Skip to content

Signals/ AI Agents/ Reel /

What if an AI-built website stayed editable after the agent finished?

Instead of regenerating a site for every change, the agent builds a content model that humans keep editing in a CMS.

EmDash Build is an MIT-licensed alpha reference app that turns a written brief into an editable Astro and EmDash website inside a Cloudflare Sandbox.

The architecture matters more than the headline:

  1. The agent designs the content model.
  2. Structured content is filled in through MCP.
  3. The agent writes the Astro pages and components.
  4. Editors can change the content later through the CMS.
  5. The site does not need to be regenerated for every update.

The repository also includes self-hosting instructions, tests, an operations guide, and a threat model. Public publishing is off by default, so the final release stays behind a deliberate, human-controlled step.

The useful commercial pattern is a narrow niche-site or lead-magnet builder:

  1. Collect structured intake data.
  2. Generate one repeatable content model.
  3. Build the microsite inside a sandbox.
  4. Let a human edit the structured content.
  5. Review the final design and links.
  6. Publish only after explicit approval.

The catch: this is still an early reference implementation, with 17 stars and one commit when it was found. Treat it as an architecture to study, not a mature production platform.

Sources & links 2

  1. GitHub: EmDash Build github.com ↗
  2. EmDash Build: hosted demo build.emdashcms.com ↗

More on AI Agent News, AI for Small Business

More signals

Reel

Your Next Customer Won't Fill Out Your Form. Their AI Agent Will.

Cube Forms launched Oct 12, 2026 (free): one form link where people see a normal web page and AI agents read the same form as Markdown or JSON and submit…

Helps Small business owners and creators with sign-up, booking, intake or lead forms

  • AI Agents
  • Cube Forms
  • MCP