TermLink FAQ: agents, machines, and what leaves your computer

Yes. Install the host with `npm install -g @termlink/cli` on any machine — a workstation, a cloud box, a home server — then run `termlink start` there. It registers under your Google account by dialing out to the TermLink relay. You then open TermLink in a browser and start a Claude session on that machine. The machine never accepts inbound connections, so you do not open an SSH port or set up a VPN.

Yes. The session lives on your own machine, not in the browser tab, so closing the tab does not interrupt the agent. Reconnect and TermLink repaints the screen exactly as Claude left it. The one requirement is that the `termlink start` process keeps running — TermLink is not a background daemon like tmux.

No. TermLink sends the stopped turn again by itself, a minute after your plan’s usage limit resets, and shows a countdown until then — a limit you hit at night finishes while you are asleep. The reset time comes from Claude Code itself rather than a guess, and typing into the session cancels the wait and sends straight away. The limits of it: Claude only for now, the turn is resent once rather than retried forever, and the wait does not survive restarting the host, because no conversation text is stored to resend.

Open TermLink from wherever you are. Each machine you registered is one line showing that session’s most recent message from the agent, so you can tell at a glance whether it finished, is still working, or stopped on a question. Tap the one you care about and you are in its terminal, at the point it actually reached.

Yes. The session belongs to the machine it runs on, not to the terminal or the app that opened it. Start it on a laptop, close the lid, and open that same session on a phone. You get the screen as the agent left it rather than a fresh one. You can also have it open in more than one place at once, because the plan cap counts sessions rather than viewers.

Each session carries its own approval setting, and it starts at `Ask for everything`. Move it to `Allow file edits` and file changes go through while a command still waits for you; move it to `Allow commands and edits` and both do. A plain terminal session has a simpler `Auto-Yes` switch instead, which confirms a highlighted Yes or Allow on screen. Either way you can change it mid-run from a phone.

Treat it as a convenience, not a safety feature. It does not read the command and decide whether it is a good idea. Two things it will not do: it never answers a question the agent asks you or a plan it wants approved, and it takes the plain allow rather than the wider “stop asking for this session” option sitting next to it. Leave it on for a run whose blast radius you already accept, and off for anything you would want to look at before it happens.

Yes. TermLink runs in a mobile browser and adds a control-key row (Esc, Tab, Ctrl+C, arrow keys) so a terminal UI is usable on touch. You can read what the agent is doing, answer its prompts, and drop a screenshot onto the terminal — TermLink uploads it to the host and types the saved path in for your next prompt.

Remote desktop tools stream an entire screen: heavy, clumsy on a phone, and they need the far machine unlocked. SSH means an exposed port and keys to manage. TermLink hosts dial out and mirror only the terminal your agent lives in. Unlike tmux it is not a background daemon, but it adds what tmux does not have: a browser and phone client, a correct TUI repaint across reconnects, several machines in one list, and approvals answered while you are away.

Yes. Four agents are on by default: Claude Code, Codex CLI, Cursor Agent, and GitHub Copilot CLI. Claude and Codex run through their own SDKs; Cursor and Copilot connect through the agent protocol server each of them ships. You install and sign in to those tools the way you normally would (`claude`, `codex login`, `cursor-agent login`, `copilot login`), and the host only checks that you did. It never reads their credentials. Claude Code is the one we support most closely and test the most. You can also open a plain terminal session, so whatever else you run in a terminal runs here too. There is more on this in the session manager for AI coding agents.

Yes. TermLink never copies your repository anywhere — the agent runs on your own machine, in the folder you started the host in, and the files stay there. What travels is the terminal screen: those bytes go over WSS through the TermLink relay so your browser can draw them, and the relay forwards them from memory to the connected clients without writing them to disk. There is no scrollback, recording, or transcript on our side, and your agent keeps using the account you already signed it in with — TermLink never reads its credentials. It is not end-to-end encrypted, though: the security page spells out exactly what the relay can see.

No. TermLink does not talk to any model itself. Its host drives each agent through that agent’s own interface, using the account you already signed in with, so you keep your own subscription and your own keys exactly as you do now. What TermLink adds is a terminal or agent session on your machine, reachable from a browser or a phone, with every machine you registered in one list.

Yes. Your TermLink account (the Google sign-in that lists your machines) and each machine’s Claude Code account are separate — TermLink never reads or shares Claude Code credentials, it just checks that a machine is signed in. Sign a personal laptop into your own Claude Code account and a work machine into your company’s seat, register both under one TermLink account, and both sessions show up in the same list.

Claude Code Remote Control lets you reach a Claude Code session on your machine from another device. TermLink is built around a different job: keeping several machines in one list, in a full terminal, from a browser or phone. You register each machine once with `termlink start`, and every session you open is the real terminal — so the same window works for whatever else you run in it, not only for the agent. If you work on one machine and want the shortest path to it, Remote Control is a fine answer. If your work is spread across a laptop, a workstation, and a box that only runs agents, that is the case TermLink was built for. There is a fuller comparison of TermLink vs Claude Code Remote Control on its own page.

Yes. The host runs on macOS, Linux, and Windows, and TermLink runs in any modern browser, including mobile. On Windows the host uses cmd.exe by default because PowerShell’s PSReadLine swallows the Esc key that terminal UIs like Claude Code rely on.

Yes, once you turn it on. TermLink raises a desktop notification when a session finishes a run or starts waiting for your input while you are looking at something else. It is off by default, and switching it on in Settings asks the browser for permission first. These are browser notifications rather than push, so they only fire while a TermLink tab is open somewhere; closing it, or locking the phone for more than a few seconds, stops them. There is no email notification. Each machine card also shows that session’s latest message, so the list tells you who is working and who is waiting even when you have notifications off.

Yes, there is a Free plan: Google sign-in, connecting to your own machines, real-time terminal access, and Auto-Yes. The plans differ in exactly one thing — how many sessions you can keep active at the same time. Free gives you four, Pro twelve, Max forty, and in every case each active session streams live with no time limit and no reconnect wait. The cap counts sessions, not viewers: you can watch them from as many browsers and phones as you like. Pro is $9.99 per month and Max is $19.99, both with priority support.