Enterprise & fleets

On-call incident response from your phone

Q: How can I restart a production service from my phone?

If the AI Commander agent is running on the server, open your AI client on your phone, quote the machine's session code or alias, and ask it to restart the service. It runs the command over the relay and streams back confirmation — no laptop or VPN needed.

Q: Can I triage an incident without a laptop or VPN?

Yes. Because the agent connects outbound and your AI client speaks to it through the relay, you can read logs, check resource usage, and restart services from a phone chat — no VPN or SSH key on the device.

The pager goes off on the train, or at 2am. You don't have your laptop, and even if you did, the VPN handshake and SSH dance cost minutes you don't have. With AI Commander you open Claude on your phone, name the machine, and start triaging — read the logs, restart the service, confirm it recovered.

The job

The thing remote-access tools are really selling on-call engineers is "investigate a system error from anywhere." AI Commander makes that a chat: the box already runs an outbound-only agent, so there's nothing to connect to — you just describe the triage and the AI client does it and reads the results back.

Find the blast radius: systemctl --failed, recent errors in journalctl.
Check the obvious culprits: disk full, OOM kills, a dependency that's down.
Restart or roll back the affected service and watch it come healthy.
Confirm recovery — hit the health endpoint, tail the log — before you go back to sleep.

What it looks like

Claude — on your phone

You: checkout is down on api-prod — find out why and fix it if it's obvious Claude: > systemctl --failed api.service loaded failed failed > df -h / ; journalctl -u api -n 20 --no-pager / 100% full ENOSPC: no space left on device > journalctl --vacuum-size=200M ; systemctl restart api ✓ api active (running) · / now 71% Claude: Disk was 100% full from logs. Vacuumed journald, freed space, restarted the API — checkout is back. Worth adding log rotation.

Why AI Commander when you're paged

	AI Commander	VPN + SSH	TeamViewer / AnyDesk
Works from a phone	✓ chat	awkward	tiny screen
Time from pager to command	seconds	VPN + keys	launch + accept
Headless server	✓	✓	needs a screen
AI helps diagnose	✓ reads output	✗	✗

Guard the code. The session code is the credential and the agent runs as root by default. Sign in with a free account so the machine auto-links under an alias for ongoing access. The owner can still block your account (aicommander-agent block-admin) or reset the code (aicommander-agent change-code). Read the security model before relying on this for production incidents.

Set it up

On each production box (Linux), install the agent ahead of time so it's ready when you're paged:

Follow the signed Linux installer steps and verify the installer before sudo.

FAQ

How can I restart a production service from my phone?

If the agent is running on the server, open your AI client on your phone, quote the session code or alias, and ask it to restart the service. It runs the command over the relay and streams back confirmation — no laptop or VPN needed.

Can I triage an incident without a laptop or VPN?

Yes. The agent connects outbound and your AI client talks to it through the relay, so you can read logs, check resource usage, and restart services from a phone chat — no VPN or SSH key on the device.

Is this safe for production access?

The session code is the credential and the agent runs as root by default, so guard the code. Sign in with a free account to auto-link the machine under an alias for ongoing access. The owner can block you with aicommander-agent block-admin or reset the code with aicommander-agent change-code. There are no inbound ports to attack. Review the security model before using it on production.

Be ready before the pager fires

Install the agent on your production boxes now so triage is one chat away when it counts.

Install the agent Read the security model