tokens
/memory Edit CLAUDE.md memory files
/doctor Installation diagnostics
/rewind Roll back to a previous checkpoint

IDE and Web

Claude Code works not only in the terminal:

  • VS Code — extension with inline diff and a change plan
  • JetBrains — plugin for IntelliJ, PyCharm, WebStorm, Rider
  • claude.ai/code — web version with GitHub repository integration

Summary

  • Install: curl -fsSL https://claude.ai/install.sh | bash
  • Configure CLAUDE.md with project description and commands
  • Restrict permissions via .claude/settings.json
  • Use /help inside a session — lots of useful stuff there

Next in the series: Skills — your own commands for the agent and MCP — connecting external tools. To see it on a concrete stack, check Claude Code for .NET + Angular.

Share: