Claude AI now remembers everything: a step-by-step guide to using memory on the free plan

Until this week, every conversation with Claude AI started from scratch. You'd explain who you are, what you do, how you like to be written to, and the next day you had to do it all over again. Annoying, especially if you use Claude several times a day for recurring things.

Anthropic just changed that. Last week they enabled chat history memory for all accounts, including free ones. It's the biggest change of the year for someone getting started with AI. And most beginners still don't know how to turn it on or what to do with it.

In this guide you'll learn three things: what this new memory really is, how to turn it on in under two minutes, and five concrete ways to use it that will save you time starting tomorrow.

What Claude's memory is — and what it isn't

Claude's memory is a space where the AI stores information you consider important across conversations. You tell it your name, your job, the projects you work on, your writing preferences, and Claude remembers it when you come back two hours or two weeks later.

It's not a microphone. It doesn't record what you say in other apps. It doesn't track your browsing. It only remembers what you tell it inside Claude, and only what gets flagged as "worth remembering" (things about you and your goals, not trivial chit-chat).

And most importantly: you can see it, edit it and delete it whenever you want. If one day you decide you don't want Claude to remember you're a doctor, you go to settings and take it out. You're in charge.

The difference with ChatGPT is subtle but interesting: Claude's memory is designed to be more "editorial" and less invasive. Claude tells you when something is about to be saved, and you can veto specific memories. ChatGPT is a bit more automatic. Depending on how controlling you are with your data, one or the other will suit you better.

How to turn memory on in 2 minutes

Open claude.ai and log in. If you've never had an account, it takes 30 seconds to make one; the free plan is enough for this.

Once inside, look for your profile icon in the top right. Click Settings. In the side menu you'll see an option called Memory or Personalization. Go there.

You'll see an "Enable memory" toggle. Turn it on. Below it an empty box appears — that's your "current memory". It's blank right now because you just enabled it.

Save the changes and go back to the main chat screen. That's it. So much for the technical part.

If you want to start with a bang, type this as your first message:

"I want you to remember the following about me: my name is [your name], I'm [job], I live in [city], and I mostly use you for [list of 3-4 tasks]. When you write to me, I prefer concrete answers in plain English, no emojis, with real examples."

Claude will respond confirming what it saved. From that moment on, in any new chat, it'll already know these things without you repeating them.

5 ways to use memory that save real time

Most tutorials stop at "turn on memory and tell it your name". That works, but that's not where the value is. Let's get to the good stuff.

1. Your writing style as a permanent template. If you write emails, posts or reports with Claude, tell it once and for all how you want everything to sound. "I prefer short sentences, professional but warm tone, no jargon, zero emojis, and always British English, not American." From then on, every email it writes for you already comes in your voice. You skip the step of fixing the tone each time.

2. Your recurring projects. If you run a business, a blog or a thesis, tell Claude once: what it's about, who the audience is, what tone you use, what internal links you repeat. The next time you ask "write me a post about X", it'll understand the context without you reminding it that your site is about fitness or cooking or law.

3. Your tools and limitations. "I use Notion for notes, Google Calendar for appointments, I have a Mac but I don't pay for Adobe." When it suggests solutions, it'll go straight to things you can actually use. No more Photoshop proposals if you've never had it.

4. Your current priorities. Every 2-3 weeks, tell it what you're focused on. "This month I'm closing end-of-quarter sales; my priority is everything related to B2B leads." Claude will tailor its answers to that focus, and when your priority changes, you just update it. It's like having an assistant you brief on Monday morning and that's it.

5. Your decision history. This is the most underrated. When you make an important decision with its help — for example, "I'm going to bet on building a course instead of a book" — ask it: "remember I decided this today and for these reasons". Three months later, if you doubt, you ask "what did we decide about the course?" and you have the audit trail.

Typical mistakes when you first turn memory on

There are three things people do wrong and that are worth avoiding.

The first is dumping too much at once. If on day one you pour two pages of information into it, Claude will keep the main points but not all of it. It's better to add in small doses, as conversations happen. Every time you spot something useful, tell it "remember this" and done.

The second is not reviewing the memory. Every now and then, go to Settings → Memory and read what's in there. Sometimes Claude stores things that aren't true anymore (you changed jobs, you moved cities, you dropped a project). Spending two minutes a month cleaning memory keeps it useful instead of noisy.

The third is treating it like a database. Don't put passwords, card numbers or confidential company information in there. It's a context memory, not a safe. Treat what you put in like a conversation with a professor who knows you well: enough to be useful, not so much that an accidental leak gets you in trouble.

What about privacy?

Fair question. The short answer: Anthropic says memory is per user and isn't used to train their models by default. You can turn it off, delete specific memories, or wipe all of it in one click.

In practice that means if you're a professional in a sensitive industry (medicine, law, HR), you probably want to use memory only for operational context (your style, your workflow) and not for client data or cases. For that, use Claude in "no memory" mode (new chat, nothing remembered) or, if the topic is really serious, consider Team/Enterprise plans that come with extra guarantees.

For someone learning to use Claude for personal productivity, study or small business, memory on the free plan is more than enough and it's safe.

Combining memory with projects: the winning combo

Claude has had a feature called Projects for months (also available on the free plan with limits). They're like folders where you group related chats and upload reference docs.

Now that you have memory, the combination is very powerful. General memory holds your cross-cutting stuff (who you are, how you write). Projects hold the specific context of an initiative (a specific client, a book you're writing, an exam you're prepping).

Example: memory identifies you as a high school maths teacher in Manchester. Projects has a folder "School Year 2026-27" with the syllabus, the exercises and the groups you teach. When you enter the project and say "give me ideas for an activity on functions", Claude already knows who you are, what level you teach, what syllabus you cover and what you've already done. No repetition.

Concrete next steps

If you've made it this far, do this in the next ten minutes: go to claude.ai, turn memory on, and write your first "remember that..." with 3-4 basic things about you. Tomorrow, when you open Claude again, you'll feel the difference.

And if you want to take it up a notch, at learnaifast.io we have free courses that teach you to combine memory, projects, professional prompts and connectors so Claude becomes your real personal assistant, not an AI that asks you who you are every Monday.

Start with the free stuff, master the basics, and in a month you'll have tripled what you get out of the same hour of work.