Set Up Your Environment
Get your Mac or Windows PC ready to build and ship real apps — install the essential tools, create your first project from a template, and run it locally with an AI coding agent.
You need a few tools before you can build anything. Set up the workshop once, then the agent does the heavy lifting.
Commands differ on Mac and Windows, so pick your OS below. Follow the steps in order; each tool depends on the ones before it. Everything is free, and the 1.5–2 hours is mostly downloads and sign-ups. If a command fails, check Troubleshooting, or paste the command and the error into Claude or ChatGPT.
We build on the Mark-Life/netxjs-monorepo template: a Next.js + Bun + Turborepo + Biome + shadcn/ui starter you copy in one command.
Prefer to let the agent do it? Install an agent from Step 8, then paste it the link to this page with a prompt like "Follow this lesson and set up my computer for coding for this course." It usually needs you only for the GitHub sign-up and the
ghbrowser login (Step 3). Use the steps below to follow along or troubleshoot.
What you'll end up with
A terminal, Git, a GitHub account, VS Code, the Bun runtime, Node.js, an AI coding agent, and a real app running on your own machine.
Make it yours: settings, configs & quality of life
Everything you installed is configurable, and your agent can set it up for you. Notice what makes you sigh and turn it into a prompt:
- "How can we make my terminal nicer to use? Set up some helpful aliases for me."
- "How can we improve my VS Code experience with settings? Turn on word wrap and anything else worth having."
- "Every time I do X it's annoying, can we configure something to fix that?"
The agent knows where these settings live and how to change them safely.
A couple of concrete examples
Shell aliases are short nicknames for longer commands. Define them once and they work in every terminal:
# In your shell config (~/.zshrc on Mac, your $PROFILE on Windows)
alias cl="clear"
alias cc="claude --dangerously-skip-permissions"Normally your agent asks for approval before every edit or command (you saw this in Step 8).
The --dangerously-skip-permissions flag, often called YOLO mode, turns that off: the
agent does the work without stopping to ask.
The trade-off: with permissions skipped, the agent can run any command, including destructive ones, so it could delete files it shouldn't. I run agents only this way, since approving every action drove me up the wall. Keep your code in Git so anything can be undone. If in doubt, leave it off and add the alias later.
Do the same for whatever you use most: Git commands, project shortcuts, your dev server. Tell
the agent "add an alias cc that runs Claude with permissions skipped" and it edits the
right file for you.
You did it
You have a working setup and a real app running on your own machine, created from the Mark-Life/netxjs-monorepo template in one command, without knowing what any of the code means yet.
What's next
Next we open that project up: how a Next.js app is organized, how pages and components work,
and how to make it look the way you want. Once bun dev shows the app at
https://web.localhost, you're ready to build. 🎉