How I set up my Claude memory in less than 15 minutes (switching from ChatGPT)

Claude just made switching free. Here’s how to set it up so it actually knows you.

Mar 04, 2026

This article can help someone you know use AI to earn more and stress less. Forward it and make their day.

Hey Adopter,

Anthropic dropped something big yesterday. Claude’s long-term memory, previously locked behind a paid subscription, is now free on all plans. And they shipped an import tool that lets you pull your entire ChatGPT history into Claude with a single copy-paste.

Claude is sitting at #1 on the App Store right now. ChatGPT started showing ads to free users last month. The lock-in game is over.

But here’s what most people will get wrong: they’ll dump their ChatGPT data into Claude, skip the setup, and wonder why the outputs feel generic. Switching without configuring is like moving into a new office and never unpacking.

This piece walks you through exactly how I set up Claude from scratch, step by step, so it works like a sharp colleague from day one.

Share AI Adopters Club

The new hire you never onboarded

You wouldn’t hand a contractor a laptop and say “figure it out.” You’d brief them. You’d tell them how the team communicates, what your priorities are, what you hate seeing in deliverables.

Claude is no different. Out of the box, it gives you polished, cautious, middle-of-the-road answers because it has zero information about who you are, what you do, or how you think.

The setup takes 15 minutes. The difference in output quality is enormous.

Step 0: enable memory

Before anything else, turn on memory. Go to Settings → Capabilities → Memory and toggle it on.

Without this, Claude forgets everything between conversations. Every chat starts cold. You’re re-explaining your job, your preferences, your writing style, over and over.

Memory lets Claude build a working picture of you across conversations. It notices patterns. It remembers your corrections. Over time, it stops making the same mistakes.

You can always check what Claude has stored by clicking “View and edit your memory” in the same settings panel. Delete anything you don’t want kept.

Settings → Capabilities → Memory → Toggle ON

Step 1: import your ChatGPT brain (if switching)

If you’ve been using ChatGPT, you have months of stored preferences and history sitting there. Anthropic built an import tool to grab it.

Their default export prompt is decent:

“I’m moving to another service and need to export my data. List every memory you have stored about me, as well as any context you’ve learned about me from past conversations.”

I use a more structured version. Paste this into ChatGPT instead:

Export all of my stored memories and any context you've 
learned about me from past conversations. Preserve my 
words verbatim where possible, especially for instructions 
and preferences.

## Categories (output in this order):
1. Instructions: Rules I've explicitly asked you to follow 
   going forward, tone, format, style, "always do X", 
   "never do Y", and corrections to your behavior.
2. Identity: Name, age, location, education, family, 
   relationships, languages, and personal interests.
3. Career: Current and past roles, companies, and general 
   skill areas.
4. Projects: Projects I meaningfully built or committed to. 
   ONE entry per project. Include what it does, current 
   status, and any key decisions.
5. Preferences: Opinions, tastes, and working-style 
   preferences that apply broadly.

## Format:
Use section headers for each category. One entry per line, 
sorted oldest first. Format each line as:
[YYYY-MM-DD] - Entry content here.
If no date is known, use [unknown] instead.

## Output:
Wrap the entire export in a single code block for easy 
copying. After the code block, state whether this is the 
complete set or if more remain.

Heres what ChatGPT’s response looked like:

if you ever set up your ChatGPT properly I would also copy your custom instructions and more about you section into the next part.

Copy ChatGPT’s response. Go to Claude’s memory import page, paste it in, done.

One thing to know: Claude’s memory focuses on work-related information and may not retain every personal detail you import. That’s fine. The structured prompt above front-loads the professional data Claude cares about most.

Step 2: tell Claude who you are and how you think

Claude has three personalisation layers and most people don’t touch any of them. Each one does something different.

2a. Your profile, the global bio

Go to Settings → Profile. The first text box asks what Claude should know about you. This is your bio. Your role, industry, team size, the tools you use, any constraints you work under. It applies to every conversation you have.

A vague “I work in marketing” gets you vague answers. Compare that to: “I’m a marketing manager at a 50-person B2B SaaS company, running a team of four, focused on pipeline generation through content and paid channels.” The second version means Claude already knows your world before you type your first question.

