What this is
Claude is one of the best tools for turning a description into working code, and Artifacts let you watch it render as you go. The catch: an Artifact lives on a claude.ai URL, runs under a strict content policy (no external scripts or API calls), and is really a capture of a session rather than a site you own. This stack uses Claude to build and Revdoku to publish, so the thing you made gets a stable, shareable address you control.
When to pick this stack
- The build is code-heavy or interactive and you want to design it live before publishing.
- You like Claude’s Artifacts loop but need the result on the open web, not behind a login.
- You want one canonical URL you can keep updating, hand to clients, or point a custom domain at.
How publishing works (Revdoku)
Copy the files out of your Artifact (or save Claude’s code) and publish them with Revdoku. You pick the access level:
- Public, an open link for anyone.
- Password-protected, the link plus a shared secret.
- Password + email, visitors sign in with a password and their email, with optional email confirmation to verify the address first.
Revdoku then shows you analytics on how the published app is used, and in password + email mode that breaks down per visitor, you see which email opened it and how they engaged. Re-publishing updates the same URL, and the free plan covers getting started.
What we’d swap
- Claude → ChatGPT: for the absolute simplest, no-frills “write me a page” flow.
- Claude → Cursor or Windsurf: when the project outgrows a single Artifact and wants a real local repo with files on disk.
Common gotchas
- Artifacts block external network requests, so anything that called an outside API inside Claude needs adjusting before (or after) you publish it as a standalone site.
- Claude’s code is clean but assumes the Artifact sandbox; double-check relative paths and asset links once it’s a real site at its own URL.