Master Claude Code: 33 Essential Tips and Commands for AI‑Powered Development

Introduction

This guide presents 33 techniques that cover everything you need to use Claude Code effectively, including keyboard shortcuts, prompt tricks, MCP servers, project rules, and automation hooks.

Part 1: Getting Started and Basic Setup

1. Switch modes with Shift+Tab

Claude Code offers three input modes. Use Shift+Tab to cycle between:

Edit mode (default): file changes require user approval.

Auto‑accept mode (single Shift+Tab): writes directly to files, suitable for most tasks.

Plan mode (double Shift+Tab): generates an action plan without modifying code, ideal for research.

# Quick mode switching
Shift+Tab -> Auto‑accept mode
Shift+Tab -> Plan mode
Shift+Tab -> Return to edit mode

2. Add Claude Code to your IDE

Install Claude Code as an extension in VS Code, Cursor, or JetBrains to combine terminal and IDE capabilities. Follow the Anthropic documentation for installation instructions.

3. Configure multiline prompts

Enable multiline prompts in the terminal:

/terminal-setup  # configure Shift+Enter for multiline prompts

4. Connect IDE for automatic context

After connecting the IDE, Claude Code automatically includes the selected lines as context:

/ide  # select VS Code, Cursor, or JetBrains IDE

5. Open Claude Code quickly

Use the following shortcut (Mac/Windows): CMD / CTRL +

Escape

6. Other common shortcuts

CMD

/ CTRL + L: clear screen ESC + ESC: jump to previous item SHIFT + ENTER: insert newline (requires /terminal-setup) CMD / CTRL + R: detailed output

Part 2: Core Features

7. Image handling

Claude Code supports multimodal input; you can drag‑and‑drop or paste screenshots directly into the prompt for “make it look like this” tasks.

8. Cost tracking

Check token usage with:

npx ccusage
# shows input/output token breakdown
# tracks cost over time

Live view: blocks --live Tip: If monthly spend exceeds $100, consider subscribing to Claude Max ($100/month) for near‑unlimited usage.

9. Custom slash commands

Create a file under .claude/commands/ (e.g., release.md) and invoke it with /release.

10. Clear context

Prevent context pollution and reduce token consumption:

/clear

11. Resume session

Continue from where you left off:

/resume
# shows past sessions and lets you pick one

12. Handling long prompts

Press CMD / CTRL + N to open a new buffer.

Enter a formatted prompt.

Select all with CMD / CTRL + A and paste into Claude Code.

Part 3: Advanced Prompt Techniques

13. Control thinking budget

Allocate “thinking tokens” via keywords:

Basic thinking: think about this and implement... More thinking: think harder about the architecture... Maximum thinking:

ultrathink about the security implications...

14. Use sub‑agents

Parallelize complex tasks: "Use subagents to refactor this codebase" Claude distributes the work among sub‑agents and merges the results.

15. Loop task execution

Iteratively fix errors:

"Run the build in a loop and fix all errors as they appear"

Part 4: Required MCP Extensions

16. TaskMaster AI MCP

{
  "mcpServers": {
    "taskmaster-ai": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "--package=task-master-ai", "task-master-ai"],
      "env": {}
    }
  }
}

17. Playwright MCP

{
  "mcpServers": {
    "playwright": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@playwright/mcp@latest"],
      "env": {}
    }
  }
}

18. Context7 MCP

{
  "mcpServers": {
    "context7": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@upstash/context7-mcp"],
      "env": {}
    }
  }
}

19. Use other AI reviewers

Integrate tools like CodeRabbit to review Claude’s changes and configure different personalities for optimal results.

Part 5: Project Settings and Rules

20. Initialise repository

/init
# generates claude.md with conventions, file structure and dependencies

21. Dynamically add rules

# "Always use async/await instead of .then()"

22. Nested Claude.md files

project/
└─ claude.md           # global rules
└─ frontend/
   └─ claude.md        # frontend rules
└─ backend/
   └─ claude.md        # backend rules

23. Reference file‑based rules

## Coding standards
see: @.cursor/rules/self-improvement.mdc

## API documentation
see: @docs/api.md

Part 6: Basic Rules

24. Automatic version control

## Git and version control
- automatically add and commit after task completion
- use descriptive commit messages

25. Post‑task IDE diagnostics

## Code quality checks
**Run before completing a task:**
- use `mcp__ide__getDiagnostics` to check linting and type errors
- fix all reported errors

26. Documentation lookup

Collaborate with the Context7 MCP to search documentation on demand.