Episode 2 · Resources
Everything from the video in one place: what an agent actually is, how it loops, remembers and stays inside your rules, how to set up Claude Code step by step, and the exact instructions for your first agent.

Use the official pages, so you're always installing the real thing.
An agent is the AI, plus tools, plus a loop. Everything else is details.
It learned by reading a huge amount of writing, and got very good at one thing: given everything so far, write the next word that makes the most sense. One word at a time.
Two things to remember: it only knows what it read plus what you show it, and it's always guessing. Usually right, sometimes wrong, and just as sure either way.
A chatbot can only talk. A tool is anything the AI is allowed to use: search your inbox, open a spreadsheet, check your calendar, write a file.
Instead of answering once and stopping, an agent keeps going until the job is done. You give it a goal; it figures out the steps.
Everything it's working on has to fit on a desk called the context window. It's big but not endless, and a new conversation wipes it clean. So good agents keep a notes file and read it every time they start.

On day one you don't hand over the company credit card. Let it look around, check its work, and give it more as it earns it.
It can read your inbox but not send. It can read the spreadsheet but not change it. Only connect accounts your company has approved.
Claude Code stops before it changes anything and asks you: yes, no, or "yes, and don't ask me about this again."
Agents are great at the first draft. Keep a person approving anything that goes to a customer.
If the steps are always the same, use an automation. If the job needs reading, judgment or "well, it depends," that's a job for an agent. Plenty of businesses use both.
A recipe: when this happens, do exactly these steps. Fast, cheap and predictable, but it gets stuck when something doesn't fit the recipe.
Like a new employee: you give it a goal and it figures out the steps. It can tell a complaint from a question and notice when something looks off. Slower, costs more to run, and it's guessing.
It would rather answer than say "I don't know," so it can fill a gap with something that sounds right and isn't: a made-up price, a date that's off, a policy you never had.
If it can't see your price list, it can't quote from it. If it doesn't know your rules, it'll guess.
The actual price list, the actual policy. And ask it to show where it found things, so you can check.
Keep a person approving anything that goes to a customer.
Take it slow. About ten minutes the first time. No coding.
Go to claude.ai. Claude Code needs a paid plan; Pro is the cheapest one that includes it.
Mac: hold ⌘ Command, press Space, type terminal, press Enter.
Windows: click Start, type PowerShell, press Enter.
A terminal is just a window where you type instead of click. You can't break anything by opening it.
On the Claude Code page, click Copy next to the install line. Click into your terminal, paste, and press Enter. Give it a minute.
You don't need to understand that line. It just downloads Claude Code and sets it up.
Make a folder on your desktop called business. In the terminal, type cd and a space, drag the folder into the terminal window, and press Enter. That tells the terminal "work in this folder."
claudePress Enter. The first time, your browser opens so you can sign in. When it asks if you trust this folder, say yes. From here on, it's just like chatting.
From easiest to most hands-on. For your first agent, all you need is way 1: your inbox.
Gmail, Outlook, Google Calendar, Google Drive and Slack have one. In Claude Code, just ask: "Connect to my Gmail." It walks you through it, and you sign in.
Some software companies make one for their product, and Claude Code is great at using them.
Most booking, accounting and customer-list software has one. Ask the company for API access, a key and the documentation, save the key in your folder, and have Claude Code build the connection.
Almost every system has an export button that gives you a spreadsheet. Save it in your business folder and Claude Code can read it. Works on day one.
Write it like you'd brief a new hire: the goal, the tools it can use, the rules, what "done" looks like, and when to come get you.
Connect to my Gmail.
Every morning, go through my inbox from the last day. Sort every email into three piles: needs a reply today, can wait, and ignore. For each one that needs a reply, write a draft in Gmail, but don't send anything. If you're not sure which pile something goes in, put it in 'needs a reply' and tell me why. When you're done, write me a one-page summary called today.
Save that as my morning briefing.
Run my morning briefing.
Never offer a refund without asking me first. Add that to your notes.

Reads new job requests, checks your calendar, and drafts a reply with three open time slots.
Pulls your price list and past jobs and drafts the quote, with the final number left for you.
Every Monday, checks who's past due and drafts a friendly reminder to each one.
Drafts answers from your own policies, and flags anything new.
Benmore builds AI agents into real businesses: your inbox, calendar and systems connected, so it just runs.
Talk to Benmore