How To Set Up Claude As Your AI Assistant: A Beginner's Guide

9 min read

Short answer: to set up Claude as your AI assistant, make a free account at claude.ai, install the desktop and phone apps, turn on memory and chat search in Settings, make one Project for one area of your life or business, and give Claude its first job with the goal, the context and the finish line spelled out. The rest of this guide shows each step and the details most people miss.

I run a one-person, home-based phone repair shop in San Diego. For a long time my AI help was a team of Grok Bots (I wrote that whole Grok Bot series). Then I sat down with Claude for one evening and it fixed my store's shipping weights, rewrote 150 search descriptions, cleaned up my warranty wording on every page and rebuilt parts of my website theme, checking its own work as it went. That evening is why this series exists. I want you to skip my learning curve.

What Is Claude, In Plain Words?

Claude is an AI assistant made by Anthropic. You talk to it like a very capable coworker: it reads, writes, researches, works with your files and, on paid plans, can work inside your browser and your apps. It runs in one account across several places:

  • claude.ai in any web browser.
  • The Claude desktop app for Mac and Windows. This is where it can reach files on your computer.
  • The Claude phone app for iPhone and Android, including voice mode.
  • Claude Code, a version built for software and website work (paid plans).
  • Claude in Chrome, a browser extension that can read, click and fill in websites with you (paid plans, source).

Everything you do in those places shares one account and one usage allowance (source). Start a chat on your laptop and you can pick it up on your phone.

Chatbot vs assistant: a chatbot answers questions. An assistant does work. The difference with Claude is mostly in how you set it up: a Project with your rules and files, memory turned on, and the right apps connected turns a chatbot into an assistant that knows your business.

What Do You Need To Get Started?

  • An email address or a Google account.
  • A phone number for verification.
  • About 20 minutes.
  • One real task you'd like off your plate this week. This matters more than anything else on the list.

You can start on the Free plan. It already includes chat, web search, file creation, Artifacts, memory, connectors and up to 5 Projects (source). If you use it daily for work, you'll likely want Pro at $20 a month ($17 a month billed yearly). It adds more usage, the Opus model, Research, Claude Code, Claude in Chrome and scheduled tasks (source). Part 2, Claude Plans, Models And Usage Limits, breaks down every plan.

How Do You Set Up Claude, Step By Step?

1. Create your account

Go to claude.ai, sign up with email or Google and verify your phone number. Then install the desktop app and the phone app and sign in with the same account. Do this on day one. The desktop app is what lets Claude reach files on your computer later.

2. Tell Claude who you are (profile preferences)

In Settings, fill in your name and your preferences. Write them like a note to a new employee. Mine boils down to: "I run a home-based phone repair shop in San Diego. I'm not a programmer. Use plain English, short answers unless I ask for detail, and ask me before changing anything live." Claude reads this in every chat, so every answer starts already knowing how you like to work.

3. Turn on memory and chat search

Claude can remember durable facts about you across chats and search your past conversations when you refer to them. Both are switches in Settings. Turn them on unless you have a reason not to. You can view, edit or delete what it remembers at any time, and an incognito chat skips memory entirely (source). Part 3, How Claude Remembers, goes deep on how this works.

4. Make your first Project

A Project is a folder for one area of work: its own instructions, its own files and its own chats. Click Projects → New Project, name it for one job ("Shop Front Desk", "Bookkeeping", "Website"), then:

  1. Write short project instructions: who it works for, what good looks like, what it must never do.
  2. Add the 3 to 10 files it needs most: your price list, policies, a sample of your writing.
  3. Start every related chat inside that Project.

Keep instructions short and remove files you no longer need. Anthropic says big instructions and unused files eat into your usage (source).

5. Connect one app (optional on day one)

Under Settings → Connectors you can link apps like Gmail, Google Calendar, Google Drive and many others so Claude can read from them and, where allowed, act in them. Start with one. Connect more when you have a job that needs them. Part 4, Let Claude Work In Your Apps, covers connectors, Claude in Chrome and scheduled tasks.

6. Pick the right model for the job

At the bottom of the chat box you can choose the model. As of November 2026 the current family is Haiku (fastest, lightest), Sonnet (fast and smart, the everyday workhorse), Opus (the default pick for serious work) and Fable (the most capable, for the hardest problems) (source). Bigger models use more of your allowance. Use the smallest one that does the job well.

How Should You Word Claude's First Job?

Most people's first message is a question. Make yours a job. Give Claude four things:

  1. The goal: what done looks like.
  2. The context: who it's for, what you already tried, where the facts live.
  3. The limits: what it must not do or say.
  4. The format: how you want the answer back.

A First Message, Written Out

S
SteveWrite the text reply I send when someone asks "how much to fix my iPhone 14 screen?" Use the price list in this Project. Goal: they book or reply with their model. Keep it under 60 words, friendly, no exclamation points, mention walk-ins are by call or text first. Give me two versions.
C
ClaudeTwo versions below, both pulled from your price list. One question first: should the reply list all four screen tiers, or just the starting price and a link?

