Your Own AI Assistant — Balibound Realty Team

01

Install Cursor

~5 min no terminal

Cursor is the editor that'll be the home for your assistant. It looks like a code editor but you'll mostly just type into a chat panel — you don't need to know how to code.

Go to cursor.com and download it
Open the downloaded file and install it (drag to Applications on a Mac; run the installer on Windows)
Open Cursor (right-click → Open the first time if your Mac complains about an unidentified developer)
Sign in with Google or email — either is fine

what success looks like

Cursor opens to a welcome screen with options like "Open Folder" and "Clone Repo". You're signed in — your name shows in the bottom-left corner.

if you get stuck

If your Mac says "Cursor can't be opened because it's from an unidentified developer", right-click the app and pick Open, then click Open again in the popup. Only needed once.

02

Install Node.js

~5 min no terminal

Node is the engine your assistant runs on. You don't need to understand it — install it once and forget about it.

Go to nodejs.org
Click the big green LTS button
Open the downloaded installer and click through it
It may ask for your password at the end — that's normal

what success looks like

The installer says "The installation was successful." You can close it.

if you get stuck

If the installer fails, restart your computer and try again — usually a permissions hiccup that clears after a reboot.

03

Install Claude Code

~2 min terminal

Open Terminal (Cmd + Space, type "terminal", press enter), then paste the command below and press enter.

npm install -g @anthropic-ai/claude-code

It'll churn for 30–60 seconds, then go quiet when it's done.

what success looks like

Terminal scrolls a bit, goes quiet, and shows your prompt again. No large red error messages.

if you get stuck

If you see permission denied or EACCES, paste this instead:

sudo npm install -g @anthropic-ai/claude-code

It'll ask for your password (typing won't show — that's normal, just type and press enter).

04

Add the Claude Code extension to Cursor

~2 min no terminal

This puts your assistant inside Cursor as a chat panel.

Open Cursor
Click the Extensions icon in the left sidebar, or press Cmd + Shift + X
Search: Claude Code
Find the one published by Anthropic — click Install

what success looks like

The Install button changes to "Uninstall". A small Claude icon appears in the left sidebar.

05

Sign in to Claude Code

~3 min browser

Option A — Claude subscription (recommended): if you have a Claude Pro or Max plan, use that. Click "Use with Claude subscription" in the panel, sign in, click Authorize.
Option B — API key: go to console.anthropic.com, sign up, go to Settings → API Keys, create one, paste it into Cursor when prompted.

Open the chat panel: click the Claude icon in Cursor's left sidebar, or press Cmd + Esc.

what success looks like

The chat panel is open with a "How can I help?" input field at the bottom. No sign-in button left on screen.

If you're not sure which option makes sense for you, ask Chris — it's a quick call.

06

Make your working folder

~1 min finder

This is the one folder where your assistant will live, remember your context, and save what it learns about how you work.

Open Finder → Desktop
Right-click in empty space → New Folder
Name it whatever makes sense — your name, "My Workspace", your call
In Cursor: File → Open Folder, navigate to it, click Open

what success looks like

Cursor's title bar shows the folder name. The file tree is empty — that's fine, it'll fill in as you work.

07

Wire up Playwright (browser automation)

~3 min terminal

This is what lets your assistant look at a real webpage for you — the live site, an admin page you already have access to, anything in a browser. Without it, your assistant can only work with text you paste in.

Make sure your working folder from step 6 is open, then in Terminal:

claude mcp add playwright -- npx @playwright/mcp@latest

Then close the chat panel and reopen it (Cmd + Esc twice) so it picks up the new tool.

what success looks like

Type /mcp in the chat panel — you should see playwright listed as connected.

if /mcp doesn't show playwright

Quit Cursor (Cmd + Q), reopen it, open your working folder again, check /mcp.

Still not there? In the chat panel: /mcp → Add MCP server → name playwright, command npx @playwright/mcp@latest, save.

08

Tell it who you are

important ~10 min

Open the chat panel (Cmd + Esc) and introduce yourself. This is the single most useful thing you'll do — everything after this session gets sharper because it knows your role and how you like to work.

You're going to be my personal AI assistant. I work at Balibound Realty. I want you to help me with my day-to-day work — [tell it what you actually spend your time on: viewings, listings, follow-ups, admin, whatever it is for you]. Let's start with you asking me a few questions so you understand how I work, then build that up into a file you keep for yourself.

Answer its questions honestly and in your own words — the more specific, the better it gets. Some things worth telling it early on:

What you actually spend most of your day doing
The tools and apps you use most (WhatsApp, email, the admin pages you have access to)
What kind of help would save you the most time — drafting, organizing, reminders, research
How you like things explained — short and plain, or more detail

what success looks like

It's asked you a handful of questions, you've answered, and it's saved a summary of who you are into a file in your folder. From now on it opens every conversation already knowing your role.

09

How to loop in Baxter

important read carefully

Nothing changes about how you talk to Baxter. You still message Baxter on WhatsApp exactly like you do today for anything real-time — a listing's current price, whether a lead is still open, an owner's contact details. Baxter is still the one place that data lives.

What your own assistant adds is everything around that question — and a cleaner way to ask it. Here's the simple version of how it works:

You're working with your assistant on something (prepping for a viewing, drafting a follow-up, organizing your week)
It hits a point where it needs a live fact only Baxter's system holds
Ask it to draft the exact question you'd send to Baxter — short, specific, ready to paste
Copy that into WhatsApp yourself and send it to Baxter
When Baxter replies, paste the answer back to your assistant so it can carry on with the task

the one rule that matters

You're always the one who hits send. Your assistant drafts the question — it never messages Baxter, Chris, a client, or anyone else on your behalf automatically. That's the same rule Chris works under with Baxter, and it keeps a real person checking every message before it goes anywhere.

Try it once now, on something small, so you've seen the shape of it:

I need to check something with Baxter — draft me a short WhatsApp question asking [something small and real, e.g. whether a specific listing is still available].

what success looks like

Your assistant hands you a clean one-line question, ready to paste into WhatsApp. You send it to Baxter yourself, get the answer, and can bring it back into the conversation with your assistant.

if this feels like a lot of copy-pasting

For most of your day it won't come up much — your assistant handles the rest without ever needing Baxter. If you find yourself doing this copy-paste step constantly, tell Chris. There may be a way to make the two talk to each other more directly down the line, but that's a bigger project for later, not something to build yourself on day one.

Set up your own AI assistant.

The nine steps (path A, full setup)

That's it, you're set up.