Claude Code Quick Reference: Commands & Shortcuts

Updated October 21, 2026: Refreshed for Claude Code 2.1.305: current shortcuts, slash commands and aliases (/review → /code-review, /cost → /usage), CLI commands (claude auth, claude purge), permission-mode flags, permission-rule syntax, AGENTS.md, and the Opus 5.5 / Sonnet 5.5 / Haiku 5.5 / Fable 5.1 lineup.

Keep this page bookmarked. Everything you need to use Claude Code efficiently, organized for quick lookup. For the complete list of commands, flags, settings, and environment variables, see the Claude Code CLI Reference.

Keyboard Shortcuts

Essential Shortcuts

Shortcut	Action
`Escape`	Interrupt Claude (stops the current response or tool call)
`Ctrl+C`	Interrupt, or clear input; press again on an empty prompt to exit
`Ctrl+L`	Redraw the screen (keeps the conversation)
`Up/Down`	Navigate command history
`Tab`	Accept an autocomplete suggestion
`Shift+Enter`	Multi-line input (`\` + `Enter` or `Ctrl+J` work in any terminal)
`Shift+Tab`	Cycle permission modes (Manual → acceptEdits → plan → auto)
`Esc` `Esc`	Clear the draft, or open the rewind menu on an empty prompt
`Ctrl+R`	Reverse-search command history

macOS Specific

Shortcut	Action
`Opt+P`	Switch model
`Opt+T`	Toggle extended thinking (no effect on Opus 5.5, Sonnet 5.5, Haiku 5.5, or Fable, which always think)
`Opt+O`	Toggle fast mode

Most Option shortcuts, including Opt+P, need Option set as Meta in your terminal. Cmd+K is not a Claude Code “clear conversation” key; use /clear.

Windows/Linux Specific

Shortcut	Action
`Alt+P`	Switch model
`Alt+T`	Toggle extended thinking
`Alt+O`	Toggle fast mode

Ctrl+K deletes to the end of the line; it does not clear the conversation.

Slash Commands

Built-in Commands

Command	Description
`/help`	Show all available commands
`/clear`	Start a new conversation with empty context (aliases: `/reset`, `/new`)
`/compact`	Summarize the conversation to free context
`/context`	Show what is filling the context window
`/init`	Initialize CLAUDE.md for project
`/model`	Switch model (saved as your default for new sessions)
`/effort`	Set effort (`low`–`xhigh`, `max`, `auto`); `/effort ultracode` toggles ultracode
`/memory`	Edit CLAUDE.md files and manage auto memory
`/usage`	Session cost, plan limits, and stats (`/cost` and `/stats` are aliases)
`/doctor`	Run a setup checkup and fix problems (alias: `/checkup`)
`/config`	Open Settings (alias: `/settings`)
`/permissions`	Manage allow, ask, and deny rules
`/bug`	Report a bug or share your conversation (alias: `/share`)

File Commands

Command	Description
`@<path>`	Mention a file or directory in your prompt (there is no `/add` command)
`/add-dir <path>`	Add a working directory for this session

Review & Version Control

Command	Description
`/code-review`	Review the current diff, or a PR number, branch, or path (`--fix`, `--comment`, `ultra`)
`/review`	Alias of `/code-review` (since v2.1.223)
`/security-review`	Check your branch’s changes for security vulnerabilities
`/diff`	Review working-tree changes, including Claude’s edits

/pr-comments was removed in v2.1.91; ask Claude to read the pull request comments instead.