Notice Claude asked a question back. That's a good sign, not a failure. A good assistant asks when a wrong guess would cost you. Answer it, and save the version you like as a file in the Project so it's there next time.

What Are The Details Most Beginners Miss?

Chats have a length limit, and that's different from your usage limit

Every chat has a context window: how much Claude can hold in mind at once. On paid plans the newest models handle up to 1 million tokens (very roughly 750,000 words), and older ones 200,000 or 500,000 (source). When a chat gets very long, Claude can summarize the earlier part automatically so you can keep going. Separately, your usage limit resets every five hours and weekly on paid plans. Long chats burn usage faster because Claude rereads the whole conversation each turn.

The habit that fixes both: one chat per task. When the task is done, start a new chat in the same Project. The Project carries your rules and files forward; the chat stays short.

Memory, Project instructions and your profile do different jobs

  • Profile preferences: how you like to work, everywhere.
  • Project instructions and files: the rules and facts for one area.
  • Memory: durable facts Claude picks up from your chats, like "walk-ins are by call or text first."

Put each rule in exactly one place. If a rule lives in all three and you change one, Claude gets mixed signals.

Claude asks before it acts

When Claude can act in your apps or browser, it's built to ask before anything you can't easily undo: sending a message, submitting a form, buying something, deleting. You'll see a confirmation and it waits. Leave that on. It's what makes it safe to hand over real work.

It can be wrong, so make checking part of the job

Claude is very good, not perfect. For anything that matters (prices, legal wording, tax numbers), add one line: "Check your answer against the file and tell me which line each number came from." That single sentence catches most mistakes before they reach a customer.

What Are The Most Common Setup Mistakes?

  • One giant chat for everything. It slows down, costs more usage and mixes topics. Use Projects and one chat per task.
  • Vague first messages. "Help with my business" gets generic advice. A job with a goal, context and limits gets work done.
  • Uploading everything. Ten good files beat a hundred random ones.
  • Using the biggest model for small jobs. Sonnet or Haiku handle quick rewrites and lookups and leave room for the hard work.
  • Not saving what works. When a reply is great, save it to the Project as an example. Examples teach Claude your style faster than any instruction.

Which Starter Prompts Can You Copy?

Interview me: "Before we do any work, interview me with 10 questions about my business, my customers and how I like to communicate. Then write project instructions I can paste in, under 200 words."

Inbox triage (with Gmail connected): "Look at my last 50 emails. Group them into: needs a reply today, can wait, newsletters. For the first group, draft replies but don't send anything."

Price check: "Here's my price list. Find any prices that look inconsistent between similar models and show me a table: model, current price, what looks off, suggested fix."

Weekly plan: "Here's my calendar and to-do list. Build my week so the urgent work happens first, leave an hour a day open, and flag anything I've double-booked."

Your Turn: Get To Work

  • ☐ Make your account and install the desktop and phone apps.
  • ☐ Write your profile preferences like a note to a new hire.
  • ☐ Turn on memory and chat search.
  • ☐ Make one Project for one job, with short instructions and 3 to 10 files.
  • ☐ Give Claude one real job using goal, context, limits and format.

Frequently Asked Questions

Is Claude free to use?

Yes. The Free plan includes chat, web search, file creation, Artifacts, memory, connectors and up to 5 Projects, with limited usage. Paid plans start at $20 a month for Pro (source).

Do I need to be technical to use Claude?

No. You talk to it in plain English. Claude Code is the developer-focused part, and you can skip it entirely until you want it.

What's the difference between a chat and a Project?

A chat is one conversation. A Project is a folder that holds instructions and files and is shared by every chat inside it, so you don't repeat yourself.

Does Claude remember me between chats?

Yes, if memory is turned on. You can see and edit what it remembers in Settings, and incognito chats don't use memory (source).

Which Claude model should I use?

Sonnet for everyday work, Opus for serious multi-step work, Haiku for quick small jobs, and Fable when nothing else gets it right. Bigger models use more of your allowance.

Is Claude better than Grok Bot or ChatGPT?

It depends on the job. I compare them honestly in Grok Bot Vs Other AI Assistants and What Is Claude Good For?.

More In This Series

  1. How To Set Up Claude As Your AI Assistant (you're here)
  2. Claude Plans, Models And Usage Limits Explained
  3. How Claude Remembers: Projects, Memory, Skills And Instructions
  4. Let Claude Work In Your Apps: Connectors, Chrome And Scheduled Tasks
  5. How To Prompt Claude Like A Pro

Unofficial guide written by a Claude customer. Not affiliated with or endorsed by Anthropic. Plans, prices and features checked against Anthropic's pages on November 17, 2026 and may change.

Back to blog