Claude Code Complete Guide

Requirements

Node.js 18+
Anthropic account (Claude Pro, Max, Teams, or API)

Installation

CLI (Terminal)

npm install -g @anthropic-ai/claude-code

VS Code Extension (Beta) ⭐ NEW

code --install-extension anthropic.claude-code

Or search "Claude Code" in VS Code Extensions (Anthropic publisher).

VS Code Extension Features:

Native sidebar panel (Spark icon ✨)
Plan mode with editing
Auto-accept edits mode
File @-mentions and image support
Multiple simultaneous sessions
Real-time inline diffs

Requires VS Code 1.98.0+

First Run

claude

Skip Permission Prompts (Optional)

Method 1: PowerShell Profile (Recommended)

notepad $PROFILE

# Add:
function claude { & "C:\Users\$env:USERNAME\AppData\Roaming\npm\claude.cmd" --dangerously-skip-permissions $args }

# Save, restart terminal

Method 2: Run with Flag

claude --dangerously-skip-permissions

Built-in Slash Commands

Command	Description
`/help`	Show all commands
`/config`	Open settings
`/permissions`	Manage permissions
`/context`	Check context usage
`/compact`	Compress conversation
`/clear`	Clear conversation
`/plugin`	Manage plugins
`/mcp`	Manage MCP servers
`/model`	Switch models
`/cost`	Show session cost
`/memory`	View/edit memory
`/ide`	Connect to IDE
`/rewind`	Go back to previous state

Memory & Context

Memory File

Claude Code stores learned preferences in memory. Access with /memory.

# Add to memory during chat
# Use # at start of message to save to memory
# my preferred language is TypeScript

Model Selection

# Switch models during session
/model

# Available models (Claude 4.5 Series):
# - claude-sonnet-4-5 (default) - Best coding model
# - claude-opus-4-5 - Most capable, complex reasoning
# - claude-haiku-4-5 - Fastest, cost-effective

Model Comparison:

Model	Best For	Speed
Sonnet 4.5	Coding, agents, daily tasks	Fast
Opus 4.5	Complex reasoning, research	Moderate
Haiku 4.5	Quick tasks, cost-sensitive	Fastest

Extended Thinking (Hybrid Reasoning)

All Claude 4.5 models support hybrid reasoning with two modes:

Instant responses - Quick answers for simple tasks
Extended thinking - Deep reasoning for complex problems

Add "ultrathink" anywhere in a skill to enable deep reasoning:

---
name: complex-analysis
description: Deep code analysis with ultrathink
---
Perform thorough analysis using extended thinking...

CLI Flags

Flag	Description
`--version`	Check version
`--resume`	Resume last session
`-p "prompt"`	Run with prompt (headless)
`--output-format json`	JSON output
`--mcp-debug`	Debug MCP issues
`--dangerously-skip-permissions`	Bypass prompts

CLAUDE.md (Project Memory)

Create CLAUDE.md in your project root:

# Project: MyApp

## Tech Stack
- Next.js 14, TypeScript, TailwindCSS
- PostgreSQL with Prisma
- Deployed on Vercel

## Coding Standards
- TypeScript strict mode
- ESLint + Prettier
- Tests required for all functions

## Commands
- `npm run dev` - Start dev server
- `npm run test` - Run tests
- `npm run lint` - Lint code

Custom Slash Commands

Create .claude/commands/your-command.md:

---
description: Fix a GitHub issue
---
Analyze and fix GitHub issue: $ARGUMENTS

1. Use `gh issue view $ARGUMENTS` to get details
2. Search codebase for relevant files
3. Implement fix
4. Write tests
5. Commit changes

Usage: /your-command 123

Skills (Auto-Invoked)

Create .claude/skills/my-skill/SKILL.md:

---
name: code-review
description: Review code for best practices
---
When reviewing code:
1. Check for security issues
2. Verify error handling
3. Ensure proper typing
4. Suggest improvements

Skills auto-activate when description matches the task.

Subagents

Built-in agents for parallel/isolated work:

Agent	Purpose	Context
`Explore`	Read-only codebase exploration	Isolated
`Plan`	Create implementation plans	Isolated
`general-purpose`	Full tool access	Forked

Using Subagents in Skills

---
name: research
description: Deep research
context: fork
agent: Explore
---
Research $ARGUMENTS thoroughly...

