Overview

Claude Code is Anthropic's official CLI AI coding assistant powered by Claude, capable of understanding project context, generating code, explaining code, and performing Git operations.

Installation

# macOS/Linux
npm install -g @anthropic-ai/claude-code

# Verify
claude --version

Permission Configuration

Claude Code requires explicit permissions for sensitive operations:

Permission Levels

Operation	Default	Description
Read files	Auto	Can read all files
Write files	Confirm	User confirmation required
Execute commands	Confirm	Terminal commands need confirmation
Git operations	Confirm	commit/push require confirmation

Configuration

# Allow all commands
claude --dangerously-permit-commands all

# Read-only mode
claude --read-only

Core Commands

/help - Help

claude
/help  # Show all available commands

/lsp - Code Understanding

/lsp find readFile
/lsp explain src/utils.js

/web - Web Search

/web how to use React hooks

/commit - Git Commit

/commit  # Auto-analyze changes and generate commit message

MCP Integration

Claude Code supports Model Context Protocol for external tools:

// ~/.claude/settings.json
{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@anthropic/mcp-server-filesystem", "/path/to/dir"]
    }
  }
}

Best Practices

1. CLAUDE.md: Create CLAUDE.md in project root for project specs
2. Permission Control: Use read-only mode in production
3. Context Management: Use /clear for large projects

References

• Claude Code Official Documentation
• Claude Code GitHub