The Complete Guide to Useful Features and Shortcuts for Claude Code

Introduction

Claude Code is a CLI-based AI coding assistant provided by Anthropic. It runs directly in your terminal, allowing the AI to autonomously handle file reading and writing, Git operations, code generation, debugging, and more.

However, with such a rich feature set, you might often find yourself thinking, "Wait, I could do that?" This article provides a comprehensive guide to mastering Claude Code, covering keyboard shortcuts, slash commands, CLI flags, as well as configuration and extension features.

---

Keyboard Shortcuts

Basic Operations

Shortcut	Function
Ctrl+C	Cancel current input or generation
Ctrl+D	Exit session
Ctrl+L	Clear terminal screen (preserves chat history)
Ctrl+O	Toggle verbose output
Ctrl+R	Reverse search command history
Ctrl+G	Open input in default text editor
Ctrl+B	Send current task to background
Ctrl+F	Terminate all background agents
Ctrl+T	Show/hide task list
Shift+Tab	Toggle permission mode

Model & Thinking Modes

Shortcut	Function
Alt+P / Option+P	Switch models
Alt+T / Option+T	Toggle Extended Thinking

Pasting Images

Shortcut	Function
Ctrl+V / Cmd+V / Alt+V	Paste image from clipboard

This is extremely useful when you paste a screenshot and ask, "Fix this UI bug."

Multi-line Input

Shortcut	Supported Terminals
\ + Enter	All terminals
Option+Enter	macOS default
Shift+Enter	iTerm2, WezTerm, Ghostty, Kitty
Ctrl+J	All terminals

Text Editing (Emacs style)

Shortcut	Function
Ctrl+A	Move to start of line
Ctrl+E	Move to end of line
Ctrl+K	Delete to end of line
Ctrl+U	Delete entire line
Ctrl+Y	Paste deleted text
Alt+B	Move back one word
Alt+F	Move forward one word

Rewind

Shortcut	Function
Esc × 2 times	Rewind to previous state or summarize conversation

If Claude makes an unintended change, pressing Esc twice will undo that operation. It's a subtle but essential shortcut.

---

Slash Commands

Session Management

Command	Function
/clear	Clear chat history (also works with /reset or /new)
/resume [session]	Resume a session (also works with /continue)
/fork [name]	Fork the conversation at the current point
/rewind	Go back to the previous checkpoint
/rename [name]	Rename the session
/export [filename]	Export conversation to a text file

Context Management

Command	Function
/context	Display context usage grid
/compact [instruction]	Compact conversation (optional: specify focus)
/cost	Show token usage statistics
/memory	Edit the CLAUDE.md memory file

Context Management Tip: Use /compact regularly during long sessions to save your context window. If you specify a focus, such as /compact Focus on TypeScript type errors, it will prioritize keeping related information.

Settings & Diagnostics

Command	Function
/config	Open settings screen (also works with /settings)
/doctor	Diagnose installation and settings
/help	Show help
/usage	Show plan usage limits
/theme	Change color theme

Git & Code Review

Command	Function
/diff	Interactive diff viewer
/pr-comments [PR]	Get comments from a GitHub PR
/security-review	Review security vulnerabilities

Extensions

Command	Function
/skills	List available skills
/mcp	Manage MCP servers
/batch <instruction>	Execute large-scale changes in parallel
/hooks	Manage hook settings

Other

Command	Function
`/fast [on	off]`
/vim	Toggle Vim editing mode
/copy	Copy last response to clipboard
/tasks	Manage background tasks

---

CLI Flags (Launch Options)

Basic Usage

# Interactive mode
claude

# Launch with an initial prompt
claude "Explain the structure of this project"

# Continue the latest conversation
claude -c

# One-shot execution (run and exit)
claude -p "List package.json dependencies"

Model Selection

claude --model claude-opus-4-6       # Use Opus 4.6
claude --model claude-sonnet-4-6     # Use Sonnet 4.6
claude --fast                        # Launch in fast mode

Permission Control

# Launch in Plan mode (read-only)
claude --permission-mode plan

# Limit available tools
claude --tools "Bash,Edit,Read"

# Allow specific commands only
claude --allowedTools "Bash(npm run *)" "Read"

# Disallow specific commands
claude --disallowedTools "Bash(rm *)"

System Prompt

# Custom system prompt
claude --system-prompt "You are a Python expert"

# Read system prompt from a file
claude --system-prompt-file ./custom-prompt.txt

# Append to existing system prompt
claude --append-system-prompt "Always use TypeScript"

Output Format (For Automation)

# JSON output
claude -p --output-format json "query"

# Streaming JSON
claude -p --output-format stream-json "query"

# Structured output (with schema)
claude -p --json-schema '{"type":"object"...}' "query"

Cost Management

# Set maximum budget (USD)
claude -p --max-budget-usd 5.00 "Large-scale refactoring"

# Limit turn count
claude -p --max-turns 3 "Simple question"

Worktree (Parallel Work)

# Execute within an isolated git worktree
claude -w feature-auth

# Create a worktree with an auto-generated branch name
claude -w

There is also a --dangerously-skip-permissions flag, but as the name suggests, it is dangerous. Only use it in trusted environments such as CI/CD pipelines.

