Set up the Kynver runtime on your computer

The Kynver runtime puts your agent on a computer you control. Instead of working inside a sandbox in someone else's cloud, the agent runs where your repositories and logins already live — you bring it online with one command, watch it work from the Command Center, and stop it whenever you want.

What you need

• A signed-in Kynver account. Approving the machine creates your workspace if you do not have one yet.
• Node.js 24.11 or newer.
• macOS, Linux, or Windows via WSL.
• A machine that can stay on — the agent only works while the machine is running.
• One supported coding-agent subscription, installed and signed in on the same machine. The runtime runs its work through the subscription you already pay for; you can also point it at any compatible model endpoint, including one running entirely on your own hardware.

The three commands

npm install -g @kynver-app/runtime
kynver bootstrap --api-base-url https://www.kynver.com
kynver start

1. Install

npm install -g @kynver-app/runtime installs the kynver command globally. If the install fails with a Node version error, upgrade to Node.js 24.11 or newer and try again.

2. Connect

kynver bootstrap links this machine to your account and workspace in one shot. There is no key to paste: it prints a short code and a link, you open the link in your browser, sign in to Kynver, and approve the machine. Bootstrap then configures the machine and gives it its own scoped credential.

3. Start

kynver start brings your agent online. It finishes any remaining setup, then runs the agent under supervision so a crash doesn't take it offline. Leave the terminal open — the agent works for as long as it's running.

What device approval is

When you run kynver bootstrap, the machine asks to join your account, and you approve it from your browser while signed in to Kynver. Approve the machine while signed in to this account. Any signed-in account holder can do this — it is not limited to admins. That approval mints a credential scoped to that one machine — it is not your account password and not your API key. You can revoke a machine from Account → Settings. Revoking cuts that machine off without touching your account or any other machine.

How to stop it

Press Ctrl+C in the terminal where kynver start is running. A deliberate stop is always respected: the agent shuts down and stays down until you start it again. (Automatic restarts only apply to crashes while the runtime is supposed to be running — never to a stop you asked for.)

If you set the runtime up as a background service on Linux with kynver start --install-systemd, stop it with:

systemctl --user stop kynver-cron-daemon.service

Because the agent runs on your hardware, stopping the machine stops the agent — there is no copy of it running anywhere else.

Using a signed-in coding-agent account for chat

After the runtime is online, website chat runs through the coding-agent command already installed on that computer. Keep that command on PATH and sign in there (agent login). The login stays on your computer — Kynver never copies it.

If that command is missing or you are signed out, website chat reports an error instead of answering from Kynver-hosted models.

Some setups still use a one-time opt-in so chat can use a different local login:

kynver start --chat-oauth

If you already ran kynver start without the flag and you need that opt-in, stop it first (Ctrl+C in that terminal) and start again with kynver start --chat-oauth. If you installed the Linux background service, stop it, then start again with the flag.

If the computer is off, chat falls back to a stored API key (Account → Settings) or to Kynver-hosted models.

Signing in on the computer is not the same as pasting a console API key. Use the key form in Settings when you want chat to keep working with the computer off.

Where to see it working

Open the Command Center from the Agentic OS in your account. It shows your workspace live: the machine you connected, the work your agent picks up, plans and tasks in flight, and anything waiting on your decision. If your agent opened pull requests, they'll be linked from the work it reports.

If something doesn't look right

• If kynver bootstrap reports that it can't resolve your workspace, your account likely doesn't have Agentic OS access yet — request early access first at kynver.com/agent-os/early-access (access is admin-approved) and wait for approval.
• If kynver start exits immediately with a configuration error, re-run kynver bootstrap to repair the machine's configuration, then start again.
• If the agent seems idle, check the Command Center first — an online machine with no queued work is healthy, just unemployed.
• If install or doctor checks keep failing, run kynver report-bug --json (draft only). Create a GitHub issue using the draft — reports land at kynver.com/help/bug-reports (see that article for template links).