Frequently asked questions
The difference, the work, your computer, your data, pricing, and how to start.
The difference
ChatGPT and Claude can work on my computer too. What's different?
Two things. yarnnn works on whichever engine you pick, so you aren't tied to one company's model. And it keeps one signed record of who changed what, across every engine and every AI you connect. An AI company can show you what its own agent did. It can't neutrally show you everything.
How is this different from ChatGPT or Claude's memory?
Their memory lives inside their own app. You can't open it, see its versions, or take it with you. yarnnn keeps your work as real files with a history, in a workspace outside any one AI. Every AI you connect works in the same files, and all of it is yours to export.
How do I see who changed something?
Open the file's history. It lists every change with who made it and when: you, your agent on the engine it used, or an AI you connected. For a workspace file you can open any earlier version and put it back.
Does it work across my team, not just me?
yarnnn starts as yours alone — your workspace, your computer, your sign-ins. You can share any file or folder with someone for viewing, and every change carries its author's name, human or AI. For an organization that wants one workspace for its people, a Team workspace — an admin, a seat for each person, one bill — is set up with us.
The work
What can it actually work on?
Files in your workspace. On your computer, with the desktop app: Word, Excel and PowerPoint files and plain text on Mac and Windows, plus Keynote, Pages and Numbers on a Mac, where it can also work in your programs through their menus and windows. PDFs are read, not changed. On the web, with the Chrome extension: the sites you use, in your own browser. And the services you connect: Slack, Notion, GitHub and WordPress.
Do I have to move my files into yarnnn?
No. A file on your computer is changed where it is, in the format it is in. yarnnn has no editor and no format of its own. A document stays a .docx, a deck stays a .pptx, a sheet stays an .xlsx. You can also keep files in your workspace, where every version is kept.
Which AI models does it work with?
Inside yarnnn you pick the engine for each chat: Claude, GPT, Gemini, Grok or DeepSeek. From outside, any AI that speaks MCP can connect to your workspace, including ChatGPT and Claude. It is neutral on purpose, which is why one record can sit across all of them.
Can it do something on a schedule?
Yes. Say when it should happen again, in the same conversation. It keeps a file current on the schedule you set, from the sources you name: a web page, a connection, or other files in your workspace. A run with nothing new to read costs nothing. Every run is listed in Work, where you can run it now or pause it.
Can my agent act on websites for me?
Yes — in your own Chrome, through the yarnnn extension. Your agent works in tabs of its own with the sign-ins you already have, does what you ask there as you, and keeps going if you close yarnnn while Chrome is open. Work you want repeated waits for you to start it. Banking, payments, trading and password managers are never allowed.
Your computer
Does it take over my computer?
No. You keep the pointer, the keys and the screen. What your agent opens stays behind what you are doing. It works only in the folders you allow, and stays out of the programs and sites you name. Hidden files, password managers, the terminal and system settings are always off.
Does it ask me before it acts?
Once — when you let it use this computer or your browser, on your own machine. After that it does what you ask, as you, sending and posting included, and every step is listed in History. Banking, payments, trading and password managers are never allowed, and you can narrow it or switch it off at any time.
What happens when it can't do something?
It says so, and tells you what would make it possible: a folder to add, a switch to turn on in macOS, a sign-in only you can give. It doesn't hand you a substitute and call it done.
Your data
Is my data mine?
Yes. Every file carries its author's name, every version is kept, and the whole thing exports as a plain git repo with its history intact. We never train on it — and nothing here deletes itself on a timer.
Where does my work go when I ask an AI to do something?
To the model provider running that task — Anthropic, OpenAI, Google, xAI or DeepSeek, depending on which engine is picked. They process it to answer, and under their API terms none of them train on it by default. File text also goes to OpenAI to build the search index that makes your workspace findable. A Word, Excel or PowerPoint file on your computer goes to yarnnn only while it is being read or changed, and is not kept. The full list of everyone who can receive your data is on the data page.
What can a connected AI actually do in my workspace?
The same things you can: read, write, move, delete, and share files. It connects through OAuth you approve and can revoke, and every change it makes is signed with its name — so you can always see what it did and walk it back. Connect assistants you trust.
Pricing & lifecycle
What does it cost?
Start free: every workspace begins with a $3 balance, enough to feel the loop before you spend a cent. The Personal plan is $20 a month and includes $15 of usage every month; if a heavy month needs more, top up any amount from $5 (a top-up never expires). AI connections are always free. A Team workspace for an organization is arranged with us.
Can I cap what it spends?
Yes. Spend is bounded by the balance itself — nothing can draw more than your plan includes plus what you have topped up, and the hard stop at zero pauses the work without losing anything. There is never an overage charge.
What if my balance runs out?
Nothing is deleted. When the balance is spent, the work pauses. Top up, or wait for the plan to renew, and it resumes. Your workspace and every file stay yours either way.
Getting started
How do I start?
Start free in your browser. Then download the desktop app, sign in, and let your agent use this computer. Ask it to change something in a file you already have, and watch it happen.
Is there a desktop app?
Yes, for Mac and Windows, free on every plan. It is how your agent works on your computer's files, and on a Mac in your programs. It's in beta, so the first time you open it your computer asks you to confirm; the download page shows how. You can also use yarnnn in any browser, for work in your workspace and on websites.
What's the best first move?
Pick a real file with something out of date in it: a deck, a sheet, a document. Ask your agent to fix that one thing. Then open the file in its own program and see the change, and open its history in yarnnn and see who made it.
Is yarnnn in beta?
Yes. It works and we run our own work on it every day, but it will keep changing, and some things will break along the way. None of that touches your files: every version is kept, every change is signed, and the whole workspace exports as a plain git repo whenever you like. If something breaks or confuses you, use Send feedback in your account menu.
Still have questions?
Start free. Ask for one change in a file you already have, and watch it happen.