---

Configuration File

Configuration File Precedence

Claude Code settings are managed across three scopes:

File	Scope	Purpose
~/.claude/settings.json	User	Personal settings shared across all projects
.claude/settings.json	Project	Team-shared settings (Git managed)
.claude/settings.local.json	Local	Personal project-specific settings (recommended in .gitignore)

Permission Settings

{
  "permissions": {
    "allow": [
      "Bash(npm run test *)",
      "Read(~/.zshrc)"
    ],
    "deny": [
      "Bash(rm *)",
      "Edit(**/.env*)"
    ],
    "defaultMode": "acceptEdits"
  }
}

There are four permission modes:

Mode	Description
plan	Read-only. No code changes are made
acceptEdits	Auto-approve file edits, confirm command execution
bypassPermissions	Auto-approve everything (dangerous)
default	Confirm each time

Status Line

{
  "statusLine": {
    "type": "command",
    "command": "~/.claude/statusline.sh"
  }
}

You can display context usage or Git branch information in the status line.

---

CLAUDE.md ── Project Memory

CLAUDE.md is a file for providing persistent instructions to Claude Code. It is automatically read at the start of every session.

Location

Location	Scope
CLAUDE.md (project root)	Entire project
.claude/CLAUDE.md	Entire project
~/.claude/CLAUDE.md	User-wide (all projects)

What to Write

# CLAUDE.md

## Basic Rules
- Respond in Japanese
- Git commit messages should be in English

## Environment
- Node.js: v24.13.0 (managed by Volta)
- Package manager: pnpm

## Coding Conventions
- Use TypeScript strict mode
- Write tests with Vitest

Modular Rule Management

In large projects, you can manage rules modularly in the .claude/rules/ directory:

.claude/rules/
  ├── git.md           # Git rules
  ├── typescript.md    # TypeScript conventions
  └── security.md      # Security rules

---

Hooks ── Automated Workflows

Hooks are processes executed automatically in response to Claude Code events.

Main Events

Event	Timing
Stop	When Claude finishes responding
PreToolUse	Before tool execution (can be blocked)
PostToolUse	After successful tool execution
Notification	When sending a notification
UserPromptSubmit	When user submits a prompt
SessionStart	When a session starts

Implementation Example: Play sound on completion

{
  "hooks": {
    "Stop": [
      {
        "matcher": "",
        "hooks": [
          {
            "type": "command",
            "command": "afplay /System/Library/Sounds/Glass.aiff"
          }
        ]
      }
    ]
  }
}

Implementation Example: Auto-format on file save

{
  "hooks": {
    "PostToolUse": [
      {
        "matcher": "Edit|Write",
        "hooks": [
          {
            "type": "command",
            "command": "npx prettier --write $CLAUDE_FILE_PATH"
          }
        ]
      }
    ]
  }
}

---

Skills ── Reusable Instruction Sets

Skills provide a mechanism to package instructions specialized for specific tasks.

Creating a Skill

mkdir -p ~/.claude/skills/my-skill

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

---
name: my-skill
description: Description of the skill
---

Write detailed instructions for the skill here...

Invoking a Skill

/my-skill arg1 arg2

For example, if you prepare an article writing skill, you can automate everything from generating an article template to Git pushing with a single command like /zenn-article-writer article-title.

---

MCP ── External Tool Integration

MCP (Model Context Protocol) is a protocol for connecting Claude Code with external services and tools.

Adding a Server

# HTTP-based server
claude mcp add --transport http notion https://mcp.notion.com/mcp

# Local stdio-based server
claude mcp add --transport stdio airtable \
  --env AIRTABLE_API_KEY=YOUR_KEY \
  -- npx -y airtable-mcp-server

Management Commands

claude mcp list       # List servers
claude mcp get github # Get details
claude mcp remove github  # Remove

Popular MCP Servers

Server	Purpose
GitHub	PR/Issue management
Sentry	Error monitoring
Notion	Document management
Slack	Messaging
PostgreSQL	Database queries

---

Power User Techniques

1. Execute commands instantly with ! prefix

! npm test
! git status

By typing ! followed by a command in the prompt input, you can run terminal commands directly without involving Claude.

2. Leverage background tasks

Use Ctrl+B to push tasks to the background. This is useful when you want to ask another question while running tests.

3. Timing for /compact

Consider running /compact when context exceeds 50%. You can check usage at any time with /context.

4. Parallel development with Worktrees

claude -w feature-login   # Develop login feature in a separate branch
claude -w fix-bug-123     # Work on bug fixes in parallel

You can develop multiple features simultaneously without polluting your main branch.

5. Accept suggestions with Tab key

Grayed-out command suggestions may appear when starting a session. You can accept them directly with the Tab key.

6. Diagnose issues with /doctor

If you feel something is "not working," run /doctor. It will automatically detect configuration errors and environmental issues.

---

Troubleshooting

Common Issues and Solutions

Issue	Solution
Insufficient context	Use /compact to compress or /clear to start fresh
Cannot connect to MCP server	Check auth state via /mcp or diagnose with /doctor
Permissions asked every time	Add to allow in settings.json
File changes not reflected	Roll back with Esc × 2 and try again
Slow response	Switch to fast mode with /fast on

---

Summary

Category	Key Items
Shortcuts	Esc×2 (rewind), Ctrl+B (background), Shift+Tab (toggle permissions)
Commands	/compact (compress), /context (usage), /doctor (diagnose)
CLI Flags	-c (continue), -p (one-shot), -w (worktree)
Settings	CLAUDE.md (persistent instructions), Hooks (automation), Skills (reuse)
Extensions	Integrate with GitHub, Slack, DB, etc., via MCP

Claude Code is more than just an AI chat; it is a development partner residing in your terminal. Mastering these shortcuts and settings will drastically improve your development efficiency.

Start by checking your context usage with /context.