At a glance
- No open ports. Your computer calls out; nobody can call in.
- A computer joins your account only after someone at that computer approves it.
- Only AI apps you signed in to with your Local Commander account can send work.
- You choose what the AI can touch. Access can be widened only at the computer itself.
- The real keys are your accounts. Protect them like your email or your bank.
No open door on your computer
Local Commander doesn’t open a port, doesn’t change your router and doesn’t need a VPN. The small program on your computer connects out to Local Commander over an encrypted connection and waits for work on that line, the way a messenger or a remote-help app does.
Nothing on your computer listens for connections from the internet. There is no address someone could find and knock on. If a stranger scans your network, there is nothing of ours to see.
Only computers you approve join your account
To add a computer, you start Local Commander on that computer and approve it in your account. A click in a browser alone isn’t enough: each computer has its own secret key that never leaves it, and every connection is signed with that key. A copied link or a stolen screenshot can’t add a computer to anyone’s account.
Helping a relative works the same way: their computer joins only after the approval is done on it, with their permission.
Only your AI apps can send work
Work reaches your computers only from AI apps — ChatGPT, Claude or another — where you signed in with your Local Commander account and allowed access. Someone else’s ChatGPT can’t reach your computers, even if they know Local Commander exists.
Your account page lists every connected AI app and every computer. Remove any of them with one click.
You set the limits
When you set up a computer, you choose: full access to your user account, or only the folders you pick. You can also block specific commands.
These limits live on the computer. Access can be widened only by someone at that computer — not from a chat, not from the website and not by the AI itself.
See everything, stop anytime
Every task your AI ran is in your account’s activity, in plain words: which computer, what it did, what came back.
To stop a task, ask your AI to cancel it, or close Local Commander on that computer. Removing a computer from your account blocks any new work right away; a task already running finishes or stops when you close the program there.
Where the real risk is
We’d rather say it plainly than hide it in small print.
- Your accounts are the keys. Anyone who gets into your ChatGPT, Claude or Local Commander account can do what you can do. That’s the same risk as with your email or online banking, and the same protection works: a strong password and two-step sign-in on your AI accounts.
- The AI can make mistakes. It can misread a request or follow bad advice it found online. On a computer that isn’t yours, such as a parent’s laptop, give access to the needed folders only.
- It isn’t a sandbox. Work runs with the permissions of the user account on that computer. For stronger separation, run Local Commander under a separate user account.
- Results pass through our service on their way to your AI. The privacy policy says what is kept and for how long.
In short: nobody gets in by accident or from outside. The access exists only in your hands — and in the accounts you protect.
Report a concern
Email Peter at peter.mamrukov@localcmd.app with a short description. Don’t send passwords, sign-in links or keys. For sensitive vulnerability details, ask for a secure way to share them first.