Cohort 2 · Week 1 Walkthrough

Your first AI agent. On your machine. Tonight.

This is the Week 1 path. Install Claude Code, set up your workspace, say hi to your agent, and understand how the whole thing works before we stack anything else on top. No Discord, no MCPs, no agents-on-cloud. Just you and your terminal talking to a real AI.

What you'll do

  1. Get a Claude Pro or Max subscription
  2. Install Claude Code (one command)
  3. Set up your side-by-side workspace layout
  4. Open Claude Code and say "hi"
  5. Trust the folder — the agent's brain
  6. Check your usage dashboard
  7. When things break (they will)
  8. What's next — Week 2 and beyond

Part 1 — Claude subscription

You need an active Claude Pro or Max plan before you install anything else. The free tier cannot run Claude Code.

1Go to claude.ai

Open claude.ai in your browser. If you don't already have an account, sign up. Use an email you'll remember — you'll sign in again in Part 4.

2Pick a plan

Click Upgrade (usually top-right of the sidebar).

API vs subscription. Claude also has a pay-per-token API. We don't use it. The subscription is cheaper and more predictable for agent work. Don't buy API credits — you don't need them.

3Confirm the email on your account

Write down the email address on your Claude account. You'll need it when Claude Code asks you to sign in. If you forget — claude.ai → bottom-left avatar → Settings → Account.

Part 2 — Install Claude Code

One command installs everything. Read what it does before you run it.

What the install actually does

This is not magic. When you run the command below, a script runs on your computer that:

  1. Checks your operating system, internet connection, and what's already installed
  2. Installs bun — a small JavaScript runtime Claude Code needs
  3. Installs Claude Code — the CLI tool that runs your agent
  4. Creates a starter workspace folder on your computer with a CLAUDE.md inside
  5. Runs a health check to prove every layer works

That's it. No account creation, no data upload. It's the same pattern Homebrew and nvm use.

If you're on Mac

Step 1: Open Terminal. Press Cmd + Space, type Terminal, press Enter.

Step 2: Paste this entire line, press Enter:

curl -fsSL https://setup.coachreemteam.com/setup.sh | bash

The script narrates itself. You'll see progress lines like:

━━━ ReemTeam Agent Setup ━━━ [1/4] Installing bun (runtime)... done (23s) [2/4] Installing Claude Code... done (18s) [3/4] Creating workspace folder... done [4/4] Running verification... all green ✓ ━━━ Setup complete ━━━
If you see a password prompt: That's macOS asking for your login password. Type it (you won't see characters as you type — that's normal, not broken), press Enter.

If you're on Windows

Step 1: Open PowerShell as Administrator. Click Start → type PowerShell → right-click the result → Run as administrator → Yes.

Step 2: Allow scripts (first time only):

Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope CurrentUser

Answer Y to the prompt.

Step 3: Run the installer:

iwr -useb https://setup.coachreemteam.com/setup.ps1 | iex
Windows Defender popup? Click More info, then Run anyway. The script only pulls from official sources.

After install — on both platforms

Fully quit your terminal and reopen it. This is the step people skip. It matters.

Why? The installer modifies your terminal's environment variables. Your current window still has the old values cached. Reopening loads the new setup.

Something went wrong

Install errored halfway? Take a screenshot of the red error text, post it in the Cohort 2 Telegram. Don't re-run blindly — we want to see the error before you try again. The script is safe to re-run once we diagnose.

Part 3 — The side-by-side layout

Before you launch Claude, set up your screen the way Reem works. This single habit 3x's how fast you execute.

Why side-by-side matters

When you work with an AI agent, you're constantly feeding it context from other places: a Notes file, a ChatGPT plan, a Discord conversation, a screenshot of an error. If you're alt-tabbing between full-screen windows, you lose the thread and you lose time.

The move is to split your screen in half:

Layout

Left half: browser / ChatGPT / notes. Right half: Terminal running Claude Code.

This is the same layout Reem uses. Learn it from Day 1.

How to set it up

On Mac: Click and hold the green dot on the top-left of your Terminal window → pick Tile Window to Right of Screen. Then pick the app you want on the left.

On Windows: Drag the Terminal window to the right edge of the screen until it snaps to half. Windows asks you to pick a window for the left half.

Reem's setup: Claude Code on the right. On the left — Discord during calls, ChatGPT/Codex when planning prompts, or a markdown file he's working from. Plan in one window, execute in the other.

Part 4 — Open Claude Code and say "hi"

This is the moment your agent comes alive. 30 seconds of work.

Good news: the setup script already created your workspace folder for you, with a starter CLAUDE.md inside. You don't need to create it — just go there and launch.

1Navigate to your workspace and launch

In your fresh Terminal window, paste this one line:

cd ~/cohort2-workspace && claude

On Windows:

cd $env:USERPROFILE\cohort2-workspace; claude
Heads up on the folder name: the installer still creates a folder called cohort2-workspace — that's a leftover from the last cohort. We're renaming it next week. For now, use the name above. You can rename the folder later.

2Sign in

A browser tab opens with a Claude sign-in page. Use the email on your Claude Pro/Max account (Part 1, Step 3).

