Warning: this will make you fall in love with Claude Code

Somewhere for Claude Code to keep everything you tell it

Tell it what happened in a meeting. It writes the note, files it under the right client, and puts what you agreed into your live work. Tomorrow it still knows all of it. It wrote it down, so you never say any of it twice. Underneath is a structured folder of plain files on your PC, and an assistant that reads and writes them while you talk to it. It turns up built, with six months of daily use already spent on working out what goes where.

Be first in line when checkout opens

Launch is close. Leave your email and you hear on the day it opens, with the price holding at US$500. Your address is used for that one email and nothing else.

Free setup guide

C:\Users\you\Documents\MYOS-Vault

MYOS-VAULT

_Inbox

_System

_Tasks

Active

Start here

Choose your starting point. Both answers land on the same setup line, and that line skips anything you already have installed.

Do you already have Claude Code running on this Windows PC?

Yes, it runs on this PCNo, or I am not sure

Then most of this page is written for someone who has not got that far yet. The fast lane has what MYOS is, the ninety seconds of setup you still have to do, and the one line that drops it into the Claude window you already have open. Take the fast lane.

Then start at the beginning. The setup guide installs Git for Windows, Claude Code and Obsidian for you, walks you through signing in, and stops when your machine is ready. You do not have to learn PowerShell. You copy one line, paste it into a black window, and read what comes back. Go to the free setup guide.

The fast lane

What a Claude Code for non-coding work looks like

Skip the install walkthrough below. The harder half is finding something worth pointing Claude Code at, and that is why this page exists.

akhilnchauhan on Hacker News described using Claude Code as driving a Ferrari to pick up some groceries. mibii, writing up a Claude Code and Obsidian setup on dev.to, named the other half: it forgets everything between sessions, so every conversation starts from scratch. MYOS answers both at once. It gives Claude a place to keep what it learns about you, and a standing job big enough to be worth the tool.

A structured folder of plain files on your PC, and an assistant already configured to read and write them while you talk. Buy a template instead and you get the empty shape. This one comes with the structure made, the loop tested and the assistant already told how to keep it current, so it runs itself instead of joining the list of things you maintain.

All the setup there is

One line, once, in PowerShell. On a machine that already has everything, it takes about ninety seconds. It checks what is present, skips it, installs whatever is missing, and then proves your Claude plan can actually run Claude Code, which is the thing that quietly catches people out later. That line is step 2 of the free guide. Nothing before it applies to you.

After that you stay in Claude

You do not go back to a terminal. Buying sends you a receipt with one plain-English line in it, and that line goes into the Claude window you already have open:

Claude

Fetch https://learnmyos.com/setup/your-order-link and follow it exactly to install my MYOS vault.

PLACEHOLDER: that address is not live yet, and the link shown is illustrative because each order gets its own. The line itself, and the helper that makes it work even if you paste it into PowerShell by mistake, are built and tested. Hosting them at a real address is a separate step.

Claude reads the instructions at the other end, installs the vault, explains each prompt before it appears, restarts itself inside your new vault and starts asking you about your work. You tell it what you do, in your own words, and it does the rest. You never write code and you never open a configuration file.

The rest of this page is the long version: what it runs on, the free guide, what you own, the price and the questions.

What it runs on

All five matter. Buy without one of them and it will not run, and you will be asking for a refund.

A Windows 10 or Windows 11 PC. MYOS is Windows only. There is no Mac version and no Linux version, and the Mac waitlist page carries no release date.

Your own paid Claude plan, Pro or above. That is about US$20 a month, it goes to Anthropic and not to us, and it is not part of the US$500. Claude Code will not run at all on the free plan, and the setup script stops and tells you so before you have spent anything here.

Git for Windows and Obsidian. Both are free and the setup script installs them for you. Obsidian is free to use for work as well as at home.

Permission to install software on the machine. A locked-down work laptop is the most common reason an install fails.

About half an hour, for the setup and your first conversation with the assistant.

The free guide below checks all five and finishes on a READY banner. Do that part first. It costs nothing, and it is how you find out your machine is up to this without giving anyone money.

Free setup guide

Set up Claude Code on Windows

It installs what Claude Code needs, gets you signed in, and proves your plan can actually run it. Nothing in this section asks you for money and nothing in it expires. If you never buy MYOS, this still leaves you with a working Claude Code on your machine.

Step 1: open PowerShell

Press the Windows key, type PowerShell, and press Enter. A black window opens. That is the only window you need.

