Claude Code - 47 PRO TIPS in 9 minutes

Mastering Claude Code: Pro Tips from a Developer’s Perspective

Hey developers! If you’re diving into Claude Code for your coding projects, you’re in for a powerful experience. Greg, a seasoned developer, recently shared his insights and pro tips on using Claude Code effectively. Drawing heavily from a detailed post by Boris Churnney, the creator of Claude Code at Anthropic, Greg highlights some invaluable tips that can take your coding workflow to the next level. Here’s a comprehensive breakdown of these tips to help you become a Claude Code pro.

1. Claude Code is a Command-Line Interface (CLI)

Claude Code operates as a bash-based CLI, which means you can leverage all the familiar command-line tricks you already know:

Pass command-line arguments that run on startup.
Use -p to run Claude Code in headless mode.
Chain it with other command-line tools and pipe data into it.
Run multiple instances simultaneously, and even have Claude Code launch sub-agents or other Claude Code instances.
When you see tasks or sub-agents in action, that’s Claude Code managing multiple processes behind the scenes.

2. Using Images to Boost Your Workflow

Images are surprisingly powerful when working with Claude Code:

On macOS, drag images directly into the terminal or use Shift + Command + Control + 4 to take a screenshot and then paste it with Control + V (not Command + V).
Mockups: Paste your UI mockups into Claude Code and have it build interfaces based on your designs.
Feedback Loop: After Claude builds something, take a screenshot of the output, feed it back into Claude, and get iterative improvements.
For automation, set up the Puppeteer MCP server locally to programmatically take screenshots and save them, enabling smoother feedback cycles.

3. Harness the Power of MCP Servers and Clients

Claude Code can act as both an MCP server and client, opening up powerful integrations:

Use MCP servers like Postgres to connect directly to your databases.
Access wrappers around APIs or up-to-date documentation via MCP servers—some companies like Cloudflare are already providing docs this way.
Even if documentation isn’t available via MCP, Claude Code can fetch URLs and extract knowledge. For example, Greg built a game by feeding Claude the official Uno rules website, letting it code game logic precisely.

4. The Magic of `claude.md`

One of the first pro tips Boris mentions is the use of a claude.md file:

This file contains project-specific instructions loaded with every request, including bash commands, style guides, linting rules, test instructions, and repo etiquette.
Initialize one with /init in your project directory—Claude scans your repo and summarizes its structure.
Add instructions anytime by prefixing with a hash sign (#).
Use a global claude.md in your home directory for universal instructions across projects.
You can also create subdirectory-specific claude.md files.
Remember: this prompt loads on every interaction, so keep it clear, specific, and free of duplication.
Use Anthropic’s prompt optimizer to refine your claude.md.

5. Slash Commands: Custom Prompt Templates

Slash commands are user-defined prompt templates you can save in the cloud/commands folder:

Examples include commands for refactoring code, linting, or reviewing pull requests.
You can pass command-line arguments to these commands, which get interpolated into the prompt.
This lets you automate repetitive tasks and customize Claude’s behavior efficiently.

6. UI Tips for Smooth Sessions

Use Tab completion for files and directories to help Claude focus on specific parts of your project.
Don’t hesitate to hit Escape to stop Claude when it goes off track.
If Claude makes a mistake, hit Escape and use the undo command to revert the last response.
Frequent interruption leads to better, more controlled sessions.

7. Version Control is Your Best Friend

One common failure mode is Claude getting overly ambitious and making breaking changes:

Always use version control alongside Claude Code.
Have Claude commit after every major change—it writes excellent commit messages.
Revert often and don’t be shy about clearing conversation history and starting fresh with more specific instructions.
Install the GitHub CLI for seamless GitHub interactions, or use the GitHub MCP server as an alternative.
Claude can help you file PRs and perform code reviews on them.

8. Managing Context and Costs

Keep an eye on the auto compacting indicator to avoid losing important context.
Compact or clear context at natural breakpoints, such as after finishing a task or committing code.
Use scratchpads or GitHub issues for planning work externally to keep Claude’s memory fresh.
If you pay per token, managing context efficiently is crucial.
For team cost tracking, leverage Cloud Code’s open telemetry support and connect it to monitoring tools like DataDog.
To avoid cost headaches, consider upgrading to Claude Max plans ($100 or $200), which bundle tokens and improve affordability.

Final Thoughts

Claude Code is a powerhouse for developers, especially when you tap into its full potential with these pro tips. From command-line tricks and image handling to MCP integrations, customizable prompts, and smart version control practices, there’s a lot to explore.

If you want to dive deeper, check out Boris Churnney’s original post linked below, along with other resources Greg recommends. With a bit of practice, Claude Code can become your go-to coding assistant that accelerates your development workflow like never before.

Resources

Happy coding with Claude Code! 🚀

Did you find these tips helpful? Share your Claude Code experiences or questions in the comments below!

← Back to Greg Baugues Blog