Your first afternoon with Claude

You watched the video. This is the whole thing in one place: the handful of moves that take Claude from "a chatbot I sometimes ask things" to something that actually knows you and does real work.

It is built for a complete beginner. No code. No jargon. Everything here works on the free plan, and most of it you set up once and never touch again.

Give it an afternoon. By the end you will have:

• Claude set up so it remembers who you are, instead of meeting you cold every time
• Claude connected to your calendar and email, so it can actually do things, not just talk about them
• Your first Project: a workspace for the jobs you do again and again

The one rule, before anything else: Claude always asks before it changes anything. It shows you a plan first, and it never sends, deletes, or books without you saying yes. You stay in control, it just does the boring part. Knowing that, you can poke at everything without worrying you will break something.

1. Chat vs Projects (the one idea that makes it click)

This is the thing most people never get told, and it changes everything.

A regular chat is a brilliant stranger with amnesia. Every new chat, it has never met you. You explain who you are, what you are doing, how you like things, and then you close the tab and it forgets all of it. Great for quick, one-off questions.

A Project is that same brilliant person, except they have read your files and they remember the brief. You set it up once with the background and the rules, and every chat you start inside it already knows all of that. Great for anything you do more than once.

You will build your first Project in section 5. For now, just hold the idea: one-off question, use a chat. Something you repeat, use a Project. On the free plan you can have up to 5 Projects, which is plenty to start.

2. Make Claude actually know you (memory)

Most people use Claude like a stranger every single time. Memory fixes that permanently, and it is free.

With memory on, Claude quietly keeps a running note of the useful stuff: what you do, what is on your plate, how you like your answers. Next time you open a fresh chat, it already knows. You stay in charge the whole time: you can read exactly what it has saved, edit it, or wipe it.

Turn it on: in Claude (web, desktop, or phone), go to Settings → Capabilities → Memory and switch it on. It may already be on by default.

Then do the 15-minute setup. I wrote a free guided version that does the hard part for you. It interviews you for a few minutes, then hands you back a clean memory file: who you are, what you are juggling, how you like answers, and what Claude should never assume. Grab it here:

The 15-minute memory setup → open it on Notion

Open a new chat, paste it in as your first message, and answer honestly. The chat is private and the file is yours.

When it hands you the memory file at the end, copy it, go back to Settings → Capabilities → Memory, click Add to memory, paste, and save. From your next chat on, Claude knows all of it without you repeating a word.

Two things worth knowing:

• You can also just say it in the moment. Type "remember that I prefer short answers" mid-chat and it sticks.
• Want a one-off private chat that is never saved? Start it in Incognito (the little ghost icon).

3. How to prompt (the four moves)

Forget the giant lists of "100 prompt tricks." Beginners get 90 percent of the result from four moves. The difference between a weak answer and a great one is almost always that the great one had more context.

1. Give it context. Who you are, what you are trying to do, who it is for. "I run a small bakery and I am writing to regulars" beats "write a post" every time.
2. Give it a role. Start with "You are my..." Claude steps into it. "You are my patient accountant explaining this to me like I am new."
3. Show it an example. Paste one thing you like, a past email, a post, a tone, and say "match this." Examples teach faster than adjectives.
4. Tell it to ask first. End with "Ask me anything you need before you start." It will catch the gaps instead of guessing.

Here is the same request, weak then strong. The weak version: "Write a LinkedIn post about productivity." The strong version:

You are me: a physio who runs a one-person clinic. Write a short LinkedIn post about one small habit that saved me time this week. Keep it under 120 words, plain language, no hashtags. Here is a post of mine to match the tone: [paste]. Ask me anything you need before you write.

You do not need to memorise this. Just keep asking yourself: have I told it enough that a smart new hire could do the job? If not, add the missing piece.

4. Give Claude the keys (connectors)

On its own, Claude is blindfolded. It cannot see your calendar, your inbox, or your files, so you end up copy-pasting everything by hand. Connectors take the blindfold off.

Set it up once (do this on a computer): open a chat, click the plus (+) near the message box, hover Connectors, then Manage connectors. Find Google (Calendar, Gmail, Drive), click Connect, sign in, and approve. On Microsoft or Outlook, same idea, pick those instead. This works on the free plan.

Then prove it works, the photo trick. This is the bit that makes people go "wait, it just did that":

1. Take a photo of anything with a date on it: a gig poster, a class flyer, a kid's party invite, even a screenshot of a text.
2. Drop the photo into Claude.
3. Type "add this to my calendar."

It reads the date, time, and place straight off the picture and offers to drop it in. You confirm, and it is done. You did not type a single detail.

Once it is connected, try:

"What is actually on my calendar tomorrow?"

"Find the email from [name] about [thing]."

"Pull up that doc about [topic] from my Drive."

The one rule again, because it matters most here: Claude reads freely once you allow it, but it is careful with actions. It will draft an email and leave it sitting in your Gmail for you to send, it does not send for you. It will create a calendar event, but it asks before each one. You always get the final yes.

5. Build your first Project (and fill it with knowledge)

This is where Claude stops being a chatbot and becomes your thing.

Pick one job you do over and over. Writing weekly updates. Replying to customer emails in your voice. Planning the family week. Drafting posts. Any of them work.

Build it:

1. Go to claude.ai/projects (or click Projects in the left sidebar), then + New Project. Give it a name.
2. Add the knowledge. Click the + inside the project and upload the documents Claude should know: a few past examples, your style notes, a price list, last month's report. These become the background for every chat in the project. (You can add as many files as you like; each one up to 30MB.)
3. Set the instructions. Click Set project instructions and write the standing rules in plain English: "You are my marketing assistant. Write in a warm, direct voice. British spelling. Always ask before assuming a date." Save it.
4. Start a chat inside the project and just ask. It already has the files and the rules, so you skip the whole re-explaining ritual.

One thing that trips people up: the project name is just a label, Claude does not read it as an instruction. Anything you actually want it to follow goes in Set project instructions.

Do this once for a single real job and you will feel the difference immediately. Then build the next one.

6. Skills, and the free Anthropic library

Last one, and it is the door to everything after this.

A skill is a small instruction pack that teaches Claude how to do one specific job really well, so it gets it right without you explaining how each time.

The free unlock you can use today: go to Settings → Capabilities and switch on Code execution and file creation. Now ask, in a normal chat, for a real document:

"Make me an Excel budget for a household, with the formulas working."

"Turn these notes into a clean five-slide PowerPoint."

"Write this up as a Word document I can download."

Claude builds the actual file and hands it back to download. No template, no formatting fight. It picks the right skill on its own, you just describe what you want. This works on the free plan.

The reveal: those document skills are just the start. Anthropic publishes a whole library of skills, and the full set is public at github.com/anthropics/skills. Inside Claude, the broader Skills area (browsing more of them and adding your own) lives under Customize → Skills, and the fuller toolkit sits on the paid plans. That is your on-ramp from "using Claude" to "building your own Claude."

One safety note: only ever add a skill from a source you trust. Treat it like installing software, because a bad one could tell Claude to do things you would not want. The built-in Anthropic ones are safe ground to start.

What now

That is the whole setup. In one afternoon you went from typing at a chatbot to something that:

• Knows who you are and how you like things (memory)
• Can reach into your calendar, inbox, and files (connectors)
• Has a dedicated workspace for the jobs you repeat (your first Project)
• Can hand you finished documents on request (skills)

Pick one thing and actually use it today. That is the part that sticks.