Step 2: paste one line

Copy this, paste it into the black window, and press Enter.

irm https://learnmyos.com/get-ready.ps1 | iex

PLACEHOLDER: this address is not live yet. The script exists and is tested; hosting it at this address is a separate step, so the line above cannot work from this draft.

It will ask Windows for permission a couple of times along the way. Say yes. Here is what it does, in order:

Checks your Windows version and free disk space.
Installs Git for Windows, Claude Code and Obsidian, skipping anything already present.
Repairs %USERPROFILE%\.local\bin on your PATH, and refreshes the PATH in the window you are standing in, so claude works there rather than only in the next window you open.
Writes a small helper so that a link you are sent later works whether you paste it into PowerShell or into Claude.
Walks you through signing in to your Claude account.
Proves your Claude plan can actually run Claude Code, by asking it a real question and reading the answer.
Prints a READY banner.

It is safe to run again. It picks up where it stopped and never repeats work it already did.

Step 3: sign in when it asks

A browser window opens for you to sign in to your Claude account. Do that, then come back to the black window.

If your account is on the free plan, the script stops here and says so. You need Pro or above, at about US$20 a month from Anthropic. Upgrade, run the same line again, and nothing you have done is lost.

Step 4: the trust question

The first time you type claude, it asks whether you trust the files in the folder you are in. That is your own home folder on your own computer, so yes is the right answer.

You will meet this question exactly twice in the whole journey: once here, and once later for your vault, if you buy. It is Claude Code asking whether it may use the settings that came inside that folder. It applies to that one folder and nothing else on your computer.

Step 5: read the banner

When it finishes, you get this:

Windows PowerShell

  ############################################
  #                                          #
  #   READY                                  #
  #   Your machine is ready for MYOS.        #
  #                                          #
  ############################################

  Everything MYOS needs is installed, and your Claude account works.
  That was the fiddly part, and it is behind you!

That was the fiddly part and it is behind you. Git, Claude Code and Obsidian are installed, your PATH is repaired, and your Claude plan is proven to work. Type claude in that same window and you are in a session.

If it says NOT READY

The banner names the item marked PROBLEM. Fix that one thing, then run the same line again. Re-running is safe.

A record of the whole run is written to %LOCALAPPDATA%\MYOS\install.log. If you are stuck, the rescue page has the read-only check to run and what every message means.

PLACEHOLDER: real screenshot of the READY banner in PowerShell, dated.
Target 1200 x 700, taken on a clean Windows 11 machine, not on a personal vault.

Where the free part ends and MYOS begins

What you have now is a very capable assistant standing in an empty folder. It knows nothing about your work, and when you close the window it forgets the lot, so tomorrow you start again from scratch. Claude Code is built to be general, and a general tool has no way of knowing what matters to you.

The usual next move is to build your own. A folder, some structure invented on the spot, a couple of instruction files, and it works well for about a fortnight. Then keeping the structure alive turns into the admin job you were trying to get rid of, and it goes quiet. That is how you end up with three abandoned note systems and a folder called notes-old.

MYOS is that same work, already done and used long enough to know which parts survive. Everything has a home from the first minute. The loop runs the same way on day one and in year three, and the assistant is configured to file, summarise and remember on your behalf. You talk to it and it writes.

The reading is free as well. Learn MYOS explains the whole framework before you spend anything: what the pieces are, why each one is shaped that way, and what it honestly does not do.

Before you buy: Windows 10 or 11, your own paid Claude plan (Pro or above, about US$20 a month to Anthropic), and about half an hour. If you cannot get it running, we refund you in full.

Your machine is ready before checkout is

That is the right way round. Leave your email and the one message you get is on the day checkout opens, at US$500. Nothing else goes to that address.

What your week looks like with it running

A call finishes and you tell your assistant what happened, in whatever order it comes out. It writes the note, files it under the right client, and puts what you agreed into your live work. You do not pick a folder, you do not name a file, and you do not open Obsidian unless you want to read something.

Later in the week you ask instead of remembering. What did we decide about the price. What did I promise them. Where did that job get to. The answers are in files it wrote, so it reads them back instead of guessing, and none of it depends on you having kept a tab open. Close a chat window and it forgets you, and tomorrow you start from the top again. That is the difference this makes, and it is the only one that compounds.

A job finishes and that is the end of it. The archiving pass folds the work back into the right files, so what you learned on that job is in the file you will look in next time, and your list of live work gets shorter without you tidying it. A folder needs you on Fridays. This does not.