After sign-in, return to your terminal. You should see:

Welcome back, [Your Name]! Sonnet 4.6 · Claude Pro your.email@example.com /Users/yourname/cohort2-workspace No recent activity >

That > at the bottom is your prompt. Anything you type there talks to Claude.

3Say hi

Type the word hi and press Enter.

hi

Claude responds. That's a working AI agent. Not a mock, not a demo — the real thing, running on your computer, reading from the folder you just created.

Why we start with "hi": it's the cheapest possible test. If Claude responds, every layer works — the install, the login, the folder trust, the network. One second, one word, full confirmation.

claude: command not found

Your terminal didn't pick up the install. Fully quit Terminal (Cmd+Q on Mac, close window on Windows), reopen, try again.

cd: no such file or directory

The setup script didn't finish cleanly. Re-run it: curl -fsSL https://setup.coachreemteam.com/setup.sh | bash (Mac) or iwr -useb https://setup.coachreemteam.com/setup.ps1 | iex (Windows). It's safe to re-run — it skips anything already done.

Part 5 — Trust the folder. The folder is the brain.

The most important concept of the whole program. Read this twice.

Your workspace folder is your agent

Claude Code is a blank slate. It doesn't know you, your business, your voice, or your goals — unless you tell it. The way you tell it is by putting files in your workspace folder.

Everything in your workspace folder becomes part of your agent's brain:

When you ask your agent to do something, it reads the folder first, then responds with that context loaded. A blank folder gets you generic answers. A rich folder gets you answers that sound like you and know your business.

Concrete example from Reem's setup with Nikky: We dropped markdown files into her stock agent folder — a strategy file, a watchlist, daily prompts, a trainer doc. That folder is now her trading agent's brain. Every time she talks to it, it reads those files first. You'll do the same for your business.

What to drop in your folder this week

Start small. Even one file helps.

You don't need to be perfect. Messy notes work. Your agent will improve as your folder grows.

Reading markdown files — install Obsidian

Markdown files (.md) look ugly in TextEdit or Notepad. Obsidian is a free app that makes them readable and lets you edit them like a proper note system.

Install Obsidian, point it at your workspace folder, and you can read and edit every file your agent uses — from the same app your agent is using.

Part 6 — Check your usage dashboard

Know your limits before you hit them.

Where to look

Open the Claude desktop app (or claude.ai). Bottom-left corner → click your avatar → SettingsUsage.

You'll see a bar showing what percentage of your weekly limit you've used. Reem's resets every Saturday at 6pm. Yours may be a different day — check the dashboard to confirm.

Rough math for Pro ($20) users

How to save usage (this is the move)

Interact with your agent through the terminal, not through extra tools or back-and-forth over Discord. Plan your prompts in ChatGPT or a notes app first, then paste the finalized version into Claude. Every back-and-forth "wait, try that again" burns tokens.

The Reem move: open ChatGPT (free) on the left side of your screen. Draft your prompt there. When it's clean, paste it into Claude Code on the right. Claude sees the polished version, not your 5 rough drafts. Your usage lasts 3x longer.

Part 7 — When things break (they will)

This isn't if. It's when. Learn the fix now so you don't panic later.

The universal first move

Something weird? Claude typing but not responding? Terminal acting stuck? Before you do anything else:

  1. Type /exit in Claude Code → press Enter (exits cleanly)
  2. Fully quit Terminal (Cmd + Q on Mac, close window on Windows)
  3. Reopen Terminal
  4. cd ~/cohort2-workspace && claude → press Enter

9 times out of 10, that's the whole fix. Don't debug for an hour — close and reopen first.

The second move — health check

If the restart doesn't fix it, run the diagnostic:

reemteam-check

Run it from a fresh Terminal (not inside Claude Code). You'll see a pass/fail for every layer of your stack. Anything red prints the exact fix next to it.

The third move — ask for help

Post a screenshot in the Cohort 2 Telegram group with three things:

  1. What you were trying to do
  2. What the terminal said (screenshot)
  3. What you've already tried

Code (the technical agent) monitors the channel. Expect a reply within 30 min during the day, first thing the next morning if you post overnight.

What NOT to do when something breaks: don't delete your workspace folder, don't reinstall everything from scratch, don't panic-Google obscure terminal errors. Restart → health check → ask. In that order.

Part 8 — What's next

You have the foundation. Here's what stacks on top over the next 6 weeks.

Week 2 — Tools and workflows

Weeks 3-6

Your Week 1 homework

Due before Week 2 session

  • Claude Code installed and working — screenshot of your terminal
  • Workspace folder exists with CLAUDE.md inside — screenshot
  • 5 real tasks given to your agent — screenshots of each result
  • Your best result posted in Cohort 2 Telegram with #Week1Win
  • Daily check-in every day — one screenshot, one action, posted to Telegram

See you Wednesday.

You've got everything you need.

When your agent says hi back, you're in. Verify your setup so we know you made it — and so we can focus on building when class starts, not debugging.

Run Verification →

Stuck? Post in the Cohort 2 Telegram — Code replies within 30 min during the day.