CLI Flags

Starting Claude Code

# Basic
claude                    # Start interactive session
claude "prompt"           # Start a session with an initial prompt
claude -c                 # Continue the most recent conversation here
claude -r                 # Resume a session by ID or name, or open the picker
claude -n "auth-refactor" # Name the session

# Model Selection (aliases on the Anthropic API)
claude --model opus       # Opus 5.5 (the default model)
claude --model sonnet     # Sonnet 5.5
claude --model haiku      # Haiku 5.5
claude --model fable      # Fable 5.1 (never the default)
claude --effort high      # low | medium | high | xhigh | max

# Output Control
claude -p "prompt"        # Print mode (no interactive)
claude -p "prompt" --output-format json  # JSON output (text | json | stream-json)
claude --verbose          # Verbose turn-by-turn output
claude --debug            # Debug logging (or --debug-file <path>)

# Permissions
claude --permission-mode plan  # manual | acceptEdits | plan | auto | dontAsk | bypassPermissions
claude --restricted            # No command-running tools or WebFetch; ignores user/project/local settings
claude --dangerously-skip-permissions  # Skip all permission prompts (isolated containers only)

Interactive terminal sessions start in auto mode when nothing sets a permission mode (v2.1.283+); claude -p runs start in Manual mode (v2.1.302+).

MCP Management

# List and manage MCP servers
claude mcp list                                # List servers and their health
claude mcp add --transport http <name> <url>   # Add a remote HTTP server
claude mcp add <name> -- <command> [args...]   # Add a local stdio server
claude mcp get <name>                          # Inspect one server
claude mcp remove <name>                       # Remove MCP server
claude mcp reset-project-choices               # Reset approvals for .mcp.json servers

# Scoped MCP (default scope: local)
claude mcp add --scope project <name> -- <command>
claude mcp add --scope user <name> -- <command>

Configuration

# Settings (`claude config` was removed in v2.0)
/config                   # In a session: open Settings
/config model=sonnet      # In a session: set one key directly
# Files: ~/.claude/settings.json (user), .claude/settings.json (project),
#        .claude/settings.local.json (local, kept out of git)

# Login/Auth
claude auth login         # Authenticate (--console for API billing)
claude auth status        # Show auth status
claude auth logout        # Log out
claude setup-token        # Long-lived OAuth token for CI and scripts

# Maintenance
claude update             # Update Claude Code
claude doctor             # Installation and settings diagnostics
claude purge [path]       # Delete a project's local state (--dry-run, --all)

Common File Patterns

@ References

# In prompts or CLAUDE.md
See @README.md for project overview.
Check @src/lib/auth.ts for auth utilities.
Reference @package.json for dependencies.

CLAUDE.md Templates

# Minimal CLAUDE.md
## Project
[Brief description]

## Commands
- npm run dev
- npm test

## Standards
- TypeScript strict mode
- Tests required for features

# Full CLAUDE.md
## Overview
[Project description]

## Tech Stack
- [Framework]
- [Language]
- [Database]

## Directory Structure
src/
├── [folder]/  # [purpose]
└── [folder]/  # [purpose]

## Commands
- [command] - [description]

## Autonomous Operations
- [allowed actions]

## Requires Confirmation
- [restricted actions]

Permission Patterns

settings.json

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

Pattern Syntax

Pattern	Matches
`Read`	All read operations
`Edit(src/**)`	Edits in src/ and subdirs (path rules use `Edit`; `Write(path)` rules are never checked)
`Bash(npm *)`	All npm commands (`Bash(npm:*)` is the equivalent older form)
`Bash(git diff *)`	git diff with any args
`Read(.env*)`	Any .env file
`WebFetch(domain:example.com)`	Fetches to example.com

Rules are evaluated deny → ask → allow, and deny rules apply in every permission mode, including bypassPermissions.

