OverviewHistoryStatsSecurity
npx skills add ...
Documentation
SKILL.md
npx skills add zeabur/agent-skills --skill zeabur-server-list
Use when listing dedicated servers. Use when checking server status, IP, or provider info. Use when user says "show my servers", "SSH into server", or "check server status". Do NOT use for browsing purchasable servers (use zeabur-server-catalog instead).
npx skills add zeabur/agent-skills --skill zeabur-server-list
Always use
npx zeabur@latestto invoke Zeabur CLI. Never usezeaburdirectly or any other installation method. Ifnpxis not available, install Node.js first.
ZeaburOS means the machine can host Zeabur projects; Ubuntu means it cannot (not until ZeaburOS is installed on it). This is the authoritative answer — do not SSH into a machine to work out which kind it is.
For machine-readable output, --json carries the same value as os on every server:
Requires CLI 0.21.0 or newer. If the
OScolumn and theosfield are both absent, the CLI is older; either letnpx zeabur@latestfetch the current version, or fall back to queryinghasK3sthrough the API (see thezeabur-server-sshskill, which documents that field's three-state trap).
Shows detailed info: OS, CPU/memory/disk usage, provider, location, managed status, creation time.
Resource columns read used/total — CPU in millicores (148/4000 m is 0.15 of 4 cores), memory and disk in MB. A column showing — means the value was not measured, not that the machine has none.
-y skips confirmation prompt — required for non-interactive use.
A server can host Zeabur projects only once it runs ZeaburOS. A project's Region.ID will be server-<server-id> when it is deployed on such a server.
Renting provisions Ubuntu — base OS and SSH, no Zeabur services — so a newly rented server cannot host projects until ZeaburOS is installed on it. The zeabur-server-rent skill covers that step. Which kind a given server is, is answered by the OS column above.
Installing directly is only for a machine that is still blank. On one the user has already set up, converting means reinstalling it in the dashboard, which erases the disk — see the
zeabur-server-rentskill before suggesting it.
On a ZeaburOS server:
zeabur-deploy skill with the project bound to that server — do NOT SSH in and manually set up web servers or copy files.On a plain Ubuntu VPS: SSH is not a debugging side-channel, it is the whole point — the machine is the user's to set up as they like (docker compose, 1Panel, 寶塔, …). There are no Zeabur services to manage, and kubectl does not exist.
The
kubectlexamples below assume a ZeaburOS server. On a plain Ubuntu VPS they fail withcommand not found— see thezeabur-server-sshskill for how to tell the two apart.
For managed servers, the password is fetched automatically. If sshpass is installed, login is fully automatic; otherwise the password is printed for manual entry.
In non-interactive environments (e.g. Claude Code, CI/CD), you cannot use an interactive SSH session. Instead, pipe commands via stdin with -i=false:
For multiple commands, separate them with ; or &&:
Important notes:
grep with no matches), the CLI will print ERROR exit status 1. This does not necessarily mean the SSH connection failed.<<'EOF') does not work reliably with this command. Always use echo '...' | instead.npx zeabur@latest server list -i=false --json | jq -r '.[] | "\(.Name)\t\(.os)"'# By server ID
npx zeabur@latest server get <server-id> -i=falsenpx zeabur@latest server reboot <server-id> -ynpx zeabur@latest server ssh <server-id>echo 'kubectl get pods -A' | npx zeabur@latest server ssh <server-id> -i=falseecho 'kubectl get pods -A; kubectl get svc -A' | npx zeabur@latest server ssh <server-id> -i=false