Setup / Claude

Get Started with Claude 5 Must Do's

Five settings you can knock out in one go, and you'll want them done before you type a real prompt.

July 2026

Written July 2026. Menus move around, so if a button isn't exactly where I say it is, look nearby, because the idea behind it doesn't change.

Most people open Claude, type a question, get something fine but generic, and never touch the settings again. That's why it feels like every other chatbot. The five things below take about an hour total, and they're the difference between a chatbot and something that actually knows how you work.


1. Work in the desktop app

This is the one almost everyone skips, and it's the biggest. The browser version is fine, but the desktop app can reach the files and folders already on your computer. That means you can drop a file into a chat instead of copying and pasting it in by hand.

Go to claude.com/download, install it, and sign in with the account you already use. It's free and takes two minutes, and from now on you do your work in the app.


2. Write your profile instructions

Click your initials in the bottom left, go to Settings, and find the field asking what Claude should consider in every response. This one loads before anything else in every single chat, so it's worth getting right.

Copy this, fill it in and paste it. If a line doesn't apply to you, delete it instead of leaving it blank.

WHO I AM
I'm [name]. I [what you actually do, in one sentence].

HOW I WANT YOU TO WORK
Push back when I'm wrong instead of agreeing by default.
Ask me one question before you write anything long, don't guess.
Give me the work itself, not a description of it.

WHAT NOT TO DO
No em dashes. No emoji.
Never use these words: [your own banned list].
Don't [the pattern that annoys you most].

Aim for 150 to 400 words. Longer isn't better, because Claude reads all of it before it ever gets to your actual question.


3. Pick the right model for the job

Click the model name at the top of your chat and you'll see a short list. Most people never touch it. That's a shame, because it changes how good your answers are.

Opus 5 is the powerhouse. Save it for your hardest, highest-stakes work, like strategy, deep analysis or anything where a wrong answer actually costs you.

Sonnet 5 is the everyday default. It's what most of your work should run on.

Haiku 4.5 is the speed pick for quick lookups, reformatting and short answers you need right now.

Fable 5, if you see it, is the newest, so reach for it when you want the best on something big.

The names will change again in a few months, but the rule won't: heaviest model for the hard stuff, lightest for the quick stuff.


4. Connect the tools where your work already lives

Think Gmail, Drive, Calendar, Notion or whatever you actually live in. Connectors let Claude work on your real files and threads instead of you copying and pasting into a chat box. That's usually where the biggest time savings show up.

In the app, open the sidebar, click Customize, then Connectors, then the plus sign. Pick one and sign in through that app's own login screen, and you're done. Then do the same for the two or three tools you actually use every day.

You stay in control. Claude signs in through the app itself, never sees your password and asks before it does anything. It can draft an email but it can't send one, so that part is still on you. Skip anything tied to money or client data until you're comfortable, and know that you can disconnect any tool in one click.


5. Create a project, and work inside it

A project is a folder for one part of your life or work, like your clients, your content or your job search (whatever your areas actually are). Keeping them separate is what keeps this organized. If you mix your client work and your personal notes into one running chat history, everything gets slightly worse, because it's all pulling in context it doesn't need.

In the sidebar, click Projects, then New project. Name it for the area instead of the tool, so "client work" rather than "writing help." Add a few files you'd hand a new assistant, drop in the instructions, and from now on start every conversation about that area inside the project.

Turn on memory while you're in there by going to Settings, then Memory, then generate memory from your chats. Each project keeps its own memory separate from the rest, so your client work and your grocery list don't end up in the same bucket.

Starting every conversation inside the project is the step people skip, and then they wonder why nothing improved. For the full walkthrough, including a fill-in template for the instructions, see Set Up Claude So It Already Knows You.


None of this replaces asking clearly once you're in there. These five settings get you a Claude that already knows you, but what you say next is still on you.

Once these five are done, Part 2 covers where to click to make it actually build things: skills, published tools, and one running on a schedule.

Back to the free library