Claude Code Overview

Overview

Every build lesson in this course runs through Claude Code: you describe what the experiment should do, it writes the code, you review and test. Before the first build lesson, this page covers the handful of controls you will actually use: how to open it, how to talk to it, how to pick a model, and how to control how much it does before asking you.

There are two ways to run Claude Code in this course, and both work for every lesson:

• The terminal version. You type into a terminal window (the same kind you used to run node --version in the prerequisites). This is what the lessons' examples assume.
• The VS Code extension. A chat panel inside VS Code, with the same brain behind it. Proposed changes show up as side-by-side comparisons, which some people find easier to review. If a terminal feels foreign, start here.

Pick whichever feels comfortable. Everything below covers both.

Opening Claude Code

Terminal: open a terminal in your project folder (in VS Code: Terminal, then New Terminal) and type:

claude

That starts a session in that folder, which matters: Claude Code can see and edit the files of the project you started it in.

VS Code extension: install it once from the Extensions view (press Cmd+Shift+X on Mac or Ctrl+Shift+X on Windows, search for "Claude Code", click Install). Then open the panel by clicking the Claude icon in the left Activity Bar, or press Cmd+Shift+P / Ctrl+Shift+P and type "Claude Code". With your project folder open in VS Code, the panel works on that project automatically.

Either way, you'll log in with your Claude account the first time.

Talking to It

You type plain language. There are only three pieces of syntax worth knowing now:

• @ attaches a file. Typing @ starts a file picker; @prisma/schema.prisma attaches the schema so Claude Code works with what's already there. Every build lesson's prompt starts this way, and the first build lesson walks through it again.
• / runs a command. Commands start with a slash, like /model and /clear below. Typing / on its own shows the list.
• Esc interrupts. If Claude Code is heading somewhere you don't want, press Escape. It stops and keeps what it's done so far, and you can redirect it.

One more habit: /clear starts a fresh conversation with an empty memory of the chat so far. Use it between unrelated tasks, like moving from one lesson to the next, so old context doesn't muddy new work.

Choosing a Model

Claude Code can run different models, which trade speed against depth. Smaller models answer faster; bigger models handle complex builds better.

To see or change the model, type:

/model

A picker opens showing the options, with names like haiku (fastest), sonnet, and opus (most capable). Arrow keys to choose, Enter to switch. You can also switch directly, for example /model opus. Choosing default returns to the recommended model for your account.

In the VS Code extension, /model works the same way in the prompt box.

For this course, the default model is fine for everything. If a build lesson's task comes back not working after a couple of revisions, switching to a more capable model for that task is a reasonable lever to pull.

Permission Modes: How Much It Does Before Asking

This is the control that matters most, because it decides how much you review. Claude Code always shows you what it wants to do; the mode decides whether it waits for your approval first.

How to switch, in the terminal: press Shift+Tab to cycle through the modes. A label near the prompt shows which one you're in.

How to switch, in VS Code: click the mode indicator at the bottom of the prompt box (it shows the current mode by name) and pick from the list.

What this course recommends: stay in Manual for the build lessons. The lessons are written for it: "approve the schema edit, approve the migration" assumes you're seeing each step, and watching the steps is how the system from the architecture lesson becomes familiar. Once you've built a few pages and trust the rhythm, Edit automatically or Auto saves clicks on tasks you've done before.

Plan Mode, Specifically

Plan mode is the one worth understanding beyond the table, because it's the mode for anything big or unfamiliar. In plan mode, Claude Code can read your files and investigate, but it cannot change anything. Instead it comes back with a plan: what it intends to do, file by file.

When the plan appears, you choose:

• Approve and let it run (it switches to automatic execution),
• Approve but review each edit (it executes in Manual style), or
• Keep planning (you give feedback and it revises the plan).

To enter plan mode, cycle to it with Shift+Tab in the terminal or pick Plan from the mode indicator in VS Code. To leave without approving anything, cycle out the same way.

A research analogy that holds up well: plan mode is the proposal meeting before anyone touches the protocol. For a one-line styling tweak you don't need a meeting. For "restructure my task to add a fourth phase," you want the proposal first.

The Short Version

• Open it in your project folder: claude in a terminal, or the Claude panel in VS Code.
• Describe outcomes in plain language; attach files with @; interrupt with Esc; reset the conversation with /clear.
• /model changes the model; the default is fine for this course.
• Shift+Tab (terminal) or the mode indicator (VS Code) switches permission modes. Stay in Manual while you're learning; use Plan for big changes.

That's everything the build lessons assume. Next, we set up the project.