Two people who are not selling you anything have described the far end of this. gist, writing on Hacker News about projects run "for my own benefit", put it this way: "I no longer find it necessary to know any details tips or tricks." Nolen Jonker at XDA went in expecting nothing and came out saying Claude Code was handling admin work better than the tool built for admin. Neither of them was running MYOS. Both were describing what happens once the tool stops needing to be managed, and that is the state this starts you in.

Watch it happen

Placeholder: demo video

60 to 90 seconds, recorded on a clean machine, showing the real paste line and ending on the assistant saying hello inside a new vault. Not recorded yet, so nothing is embedded here. No stock footage stands in for it.

What you own

Files on your own machine that you own outright. There are five parts.

1. The vault

A structured folder. Your work streams live in Areas/, your live and finished work in _Tasks/Active/ and _Tasks/Archive/, and your profile, your inbox and the structural rules sit in _System/. Everything has a home from the first minute, so nothing you tell it has nowhere to go.

2. The assistant configuration

The instruction file your assistant reads at the start of every session, a permission set that lets it write your notes without asking and stops it dead before it runs a command, and a status line so you can see what it is doing. It behaves the same way in session one and session two hundred, and you never re-explain yourself.

3. The working crew

Twelve specialist agents, seventeen skills and six commands, wired in already: research, scoping, planning, building, reviewing, linting, cleanup, and the archiving pass that folds finished work back into the right files. You say what you want and the routing happens underneath.

4. The onboarding and the nine-task tutorial

Your assistant interviews you, personalises the vault, and then walks you through nine tasks that teach the system by running it on work you actually have: Claude Code basics, how your assistant works for you, telling it about you, the task system, areas, Obsidian, connecting your tools, meeting debriefs and hooks, and finishing your first real task. You come out the other end with a system that already has your work in it.

5. The rules that keep it consistent

The structural model, the writing conventions and the templates your assistant follows. This is why a vault with three years of material in it is still readable, and why the files it writes in month twenty look like the ones it wrote in week one.

specialist agents

skills

commands

tutorial tasks

Fixes and improvements are free

You buy the kit as it stands today. It carries a version marker, it is yours at that version, and nothing lands in your vault on its own afterwards. If we find a defect in what you bought, or we improve the framework itself, that fix is free to you: honouring the sale includes fixing what is broken, so there is no paid upgrade to get a bug sorted and nothing to keep paying to stay current. What we will not do is push changes at you or write over anything you have customised. From there you extend it by talking to your assistant, which is the whole point of the thing, and Growing your system explains what that looks like.

What people who ran it say

Placeholder: beta cohort quotes

Three to five named quotes from the stranger beta cohort, each with written permission on file, each naming what the person actually does. No quote goes here until the cohort has run and permission is recorded. Nothing is written for them in the meantime.

PLACEHOLDER 1: quote, name, one line about their work.

PLACEHOLDER 2: quote, name, one line about their work.

PLACEHOLDER 3: quote, name, one line about their work.

Price

Six months of decisions, already made

Do not waste months building your own second brain. This one is already built, and six months of daily use proved which parts hold up. That is what the money buys: which structure survives contact with real work, what the assistant has to be told, what the loop is, what happens the moment a task finishes, and what runs on its own so the whole thing does not rot. You get all of it settled, in the files, working the first time you open it.

All of it arrives at once.

The vault, with a home for everything from the first minute.
The assistant configuration it reads at the start of every session.
Twelve agents, seventeen skills and six commands doing the research, planning, reviewing and archiving without you learning a single name.
The interview that makes the vault yours.
Nine tasks that teach you the system by running it on work you already have.

None of that is a bonus or a module, and none of it is waiting behind a second payment.

US$500, tax included, paid once

There is no subscription, no login and no account that has to stay alive for your files to keep working. What you end up holding is a folder of plain text files on your own machine, which is about the least fragile format there is: they open in anything, they will still be readable in ten years, and they carry on working whether or not we do.

The free alternatives, named

They exist and they are one search away, so here they are. There is claude-obsidian on GitHub, MIT licensed and actively maintained. There is second-brain-starter. There are free vault templates on GitHub, a free second-brain starter kit on Gumroad, and some good free tutorials on wiring Obsidian to Claude. If you want to assemble your own system, start there. They are genuinely useful and they cost nothing.

What else you could do

Build it yourself, which costs nothing but the months, and the free repos above are the right place to start. Or buy a template, which costs less than this and hands you the same empty shape to fill in on evenings you were going to spend on something else. Both land in the same place if the structure does not survive a real week, and you find that out late. The third option is to stop, and go back to typing into a chat window that tells you what to do and does none of it.

