Two different publish models
Drag a file in, get a link out — no project, no build, no repository. Against that: connect a repository, configure a build, get deployments, previews, redirects, functions and a team. The first is a verb, the second is a platform.
Where the drag-and-drop tool wins
Time to link, and nothing else comes close. For a one-page site, a client mockup, or a file somebody emailed you, the platform's setup cost is pure overhead. It also stays comprehensible to people who do not deploy things for a living, which is a genuine product quality.
Where the platform wins
Anything with a lifecycle. Build pipelines, per-branch previews, redirect rules, serverless functions, form handling, staged rollouts, an audit trail of what shipped when. If two people will change the site over the next year, the platform's overhead becomes its value.
Pricing shape
No numbers, on purpose. The shapes: a drag-and-drop tool prices per site or per small plan, and a platform prices per seat plus usage. Which is cheaper depends entirely on whether you are counting artifacts or people — and an agent is neither, which is where both models start to creak.
Side by side
| Drag-and-drop host | Deploy platform | |
|---|---|---|
| Unit | A file or folder | A project |
| Setup | Seconds | Minutes to hours |
| Build step | None | Yes |
| Previews | No | Per branch |
| Runtime code | No | Yes |
| API for a program | Limited | Yes, and project-shaped |
The third option, for agent output
Both assume a human at the keyboard: one at a drop zone, one at a dashboard. When the author is an agent, what you want is the drag-and-drop shape reachable as a tool call, with privacy on by default because the file was assembled out of a context window rather than written by hand.
# the same act, without a browser
npx -y opentree-mcp
# then, from the agent: publish_html { content: "<html>…</html>" }Can a drag-and-drop host be automated?
Some expose an API, but the abstraction stays "an upload a person made". The distinction that matters for a loop is whether re-publishing the same artifact keeps the same URL.
Is a deploy platform overkill for one HTML file?
For one file that is finished, yes. For one file that is the beginning of an application, no — and the second case is more common than it looks.
What about privacy on either?
Both express it as a plan feature. Per-link privacy — password, email gate, one-time view, expiry — is a different product shape, and it is the one this deployment implements.
Which handles a zip?
Both, and so does this: a zip becomes a multi-file site rather than a single document.