AI NewsOpen sourceAnnouncement
Mike Hasa released GoLive, a Claude Code and Codex skill that takes an agent-built app live on the user's own accounts
Mike Hasa released GoLive on 23 September, an MIT-licensed Claude Code and Codex skill that walks an agent-built app through hosting, database, custom domain, email and test-mode payments on the user's own accounts, then can tear it all down. The repository passed 241 stars in one day.
Image: GitHub
Why it mattersA builder using Claude Code or Codex to make apps now has a repeatable last mile that inspects the code, prints the destination accounts and cost before writing anything, and records everything it created so the same command can remove it.
A coding agent can build a working app in an afternoon, and reaching a real user still costs half a day of signups for hosting, a database, a domain and an email provider. Mike Hasa released GoLive on 23 September, an open-source agent skill that carries a Claude Code or Codex session through those signups: hosting, database, custom domain, transactional email and test-mode payments, on the user's own accounts. The repository was created two days ago and passed 241 stars on 24 September under the MIT licence.
The skill installs as one command inside the agent. A chat prompt like "use golive to take this app live" then makes the agent inspect the code, ask about the providers it does not yet know, and print a plan naming the exact accounts, resources and cost before it does anything. The user approves each step, so a live payment webhook or a real DNS record is never written without a "yes" first.
Which providers the alpha covers
Hasa lists the paths that have been exercised on real accounts with disposable resources: frontend hosting on Vercel and Netlify; a database on Supabase and Neon; a custom domain on Porkbun and GoDaddy; transactional email on Resend; and test-mode payments on Stripe. The full launch checklist in the README is longer than that, and Hasa marks each row on the roadmap with a strikethrough on what is proven and a plain note on what is still planned.
Two limits are stated on the page. Cross-pairings that were not exercised carry mock coverage rather than a live run. Providers outside the tested list fall to a general guided flow that follows their official CLI or dashboard, and Hasa calls that best-effort guidance rather than a validated adapter.
The teardown command
The other half of the pitch is the reverse direction. golive teardown plans the removal of everything the skill created, again for approval, and Hasa's live-test log shows the disposable Netlify project it created being removed cleanly along with the Porkbun and GoDaddy records, the Resend sending keys and the Stripe test-mode endpoint.
For a team using an agent to try out experiments, that is the piece a written checklist cannot easily provide: a record of what was created and a way to remove it that another person can repeat. There is no GoLive account, no hosted backend and no product telemetry, so a project that stops using the skill leaves nothing behind except what the user's providers already know about.
Source
- Repository: mikehasa/golive-skill
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.