I don’t like repeating myself. Never have. If I tell somebody how I want something done, I want to tell them once.
That’s one of the main reasons I’m making this whole series. Generic AI forgets you the second the chat ends. It gives you the same bland answer it gives everybody else. The version of Claude I want you to have knows your business, knows your clients, knows how you talk, and gets sharper every week you work with it.
That comes down to memory. Here’s how it works, and the simple habit that makes it work a lot better.
What Claude remembers about you
Claude Code keeps notes on what it learns about you and your business, and it reads them every session. It does a great job of this out of the box. Memory is on by default, so it’s already capturing things as you go.
What it keeps falls into four buckets:
- About you. Your role, your business, how you like answers. “I’m a solo coach, keep replies short.”
- How you like things done. Your corrections, and what you said yes to. “Never send without showing me the draft.”
- What you’re working on. Projects, deadlines and decisions you made. “New program launches March 1.”
- Where things live. Which folder, list or link something belongs in. “Renewals go in the Renewals list.”
That last one is underrated. It’s so helpful when Claude just knows where your stuff is instead of you having to point every time.
Where memory lives
Your memory isn’t some black box in the cloud. It’s plain text files on your own computer that you can read, change and take with you.
Each thing Claude learns gets its own small file. There’s an index file called MEMORY.md that lists them all. Every new chat starts by reading that index, and Claude opens the full note when it’s relevant.
They live in a hidden folder, because the folder name starts with a dot. On a Mac, open Finder, press Command + Shift + G, paste ~/.claude/projects/, and press Return. Open the folder named after your workspace, then memory. On Windows, paste %USERPROFILE%\.claude\projects\ into File Explorer’s address bar.
You don’t need to go in there often. But I want you to know it’s yours.
The three parts of Claude’s memory
When I talk about memory, I’m really talking about three things working together:
- CLAUDE.md. A file you write with your standing instructions. Claude reads it at the start of every session. Keep it short and point to other files for details. There’s a whole video on this coming in the series.
- Memory. The notes Claude writes as you work: your rules, your clients, your preferences.
- Hooks. A hook is a “when this happens, do that” rule that fires every time. A rule in memory is something Claude usually follows. If it can never be missed, make it a hook.
The trigger words that make it stick
Here’s the habit. Out of the box, Claude decides on its own what’s worth remembering, and it won’t get it 100% right. But if you use the right words, you’ll nail it.
When you want something kept, say it with one of these:
- Always… “Always use the client’s first name in check-in replies.”
- Never… “Never message a client without showing me the draft first.”
- Remember… “Remember that Kayla is training for a half marathon in March.”
- Don’t forget… “Don’t forget to tag new leads ‘Consult booked’ in my CRM.”
- From now on… “From now on, save every new program in the Programs folder.”
- Whenever… “Whenever a client misses two check-ins, draft a nudge for me.”
When you use language like this, Claude is a lot more likely to grab it and save it right then.
Talk to it like a new hire on day one. Say the rule once, out loud, and it’s kept.
Make it reliable: one paragraph in your CLAUDE.md
Funny thing: Claude is actually the one that taught me to use those words. Then I took it a step further and put them in my CLAUDE.md, so it saves every time I use them and tells me what it saved.
It’s a short amount of text for a big bang for your buck:
## Memory
When I say "always", "never", "remember", "don't forget", "from now on", "rule:", or "whenever this happens, do that", save it to memory right away. Don't wait until the end of the session.
Then confirm in one line: "Saved to memory: [what you saved]."
If what I said changes something you already saved, update that note instead of adding a second one.That confirmation line is the key. Now you know it stuck.
Check it, fix it, tidy it
Memory isn’t set it and forget it. You can quiz Claude any time:
- Check: “What do you remember about my check-in process?”
- Fix: “Update that: nudges go out after three missed check-ins, not two.”
- Forget: “Forget what I said about the Programs folder.”
- Tidy: “Review your memory. Merge duplicates and remove anything out of date.”
I’d do a tidy every so often, the same way you’d clean up your client list.
Coming from ChatGPT?
You don’t have to start from zero. Claude has an import memory option that brings over what ChatGPT already knew about you. Look for it when you type /, or in settings. So if you’ve been living in ChatGPT and you’re thinking “oh no, I have to start over,” you don’t. Good news.
Do this today
Give it 15 minutes:
- Add the memory paragraph above to your CLAUDE.md. Don’t have one yet? Open your folder in the Code tab and ask Claude to create a CLAUDE.md with that paragraph in it.
- Say three rules out loud. One “always,” one “never,” one “from now on.” Pick things you’re tired of repeating. Watch for the “Saved to memory” confirmation.
- Quiz it. Ask, “What do you remember about how I run check-ins?” Fix anything that’s off.
- If you’re coming from ChatGPT, run the import so your history comes with you.
- Pick the one rule that can never be missed and make it a hook instead. The unread-message guard in the Skills Kit is a great first one.
Get the cheat sheet and watch the full video
The Teach Claude to Remember cheat sheet, with every trigger word and the CLAUDE.md paragraph ready to copy, is in the Coach Packet in the free guide at thefmm.click/guide. You also get the Skills Kit, with ready-made skills and hooks to install.
Watch the full video: Claude Code for Trainerize Coaches, the complete walkthrough
New here? Start with Trainerize AI: what it does and what it still can’t do, then come back to this one.
AI that amplifies you. Always authentic, never generic.
