Questions

Fair questions.

Your AI can tell you how. Then you’re the one clicking through twelve steps. Lyndran does them for you, on your Mac, and asks before anything changes.

How is this different from Claude on its own?+

Claude tells you how to do the thing: open this, click that, sort by size. Lyndran does those steps on your own Mac, with your files, your mail and the browser you are signed into, and stops at a card before anything changes. You keep asking in Claude; Lyndran is the part that can reach your Mac. And an answer that works can be kept as a tool you run again, without asking anyone.

What about the desktop assistants that work on my files?+

They are real, and if what you want is a folder rearranged, several will do it. Two things decide whether this is the one you want. The first is reach: Lyndran does the task inside the things you already use — your mail, your Finder, the browser you are signed into — so “find the invoice they sent in March and file it” is one task, not a file operation with the rest still left to you. The second is the press: confirming a deletion is a dialog somebody added, and this is the other thing — the ability to commit is not in the app, so there is nothing to switch off or forget. And one thing we are plainly not: this is not for writing software. No terminal, no code, no repository. It is for the invoice, the supplier and the claim.

What does it cost?+

7 days free, starting the first time your AI uses Lyndran on your Mac, then $5 a month. Claude or ChatGPT itself is paid for by your own plan with them. Nobody is moved onto a paid plan quietly.

Can it send or delete things by itself?+

No — but this page used to say something stronger than we can say now, so here is the exact shape of it. An email is written for you and put in front of you whole: every recipient, the subject, the entire body. You press send on that card. Lyndran presses the mail app’s own Send button afterwards, so you are not sent off to find it — your press is the approval, and without it nothing goes. There is no bcc field anywhere in the code, five recipients is the most a card can carry, and “always allow” is never offered for sending, because a sent email cannot be recalled. One more thing sends in your name, and only when you ask for it: an out-of-office. The card shows the Gmail account, the days and the exact reply, and after you press it, Gmail sends that reply to everyone who emails you until the last day, or until you ask Lyndran to turn it off. Replies it has sent cannot be recalled either. Nothing is deleted for good either: what it clears goes to the Trash, and every change it makes has a one-press undo. Say no and nothing happens at all.

Can I use it from ChatGPT, or from Claude on the web?+

Yes. In Lyndran’s Settings, press Connect ChatGPT or Connect Claude: it copies Lyndran’s address and opens the page where you add it. You sign in with Google once, press Allow on your Mac, and from then on you ask in ChatGPT or Claude and Lyndran does it on your Mac, with the same cards before anything changes. The steps, and what to do if it does not connect, are on the connect page. The same sign-in and Allow also open your Mac in any browser, at app.lyndran.com.

Can I give ChatGPT just one project?+

Yes. On a project’s page on your Mac, press Connect this project to ChatGPT (or Claude). It copies that project’s own address; add it in ChatGPT or Claude, sign in with Google, and press Allow on your Mac. That connection works on that project only: its files, databases, notes and memory. Your mail, your screen and your other projects stay out of its reach, and every change still comes to you as a card.

What about sites that send me a code to sign in?+

Lyndran asks you for it. When a site wants the code it just texted or emailed you, a card comes up on your Mac with a box in it: you type the code there and press the card, and Lyndran puts it into the page. Your AI never sees the code, and it is not kept anywhere. A saved password works the same way — it goes from Lyndran’s vault on your Mac straight into the site it was saved for, never through your AI.

Can it change a password for me?+

Yes. Ask for a new password on a site you have a login saved for, and a card comes up on your Mac saying what will happen: Lyndran makes a strong password, types it into the site’s new-password boxes, and saves it in the vault under that login’s name. The one you had before stays saved as “… (before)” until you remove it in Settings › Saved logins, so a site that quietly refuses the change has not locked you out. Nobody is shown the new password, not your AI and not the card; your Mac’s keychain holds it.

Is there a Windows version?+

Not now. Lyndran is for the Mac.

What does my AI see, and is my data uploaded?+

Your files stay on your own Mac — the one exception is a photo search for something your Mac can’t recognise by itself, like damage or a paint colour, where small copies of the photos go to the AI only after you press Yes on a card — and everything Lyndran keeps — tools, history, preferences — lives there too. Claude or ChatGPT sees only the readings you switch on for it, kind by kind, and Lyndran shows each one with how many words went; mail and your screen also need a short allowance pressed on the Mac. What your AI reads goes to Anthropic or OpenAI under your own account with them; from Claude on the web or ChatGPT it passes through Mainspring’s relay on the way, encrypted end to end, and nothing is kept there. Your passwords and saved logins never go to your AI, not even their names. And “forget everything about me” is one press away.

Still unsure?

Give it one task. Something small from today’s work. Watch it stop and ask.

What it will never do on its own ›