Zomg is built for agent-assisted setup. Hand your coding agent the hosted quickstart prompt and it installs zomg, collects the GCP/domain/TLS values it needs, deploys your environment, and proves it works.

​
Start with a prompt

Full prompt: https://zomg.ai/go.md For Claude Code: 
curl https://zomg.ai/go.md | claude
For Codex from the CLI: 
codex "$(curl -fsSL https://zomg.ai/go.md)"
For Codex in the desktop app, start a new session and paste: 
https://zomg.ai/go.md
The prompt tells the agent to install the CLI from zomg.ai, load zomg setup agent, use the deploying-zomg and zomg-cli skills when available, start the setup UI on a local computer, and keep going until your environment is deployed.

​
Install the agent plugin

You can also install the Zomg plugin so /quickstart and the Zomg skills are ready in future sessions. Both agents install the marketplace straight from the public GitHub repo—no clone required. Install for Claude Code: 
claude plugin marketplace add loopwork/forkr
claude plugin install zomg-plugin@zomg
Install for Codex: 
codex plugin marketplace add loopwork/forkr
codex plugin add zomg-plugin@zomg
Then start a new agent session and run: 
/quickstart

​
What the agent will ask for

You need:
  • A GCP account with billing enabled.
  • A Zomg domain, for example zomg.ai.
  • DNS access for <zomg_domain>, api.<zomg_domain>, ssh.<zomg_domain>, and *.<zomg_domain>.
  • TLS certificates for that domain, or DNS/API access so the agent can help issue wildcard certificates.
The agent will use the setup UI and CLI commands such as: 
curl -fsSL https://zomg.ai | sh
zomg setup ui
zomg setup check
zomg setup deploy
The agent verifies that the selected GCP project is active before deployment, updates the ADC quota project when using GCP KMS/SOPS, and asks before mutating cloud resources or DNS records unless you are running a disposable verification. After deployment, it exports the runtime profile, creates a box, runs a command in it, copies in an HTML file, starts a small web server, publishes it, and gives you the URL.

​
Try Zomg after setup

When the agent finishes, your local CLI points at the deployed profile. These commands create a box, run Linux commands, copy in a page, and publish it: 
zomg health
zomg project current
zomg create hello-zomg
zomg exec hello-zomg -- cat /etc/os-release

cat > /tmp/zomg-hello.html <<'HTML'
<!doctype html>
<html>
  <head><meta charset="utf-8"><title>Hello from Zomg</title></head>
  <body><h1>Hello from Zomg</h1><p>This page is served from a box.</p></body>
</html>
HTML

zomg cp /tmp/zomg-hello.html hello-zomg:/root/index.html
zomg exec hello-zomg -- sh -lc 'mkdir -p /srv/hello && cp /root/index.html /srv/hello/index.html'
zomg service start hello-zomg hello-web --port 8080 --publish hello -- sh -lc 'cd /srv/hello && python3 -m http.server 8080 --bind 0.0.0.0'
zomg service status hello-zomg hello-web
The published page is usually at https://hello.<zomg_domain>. Useful follow-up commands: 
zomg list
zomg exec hello-zomg -- pwd
zomg service logs hello-zomg hello-web --lines 50
zomg service list hello-zomg
zomg delete hello-zomg --force