Custom Agent Definition

Create in .claude/agents/my-agent.md:

---
name: security-auditor
description: Security-focused code analysis
tools: Read, Glob, Grep
context: fork
---
Audit code for security vulnerabilities:
1. SQL injection patterns
2. XSS vulnerabilities
3. Authentication flaws
4. Secrets exposure

Output as security report with severity ratings.

See Agent Collections for 100+ ready-made agents.

MCP Servers (External Tools)

# Add MCP server (HTTP transport - new syntax)
claude mcp add --transport http github https://api.githubcopilot.com/mcp/

# Add MCP server (NPX - new syntax)
claude mcp add playwright -- npx @playwright/mcp@latest

# List servers
claude mcp list

# Remove server
claude mcp remove github

# Debug MCP issues
claude --mcp-debug

Popular MCP Servers

Official (HTTP Transport):

claude mcp add --transport http github https://api.githubcopilot.com/mcp/
claude mcp add --transport http notion https://mcp.notion.com/mcp
claude mcp add --transport http sentry https://mcp.sentry.dev/mcp

NPX-Based:

claude mcp add playwright -- npx @playwright/mcp@latest
claude mcp add memory -- npx -y @modelcontextprotocol/server-memory
claude mcp add filesystem -- npx -y @modelcontextprotocol/server-filesystem "/path"
claude mcp add db -- npx -y @bytebase/dbhub --dsn "postgresql://..."

Registry: https://registry.modelcontextprotocol.io

Plugins

# Install plugin
/plugin install https://github.com/user/plugin

# List plugins
/plugin list

# Remove plugin
/plugin remove plugin-name

Hooks (Automation)

Hooks are configured in .claude/settings.json using the event-based system:

{
  "hooks": {
    "PostToolUse": [
      {
        "matcher": { "tool": "Write", "path": "*.ts" },
        "command": "npx eslint --fix $TOOL_INPUT_PATH"
      }
    ],
    "PreToolUse": [
      {
        "matcher": { "tool": "Bash", "command": "git commit*" },
        "command": "npm run lint && npm test"
      }
    ]
  }
}

Hook Events

Event	When Triggered
`UserPromptSubmit`	Before prompt sent
`PreToolUse`	Before tool runs
`PostToolUse`	After tool completes
`SessionStart`	Session begins
`Stop`	Claude stops

See Claude Code Resources for more examples.

Keyboard Shortcuts

Key	Action
`Ctrl+C`	Cancel current action
`Ctrl+B`	Send agent to background
`Esc`	Exit
`Tab`	Autocomplete
`↑/↓`	History navigation

Headless Mode (CI/Automation)

# Run single task
claude -p "fix all lint errors" --output-format json

# In CI pipeline
claude -p "run tests and report" --dangerously-skip-permissions

Best Practices

Start with CLAUDE.md - Give project context
Use /compact - When context gets large
Create slash commands - For repeated workflows
Use subagents - For parallel/isolated tasks
Check /context - Monitor token usage
Use MCP - For external integrations

Update

npm update -g @anthropic-ai/claude-code

Verify

claude --version

Resources

Docs: https://docs.anthropic.com/claude-code
Best Practices: https://anthropic.com/engineering/claude-code-best-practices
Awesome List: https://github.com/hesreallyhim/awesome-claude-code

Settings Location

Claude Code settings are stored in:

# Global settings
~/.claude/settings.json

# Project settings
.claude/settings.json

Example settings.json:

{
  "permissions": {
    "allow": ["Read", "Write", "Bash(npm *)"],
    "deny": ["Bash(rm -rf *)"]
  },
  "env": {
    "NODE_ENV": "development"
  }
}

Troubleshooting

Claude Code Not Responding

Check internet connection
Verify login: claude --version
Clear cache: rm -rf ~/.claude/cache
Restart terminal

Extension Not Working in VS Code

Ensure VS Code 1.98.0+
Reload window: Ctrl+Shift+P → "Reload Window"
Check for extension updates

Permission Denied Errors

# Grant permissions
/permissions

# Or use flag
claude --dangerously-skip-permissions

MCP Server Issues

# Debug mode
claude --mcp-debug

# Check server status
claude mcp list

High Token Usage

Use /compact to compress history
Check /context for usage
Start fresh with /clear

FilesExpand file tree

claude-code-guide.md

Latest commit

History