Core Commands & Shortcuts (Quick Reference)

A collection of essential commands and shortcuts to boost your token efficiency and workflow speed in Claude Code.

---

⌨️ Must-Know Key Shortcuts

Shortcut	Action
Shift + Tab	Toggle Plan ↔ Edit mode — Essential when you want to plan before making direct code changes
ESC once	Immediately interrupt the current operation
ESC twice	Rewind — Returns to the previous prompt state

---

⚡ Running Bash Commands in Terminal (!)

You don't need to open a new terminal to check git status or start a server inside Claude Code. Prefix any command with ! to execute it as a local Bash command immediately.

!git status        → Check changed files
!git diff          → View detailed changes
!npm run dev       → Start local server
!npm run test      → Run tests

---

💬 Essential Slash Commands

Session & Context Management

Command	Description
/init	Auto-generates CLAUDE.md, your project's instruction file (required during initial project setup)
/clear	Completely resets the accumulated conversation context
/compact	Compresses context by keeping only key summaries to save tokens
/context	Visually displays current token usage in the session as a bar chart

💡 Running /context shows token usage broken down by System Prompt, System Tools, and User/Assistant. If System Tools is consuming a lot after connecting MCP, consider deactivating it. → MCP Token Management

Memory & Settings

Command	Description
/memory	Displays the contents of currently loaded memory (CLAUDE.md, MEMORY.md, etc.)
/status	Shows a comprehensive overview of your logged-in account, subscription plan, and connected MCP server status
/allowed-tools	View and manage the list of tools permitted for the AI (file writing, local command execution, etc.)

Model & Output

Command	Description
/model	Switch models based on task complexity to optimize token usage
/output-style	Change the AI's response style

/model Selection Guide:

Model	Best For
Opus	Complex architecture design, mysterious bug resolution, critical decision-making
Sonnet	General coding tasks, delegating to parallel sub-agents (default)
Haiku	File exploration, simple renaming, quick questions

/output-style Options:

Style	Description	Recommended When
auto	Automatically adjusts to the situation (default)	General use
concise	Short and to the point	Fast iterative work, code edits
verbose	Includes detailed explanations	Learning purposes, unfamiliar codebases
explanatory	Explains reasons behind code changes	Learning to code, code reviews
warning	Displays strong warnings for risky operations	Safety net for production work

Voice Input

Command	Description
/voice	Enable/disable voice input mode (hold Space and speak)
/config	Settings menu (select Korean under Language for Korean voice recognition)

Agents & Automation

Command	Description
/agents	View and manage the list of sub-agents defined in the current project
/mcp	View connected MCP servers and enable/disable them

Export

Command	Description
/export	Extracts the current session context so it can be imported into other chatbots (ChatGPT, Gemini, etc.)
/export --clipboard	Copies directly to clipboard

---

🛠️ Creating Your Own Custom Slash Commands

You can modularize frequently used pipelines as slash commands.

1. Create a .claude/commands/ folder inside your project.
2. Create a file with any name you want, such as ship.md, and write the workflow to execute.
3. After restarting Claude Code, typing /ship will automatically run that pipeline.

→ See Custom Slash Commands for details