Getting started with Claude

I opened claude.ai on a saturday because I had a landlord email sitting in my inbox and I didnt want to answer it yet. Not because I wanted a new tool. The email was polite and still somehow made me feel behind on a thing I did not agree to.

If you already bounced around another chat app this will feel familiar. Account, box, type. The long messy paste can wait.

This is step 1 of a short series. If you have not even named the problem yet, go read You dont need another tool first. Picking Claude vs Cursor vs Grok is a later question, and Which tool when is there when you want the honest version.

Make the account and stop decorating it

Go to claude.ai in a normal browser. Anthropic's own start page says you can also use a desktop app or the iOS and Android apps, and you have to be 18 and in a supported location. I started in the browser. That is enough.

You will see two login paths. Continue with Google, or type an email and hit Continue with email. The email path sends a message with the subject "Secure link to log in to Claude.ai". There is no password to invent. I spent a minute looking for the password field anyway because that is what every other site trained me to do.

Use the email you actually check. You can use the same account later on your phone by entering that same address. If the login email does not show up, it is usually sitting in junk from @mail.anthropic.com.

When you get in you are looking at a chat. Type in the box and hit the submit button. That is the whole first session. The leftover chrome can wait.

There is a + button on the lower left of the box, and you can type / to see extra options. Fine. Do not go hunting through settings. Do not start a Project. Do not download the desktop app unless you already wanted an app. Desktop is real on Mac and Windows, Linux still in beta, and chat works on free. You still do not need it tonight.

Under the text box on web and desktop you will see which model you are talking to. You can click it and switch if your plan lets you. I would not pick a model on day one like you are buying a car. Use the default, send one real thing, see if the reply is even in the right zip code.

What to paste the first night

Keep the first ask small and real. Anthropic's own getting-started note says start simple and be specific. That is the one bit of official advice I actually followed.

A good first paste looks like this. You copy the landlord email, the one that mentions a "routine inspection" and also a date that does not work and also a sentence about "access if you are not home." You drop the whole thing in the box. Then you write one sentence under it.

Something like: tell me in plain language what they are asking me to do, what I can say no to, and draft a short reply I can edit.

That is a job. "Help me with my apartment" is not a job. "Write a professional email" with no source is how you get a letter that sounds like a hotel.

Other first asks that worked for me or people I know:

Paste three messy bullet notes from a meeting and ask Claude to turn them into a list of decisions and open questions. Not minutes. Decisions.

Paste a half-written product paragraph from your site and ask what a stranger would still not understand.

Paste a school email about a form and ask what the deadline actually is, quoting the line.

I tried the empty prompt first, once, because I was tired. I typed "I need to be better about email." Claude wrote a little essay about inboxes. It was fine. I closed the tab and still had the landlord email.

Why small is not you being timid

The free plan is a real plan. Anthropic lists Free, then Pro at $20 a month or $200 a year, then Max tiers above that. Free has a session usage limit that resets every five hours. How many messages you get inside that window moves around with demand and how long the chat gets. Claude will tell you when you are done.

So if you spend the first hour asking it to invent a content calendar, you will hit the wall before you get to the email you actually have to send. I did this. It feels productive. It is not.

Small also means you can tell if the answer is wrong. If you ask Claude to rewrite your whole business, you cannot check it. If you ask it what the landlord wants by Tuesday, you can.

A first chat I still like: paste the thing, ask for a short reading of it, then ask for a draft. Two turns. You stay in the same thread. You do not open a second tab to "try a better prompt."

If the reply is stiff, say so. "Shorter, and I do not want to sound like a lawyer." Claude takes that kind of note pretty well. If it invents a building policy you never mentioned, that is your warning for step 4. Even tonight, do not send the draft without reading it against the original email.

Things I would skip until next week

Projects. They exist, they are useful, free accounts can make up to five of them. Next week. Tonight you have one chat.

Artifacts, the extra window that opens on the right when Claude makes a longer standalone thing. Cool when you want a document you can poke at. Not the first night.

Saved instructions in Settings. Same story.

Uploading a folder of PDFs. You can attach files later with the + button and Add files or photos, and you can drag a file onto the chat. Day one I would rather paste the one page you care about so you can see what Claude does with it.

Claude Code, Cowork, connectors. Current products. Also how people delay the email.

If your actual work lives in a repo and you need an editor, that is Cursor, not this tab. If you need a current take on something that changed last week, Grok is often the less painful start. You can still use Claude tonight for the email.

A first night that does not waste the limit

Sit down with one thing you were going to deal with anyway.

Paste the source. Ask for a plain reading. Ask for a draft you will edit. If the draft is close, tell it which sentence is wrong. If the draft is generic, paste two lines you actually sent last month and say write more like that.

Then stop.

Close the tab before you ask it to also rewrite your about page. Copy the draft into a note so you still have it if the chat gets lost in the sidebar tomorrow.

I still do this. The first useful Claude session I had was 12 minutes and a reply I changed four words in. The long sessions came later, when I already knew it would not just invent a lease clause.

Next is give it the mess, which is the part most people skip because pasting the real notes feels embarrassing. After that the boring middle is where the work actually happens.

If you only do this one post, do the landlord email or the meeting notes. Leave the rest of the product alone until you have one reply you would actually send.