The assistant is the course

The teaching is not a second product you have to get through. Your assistant runs the nine tasks with you on work you already have, answers in the same window when something confuses you, and is still doing it in month twenty. So there is no video library to sit through, no cohort to keep up with and no membership renewing quietly in the background. Nothing in the price is waiting for you to make time for it. If the teaching is the part you would be paying for, read Learn MYOS first and judge it for nothing.

If you cannot get it running, we refund you in full

Write and tell us it will not run. Give us a fair chance to help first, which usually means answering a couple of emails and, if you are willing, sending us the install log off your own machine. If we still cannot get it running, you get the full US$500 back. We pay that even when the cause is at your end. A work laptop that blocks scripts, antivirus that quarantines the download, a network that will not pass the file: all of those count, because the promise is about the outcome and we do not argue about whose fault it was. All of this sits on top of the rights the law already gives you and takes nothing away from them.

Before you buy: Windows 10 or 11, your own paid Claude plan (Pro or above, about US$20 a month to Anthropic, not included), Git for Windows and Obsidian (both free, installed for you), and permission to install software on the machine.

If it will not run on your machine, you do not pay for it. Do the free guide first and you know that before you spend anything.

What this does on your computer

Read this part before you pay. It sets out what your assistant can do on your machine and what leaves it.

What your assistant may do without asking

It writes and edits files inside your vault without asking each time. That is the entire point of the system and it is confined to the vault. Running commands works differently. Every time it wants to run something in a terminal it shows you the command and waits for you to approve or deny it, and that stays switched on until you decide otherwise. When you are comfortable you can tell it to stop asking, and it will widen its own settings after telling you in plain terms what that means. Claude Code has a middle path built in as well: when a prompt appears, one of the options stops the asking for that one kind of action. There is also a mode that turns permission checks off completely. MYOS never uses it and never asks you to.

What leaves your machine

Your files sit on your disk. There is no MYOS account, no MYOS server, and nobody receives a copy of your vault because you installed this. What does leave is the conversation. When you talk to your assistant, the parts of your files it reads go to Anthropic as part of that conversation, exactly as they would in any Claude session. That is the honest boundary and it is worth knowing before you write something sensitive into a file and then ask about it.

The one thing we can be sent, only if you say yes

After an install, your assistant may offer once to send us a one-line note about how it went. That note holds five things: the order number, whether it worked, which step it reached, any error code, and your Windows version. Nothing from inside your vault, nothing else, and never without you agreeing to it. Say no and it moves on.

Where the vault goes

Into Documents\MYOS-Vault. If your Documents folder lives inside OneDrive it goes to %USERPROFILE%\MYOS-Vault instead and says so on screen, because a folder that keeps its own history every fifteen minutes and a sync client that keeps its own copy of everything will fight, and that fight surfaces as mangled files a week later. You can override it and put the vault wherever you want.

Your existing files are not touched

If there is already something sitting where the vault goes, and this machine has no record of putting it there, the install stops and asks you. It never writes over a folder on its own initiative, and nothing is deleted by any path through the installer.

Support, and who is answering

Support is by email. It covers getting the kit installed and running, how the system works, and anything that looks like a defect. It does not cover building custom workflows for you, business advice, or support for Claude Code, Obsidian or Windows themselves. Replies are drafted with the help of an AI assistant and they say so. Anything about refunds, the guarantee or pricing is answered by a person.

Questions

I know literally nothing about coding. Can I actually do this?

Yes, and the whole design assumes it. You copy one line into a window and press Enter. After that you are talking, not typing commands: your assistant runs everything and explains each prompt before it appears. You never write code and you never open a configuration file. If a prompt confuses you, ask your assistant what it means, in the same window, in your own words.

I got it all installed and then found out you need a paid plan. Is that right?

Yes: Pro or above, about US$20 a month, paid to Anthropic and not to us. It is the most common nasty surprise in this whole area, which is why the free setup line checks it before you spend anything here. It asks your account a real question, reads the answer, and stops with a plain explanation if the plan cannot run Claude Code. Upgrade, run the same line again, and nothing you have already done is lost.

It runs, it looks scary, and sometimes nothing appears to happen. Is that normal?

Mostly, yes. Installing three pieces of software takes a few minutes and some of that time looks like a still screen. The setup narrates itself as it goes, so there is always a "Step 3 of 8" line telling you where you are, and it writes a full record to %LOCALAPPDATA%\MYOS\install.log while it runs. If it genuinely stops, it prints a block marked PROBLEM saying what happened and what to do about it. Running the same line again is safe: it keeps everything that already worked and carries on.