This 2-minute investment means you never re-explain your job, your tech stack, or your approval process again.

2b. Preferences, the global rulebook

The second text box on the same page asks what preferences Claude should follow. This controls how Claude talks to you. Tone, formatting, communication style, things to always do, things to never do. It’s applied across every chat.

Claude’s default personality is polished and agreeable. Fine for casual questions. Useless when you need someone to challenge sloppy thinking.

Most people stare at that empty box and close the tab. Don’t. Paste this into a fresh Claude conversation instead:

Ask me what tone, style, and level of directness 
you should use when responding to me. Check if I 
want you to challenge my thinking or support my 
ideas. Find out my writing preferences, formatting 
rules, or words to avoid. Ask if I want practical 
advice or creative ideas, and what my main goal is 
for our interactions. Check for any personal habits 
or biases you should flag, and if there are any 
rules you should always follow.

Ask ONE question at a time to help me articulate 
my ideal working style. When we're done, create a 
concise summary I can paste into my User Preferences 
in Settings.

Claude will walk you through 5-8 questions. At the end, it hands you a clean block of text. Copy it, paste it into your preferences, done.

For reference, here’s what I use (sanitised):

Operate with a confident, clear, and dry-humoured 
voice. Never saccharine, never verbose. Be critical 
but constructive, and take a strategic, 
action-oriented approach in every response. 

Never use the em dash. Use commas or full stops.

When I present an idea, don't play the role of a 
passive assistant. Your job is to sharpen my 
thinking: identify assumptions I'm making, present 
the best counterarguments a well-informed sceptic 
might raise, and stress-test my logic for gaps. 

Offer alternative perspectives where relevant. 
Prioritise truth and clarity over agreement. If my 
thinking is off-base, say so directly and without 
sugar-coating. 

Be practical above all. Get straight to the point. 
Don't shy away from strong opinions when they're 
warranted. If you see confirmation bias or lazy 
assumptions creeping in, call them out.

You don’t need to copy mine. The guided prompt above builds one that fits the way you work.

2c. Styles, the delivery layer

The third personalisation feature is Styles, found under Settings → Styles. Preferences control what Claude says. Styles control how it formats and delivers the response.

Pick from Claude’s built-in options or create a custom style. I run a custom one that keeps responses tight and avoids filler. The key feature: you can switch styles per conversation. Writing a long report? Use your detailed style. Quick Slack reply? Switch to concise.

Think of preferences as your standing orders. Styles as your dress code. Both matter. Preferences rarely change. Styles switch depending on the task.

Step 3: create your first project

Projects are where Claude gets truly useful. A project is a contained workspace with its own instructions and reference files that Claude follows every time you open a chat inside it.

Go to Projects in the left sidebar and create one for your main work area.

Add project instructions that tell Claude your role, your team, your recurring tasks, and any rules specific to that work.

You are helping me manage content marketing 
for a B2B SaaS company (50 employees, Series B). 

My team: 2 writers, 1 designer, 1 SEO specialist.
Our channels: blog, LinkedIn, email newsletter.
Tone: professional but not corporate. Clear, 
direct, no jargon.
Current priority: launching a product comparison 
page by end of Q1.

Upload relevant files, style guides, brand documents. When you ask Claude a question inside this project, it already knows the answers to a dozen questions you’d otherwise spell out every time.

This is where you can find your projects and set them up. This is just scratching the surface now that you have switched to Claude. I don’t even want to start with Cowork or Code in this edition because it would be way too much ground to cover.

Step 4: fine-tune with styles

Claude has a Styles feature under Settings that controls response formatting. Choose from built-in options or create a custom style. I run a custom one that keeps responses tight and avoids filler.

You can switch styles per conversation. Writing a report? Use your detailed style. Quick Slack reply? Switch to concise.

Are you going to do it?

Four steps. Under 15 minutes total.

Enable memory in Settings → Capabilities. Takes 10 seconds.
If you’re switching from ChatGPT, run the enhanced export prompt above, paste the result into Claude’s import tool. 3 minutes.
Write your user preferences. Start with three things that annoy you about AI responses and turn them into instructions. 5 minutes.
Create one project for your primary work area. Add a 4-5 sentence brief and upload one reference file. 5 minutes.

Discussion about this post

Ready for more?