Model Comparison

Model	Best For	Speed
Fable 5.1	Hardest problems, long autonomous tasks	Slower
Opus 5.5	Default for most work, complex reasoning	Fast; `/fast` for up to 2.5x
Sonnet 5.5	Everyday coding at lower cost	Fast
Haiku 5.5	Quick tasks, exploration, subagents	Fastest

When to Use Each

Fable 5.1: Root-cause investigations, architecture decisions, work larger than a single sitting (select it with /model fable; it is never the default and can bill to usage credits)
Opus 5.5: The default on Pro, Max, Team, Enterprise, and the Anthropic API, and Anthropic’s suggested starting point; toggle /fast when latency matters
Sonnet 5.5: Day-to-day development, feature implementation
Haiku 5.5: Quick questions, file exploration, simple edits

On the Anthropic API, opus, sonnet, haiku, and fable resolve to these models from Claude Code v2.1.280, v2.1.284, v2.1.293, and v2.1.257. Bedrock, Google Cloud’s Agent Platform, Microsoft Foundry, and Claude Platform on AWS still map some aliases to older models. Opus 5.5, Sonnet 5.5, and Haiku 5.5 default to medium effort; Fable 5.1 defaults to high.

Memory System

Memory Files

Location	Scope	Use Case
`~/.claude/CLAUDE.md`	Global	Personal preferences
`./CLAUDE.md`	Project	Team standards
`./.claude/CLAUDE.md`	Project	Team standards (hidden)
`./CLAUDE.local.md`	Personal	Individual overrides
`./AGENTS.md`	Project	Read instead when there is no CLAUDE.md (v2.1.277+)
`~/.claude/projects/<project>/memory/`	Per project	Auto memory Claude writes itself

Load Order

Files are combined, not overridden; later = more specific:
1. ~/.claude/CLAUDE.md (user)
2. ./CLAUDE.md or ./.claude/CLAUDE.md (project)
3. ./CLAUDE.local.md (local, read last)

Common Workflows

Start New Feature

claude
> Implement [feature description]
> Follow TDD - write tests first

Debug Issue

claude
> Debug: [error message]
> Check logs at [path]

Code Review

claude
> /code-review
> Focus on: security, performance, maintainability

Refactor

claude
> Refactor [component] to [pattern]
> Maintain backward compatibility

Troubleshooting Quick Fixes

Problem	Solution
CLAUDE.md ignored	Check file location (./CLAUDE.md or ./.claude/CLAUDE.md); `/context` lists the memory files that loaded
MCP not loading	`/mcp` or `claude mcp get <name>`; `claude mcp reset-project-choices`
Permission denied	Add a rule with `/permissions` or in the settings.json allow list
High token usage	Check `/context`, then `/compact` or `/clear`
Wrong model	`/model`, or `Option+P` (Mac) / `Alt+P` (Win/Linux)
Config not applied	Run `claude doctor` for settings errors, then restart Claude Code

Environment Variables

# API Configuration
ANTHROPIC_API_KEY=sk-...

# Proxy (if needed)
HTTP_PROXY=http://...
HTTPS_PROXY=https://...

# Model and tools
ANTHROPIC_MODEL=opus               # Model for new sessions
CLAUDE_CODE_EFFORT_LEVEL=high      # low | medium | high | xhigh | max | auto
CLAUDE_CODE_ENABLE_TODO_TOOLS=1    # Task/TodoWrite tools on newer models (off by default)

# Debug
DEBUG=1                            # Same as --debug

Print-Friendly Summary

CLAUDE CODE QUICK REFERENCE
===========================

SHORTCUTS
---------
Escape      Interrupt Claude
Ctrl+C      Interrupt / clear / exit
Shift+Tab   Cycle permission modes
Opt/Alt+P   Switch model

COMMANDS
--------
/help       Show commands
/clear      New conversation
/init       Create CLAUDE.md
/doctor     Diagnose issues

CLI
---
claude              Start session
claude -p "prompt"  Single prompt, then exit
claude -c           Continue session
claude --model X    Use model X

FILES
-----
./CLAUDE.md         Project config
~/.claude/CLAUDE.md Global config
CLAUDE.local.md     Personal overrides
.claude/settings.json  Permissions

MCP
---
claude mcp list     Show servers
claude mcp add      Add server
claude mcp remove   Remove server

Bookmark this page and keep it handy. You’ll reference it more than you think.

Sources: Claude Code Documentation, CLI reference, Commands, Interactive mode, Model configuration, Claude Code GitHub