Watching my machine run a script I do not understand is unnerving. What is it doing?

The setup script is a plain text file at a fixed address. You can open that address in your browser and read the whole thing before you run anything, and it is the same file for everyone. It installs Git, Claude Code and Obsidian through Windows' own app installer, or straight from the people who make them if this machine does not have that, then fixes your PATH, walks you through signing in and checks your plan. It deletes nothing, it touches no file of yours, and the record of what it did stays on your machine. Later on, when your assistant wants to run a command, it shows you the command and waits for you to say yes.

I already use Claude Code. Will this get in the way?

No. The setup detects what you already have and skips it, and the vault's own settings only apply inside the vault folder. Your existing setup and any other project you have configured are left alone.

Do I have to learn Obsidian? Is it free, even for work?

It is free, work use included, and it is not really somewhere you work. Obsidian is the window you read your vault through. Someone who tried this setup put it well: Obsidian is not for you, it is for Claude Code. The writing happens on the other side while you talk.

Do I need to know git?

No. Your vault can keep its own history, so a bad edit is a two-minute fix instead of a bad afternoon. Your assistant sets that up and uses it. You never type a git command unless you want to.

Does my data leave my computer?

MYOS adds no new party. There is no MYOS account and no MYOS server, and nothing from your vault comes to us because you installed it. Your files sit on your own disk. What goes to Anthropic is the conversation, exactly as it does in the Claude sessions you already have, so if you are already using Claude Code this changes nothing about where your words go. The one exception is deliberate and small: after an install your assistant offers once to send us a five-field note about how it went, and it only sends it if you say yes.

I already have an Obsidian vault. What happens to it?

Nothing, unless you ask. The install goes into its own folder, and if it finds files it did not put there it stops and asks you before doing anything. Moving your existing notes in afterwards is a conversation with your assistant, and there is no migration tool.

Someone said they burned through their credits in three hours. Will this eat my Claude plan?

The first session is a busy one: the interview, the setup and a nine-task tutorial. We have not measured what that costs on a Pro plan carefully enough to publish a number, so we are not going to invent one, and nobody selling you a template can tell you what your own usage will cost either. What we can say is that hitting a limit here is not a disaster. The work is written into files as it goes, so a new session picks up where the last one stopped.

Mac? Linux?

Neither, today. The installer is Windows-only and everything here is tested on Windows 10 and 11. There is a waitlist and no date.

Is this a course? Is there a community?

No and no. There are no videos, no cohort, no forum and no membership. What replaces them is an assistant that teaches you the system while you use it on your own work, plus a public explanation of the whole framework you can read before you buy.

There are cheaper Claude Code templates that do this. Why this one?

Some of them are good, and the free ones are named further up this page. The honest difference is where each one starts. A template hands you a folder of files and assumes the hard part already happened: that you can get Claude Code installed on Windows, sign in, work out for yourself that the free plan will not run it, and repair your own PATH when the command comes back unrecognised. That assumption is exactly where people fall over. Here that part is the free guide on this page and you run it before paying anything, so you either finish on a READY banner or you find out on our time. Two other differences. You end up with plain text files. They are not rows in somebody's database, so they open in anything and stay readable whatever happens to us. And the onboarding teaches instead of configuring: nine tasks on your own real work, so you come out of it able to change the system yourself.

What if I get stuck?

Four things, in order. The assistant on your own machine can see the actual error and diagnose it. There is a read-only check that prints a status table you can paste into an email. There is a rescue page explaining every message the installer can produce. And there is email support. If none of that gets it running, the refund is the rescue.

Can I share it with my team?

One purchase covers one person. Everything you create in your vault is yours. The kit's own files are not for reselling, republishing or passing on.

What happens when the kit improves? Do I have to buy it again?

No. You own what you bought, at the version you bought it, and defect fixes and framework improvements to it are free, because fixing something broken is part of the sale. Nothing lands in your vault overnight and nothing you customised gets overwritten by somebody else's improvement. Your upgrades are the changes you make and keep.

If you cannot get it running, we refund you in full

Finish the free setup guide first. It proves your machine and your Claude plan work, and it costs nothing. If you buy after that and still cannot get it running, tell us and take your money back.

Before you buy: Windows 10 or 11, your own paid Claude plan (Pro or above, about US$20 a month to Anthropic), and about half an hour.

Get in line for launch day

Leave your email and you hear on the day checkout opens, at US$500. That is the only message sent, and the address is used for nothing else.