1 Welcome

Everything this app needs to talk to the rest of Spectrum's systems — sign-in, a shared database, hosting — is already built and waiting for you. You won't set any of that up from scratch. Your job is to describe what you want to build, review what Claude proposes, and test it.

2 Sign in — get your accounts ready

Claude Code needs you to be signed in before it will do anything, so grab these two logins from a mentor before you go any further:

Get both passwords from a mentor now — you'll need the Claude account in the very next section, and the GitHub account a little further along.

3 Install your tools

There's really only one thing you install by hand: the Claude Desktop app. Once that's running, you can ask it to install everything else for you — and it will ask your permission at every step, so nothing happens without you clicking Accept. Windows and Mac only below.

Please check whether I have GitHub CLI (gh), Node.js, and the Firebase CLI installed on this computer. Install whichever ones are missing yourself — on Windows use winget, on Mac use Homebrew (install Homebrew first if I don't have it). Ask me before running anything, and tell me in plain English what each command does before you run it.

git --version
gh --version
node --version
firebase --version

4 Meet Claude Code

Before you start building, here's what a session in the Code tab looks like:

• You type what you want in plain English, in the message box at the bottom.
• Claude reads and writes files in your project folder, and can run commands — like installing something, or starting a test server.
• Every file change and every command shows up as a card you can Accept or Reject before it happens. This is called Manual mode, and it's the default — nothing changes on your computer until you say yes.
• Your project has a file called CLAUDE.md sitting in the folder. Claude automatically reads it at the start of every session — it's a cheat sheet that tells Claude exactly how Spectrum apps are built, so you don't have to explain any of that yourself.

Choosing a mode

Next to the message box there's a mode selector. It controls how much Claude does on its own before checking in with you:

A good rhythm for a new feature: start in Plan so Claude explains its approach first, then switch to Manual or Accept edits to actually build it.

5 Pick your model, and know your usage limits

Which model to use

The model dropdown next to the message box lets you pick which version of Claude does the work. For building your app, use the latest Claude Sonnet model — it's what Team 3847's account is set up for, and it's plenty capable for everything in this guide.

If a mentor suggests trying Opus for something especially tricky, that's fine — but Sonnet should be your default for basically everything you build here.

Usage limits — what to expect

Claude accounts refill on a rolling basis — a short window that resets every few hours, plus a bigger weekly allowance. Since everyone on the team signs into the same shared account (Spectrum3847@radiantrobotics.org), that allowance is shared across every student building at the same time, not just you.

• If you see a message about hitting a session or weekly limit, that's normal — not something you broke. It tells you when it resets, so just come back then.
• Busy times (like everyone building right before a meeting) burn through the shared limit faster than working solo. Spread out big building sessions when you can.
• Long, rambling sessions use more of the allowance than focused ones. Starting a fresh session for each new feature — instead of one giant session for your whole project — helps a lot.
• Being specific about what you ask for ("add a search box that filters the parts list by material") uses less than a vague ask ("make this better"), because Claude has to do less guessing and re-work.

Still stuck on a limit and it's blocking real work? Ask a mentor — there may be a way to add usage headroom for the team.

6 Plan your app

Before you open Claude Code and start building, spend a few minutes deciding what you're actually building. The best student projects start small: one clear problem, one main type of thing you're tracking, for one group of people. You can always add more later.

Answer these three questions — even just to yourself:

1. What problem does this solve for the team?
2. Who's going to use it, and what's the one thing they need to do with it?
3. What's the one "main data type" — the noun the whole app revolves around? (an alumnus, a part, a meeting, a sponsor…)

Still brainstorming? A few more ideas that follow the same one-clear-problem shape:

• Meeting Notes Archive — a searchable log of team meeting notes by date
• Competition Day Checklist — a shared checklist the drive team runs through at every event
• Sponsor Contact Tracker — who to contact at each sponsor, when you last reached out, what they've given
• Event RSVP Form — lets members sign up for an event and see who else is going
• Scouting Notes Viewer — a simple read-only page that displays match scouting data

7 Build it with Claude Code

This is where two tabs work together: Register My App (right next to this one) and the Code tab you already opened in Section 3.

8 Go live and keep improving it

Once Claude pushes your first commit, your app is live at https://spectrum3847.github.io/YourRepoName — usually within about a minute. Open that link and actually sign in with your Google account to make sure the whole flow works end to end (sign in → get approved by Allen → see your app).

From here, building is just a conversation:

• Show it to a mentor or another student and get feedback
• Come back to the Code tab any time and ask Claude for the next feature, a bug fix, or a design tweak
• Small, frequent changes are easier to review than one giant one — accept changes as you go rather than saving up a huge pile of edits

9 Getting unstuck

Still stuck? Ask a mentor, or post in the team Slack. Nobody expects you to have all the answers on your first app.