Claude Code tips and tricks

If you’re familiar with all the AI news, you probably have heard about Claude Code before. It’s an agentic coding tool developed by Anthropic that operates directly in the terminal, assisting you in refactoring, documenting, and debugging code. It sure had been one of the first of its kind and unlike any other traditional AI-fused IDEs or chat-based coding assistants at the time.

Like many developing tools out there, Claude Code only supported macOS, Linux, and WSL on Windows when it first came out. But that has changed last week, when the new versions finally support Windows “natively” (still need Git for Windows installed for that sweet Git Bash to work, but hey, it’s still progress 🚀).

Loading embed...

So now that everyone is happy, let’s dive in for some more details.

Essential setup and configuration

Quick installation

It’s pretty straightforward and the same no matter what OS you’re on:

# Install Claude Code
npm install -g @anthropic-ai/claude-code
# Navigate to your project
cd your-awesome-project
# Start coding with Claude
claude

* If you’re behind a corporate proxy, make sure you’ve set the NODE_EXTRA_CA_CERTS to point to your certificate file before running claude for Claude Code to run correctly.

IDE integration

You can also install the Claude Code extension for VS Code, Cursor, or Windsurf. While it’s essentially just a launcher, it makes opening Claude Code seamless and allows you to run multiple instances in parallel.

Skip permission prompts

One thing about Claude Code is the constant permission requests, which can be both good and bad depending on your view. But if you like moving fast, you can bypass all that by using:

claude --dangerously-skip-permissions

Core workflow tips

Use /clear frequently

This is probably the most important tip I can give you. Call /clear often—like, really often. AI agents tend to get a bit… unpredictable the longer a conversation goes on, especially when you’re jumping between different tasks. So whenever you’re starting something new, just clear that chat history and start fresh.

Master the CLAUDE.md file

Here’s something cool: Claude automatically looks for a CLAUDE.md file when it starts up and pulls it into context. Like giving Claude a cheat sheet about your project. You can document things like:

  • Your coding standards and conventions
  • How you handle branches (merge vs. rebase, branch naming, etc.)
  • Environment setup quirks
  • Common patterns you use in the codebase
  • Any weird behaviors or gotchas to watch out for

And here’s a pro tip: when Claude messes something up, ask it to update the CLAUDE.md file so it doesn’t make the same mistake twice.

See Claude as your very fast intern…

…or the way Microsoft sees it, “COPILOT! COPILOT! COPILOT!”

Claude is like having an intern with perfect memory but not much real-world experience. It’s incredibly fast and remembers everything you tell it, but you still need to check its work and give clear directions. Don’t expect it to read your mind, you need to spell out what you want.

Context is everything

The more context you give Claude, the better it performs. If you’re thinking about edge cases, mention them. If there’s a similar pattern elsewhere in your codebase, point it out. Don’t assume Claude knows the implicit stuff that’s obvious to you.

Advanced features worth knowing

Extended thinking modes

Claude Code has these thinking modes that basically give it more time to think through complex problems. Just use these keywords in your prompts:

  • think - Basic extended thinking
  • think hard - More thinking time
  • think harder - Even more thinking time
  • ultrathink - Maximum thinking power

Use these when you’re asking Claude to plan something complex or work through a tricky problem.

Sub-agents and parallel tasks

Claude can actually spawn multiple instances of itself to work on different parts of a problem simultaneously. You’ll see something like Task(...) in the output when this happens. It’s pretty neat for complex tasks that benefit from multiple perspectives.

Working with images and visual feedback

Yes, Claude Code does work with images, don’t assume it cannot because of, you know, “TErMInAL”. You actually can:

  • Take screenshots with Shift ⇧ + Cmd ⌘ + Ctrl ⌃ + 4 on macOS, then paste with Ctrl ⌃ + V (not Cmd ⌘ + V!)
  • Drag images directly into the terminal
  • Show Claude UI mockups and ask it to build the interface
  • Take screenshots of bugs for better debugging context

Custom slash commands

You can create your own slash commands by putting Markdown files in the .claude/commands folder. These show up in the slash commands menu and you can even commit them to git to share with your team. Use $ARGUMENTS to pass parameters to your commands.

Productivity hacks and shortcuts

Terminal shortcuts you need to know

  • Shift ⇧ + Enter for new lines: Run /terminal-setup to enable this. Or on Windows cases where this does not show up, type \ + Enter for new lines.
  • Stop Claude: Use Esc (not Ctrl + C, which exits entirely)
  • Jump to previous messages: Esc twice to see all previous messages
  • Resume old conversations: Use --resume flag to list prior threads

Managing costs and context

Claude Code pulls a lot of context automatically, which is great but can get expensive. Keep your CLAUDE.md files concise, use /clear liberally, and be specific about what context you actually need.

MCP integrations

MCPs (Model Context Protocol) are where Claude Code really shines. Popular ones include:

  • Context7: Grabs documentation for any tech stack on demand
  • Puppeteer: For automated screenshots and testing
  • Various project-specific MCPs: Check the community for your specific needs

Use --mcp-debug if you run into configuration issues.

Common pitfalls to avoid

Over-engineering

Claude often suggests overly complicated solutions because, well, it’s like an intern trying to impress you. Don’t be afraid to ask for simpler approaches.

Context overload

It’s easy to let conversations get too long and unfocused. Remember: /clear is your friend.

Fighting the abstraction

Claude Code works differently than traditional IDEs, it abstracts you above the code level to focus on strategy. Don’t fight this; embrace it.

Wrapping up

Claude Code represents an interesting approach to AI-assisted development, moving the interaction from sidebars and autocomplete to a conversational terminal interface. Whether it fits your workflow depends on your preferences and project needs.

The tips above should help you get started if you decide to give it a try. Remember the basics: use /clear frequently, provide good context, and treat it like you would any other development tool, which can be useful for some tasks, maybe not for others.

As with any new tool, there’s a learning curve. Some developers love the conversational approach, while others prefer more traditional IDE integrations. The choice is yours.

If you do try Claude Code, start small with simple tasks and gradually work up to more complex workflows. And remember, it’s still a tool, not a replacement for understanding your code and making thoughtful decisions.