Manus, the agent startup that turns chat prompts into deployed web apps, now lets that agent publish finished work to a live URL with nobody clicking confirm. The feature, called Auto-Publish, is available today across web, iOS, and Android, according to a post on Manus’s company blog. It ships while the question of who actually controls Manus, after a blocked Meta acquisition and reported talks with Tencent, remains open.
The mechanic is simple. Inside the WebDev tool’s publish menu, one switch instructs Manus to deploy automatically the instant a build finishes without errors. Off by default, the setting is meant for sessions where a builder already knows the direction and no longer wants to click Publish after each small edit.
Manus frames this as removing “ceremony” from finishing work, not exploration. That framing is doing real work. The company’s own FAQ confirms that only completed, successful builds reach the public URL: a failed or in-progress build leaves the last stable version live. Manus has not disclosed how it defines a “successful” build beyond a clean compile, which is a lower bar than a correct one.
Combined with message queuing, an existing Manus feature, a user can drop a full list of changes into the chat, close the laptop, and return later to a finished, live site with no review step in between. That is the more consequential shift buried in a product update. Coding agents that generate output for a human to review before deployment keep a person between the work and the audience. An agent that queues its own task list and publishes the result unattended removes that checkpoint entirely.
This is where the ownership noise around Manus stops being gossip and starts being relevant. When an agent can push work to a public URL with nobody in the loop, the operative question is not whether the code is right. It is who answers for it when the code is wrong, and whether that answer changes depending on who owns the company that shipped the feature. A blocked Meta deal and reported Tencent talks do not change how Auto-Publish itself works. They do change who is accountable if an unattended run pushes sensitive data, a broken payment form, or unmoderated content onto a live domain.
Compare this to how established platforms handle continuous deployment. Vercel and Netlify built businesses on automatic deploys triggered by a git push, but that push is still a human-authored commit reviewed in a pull request. Manus collapses that layer further: the “commit” is a chat message, and the reviewer is the agent’s own build-success signal.
For teams testing agentic app builders, the takeaway is narrow but concrete: audit what Manus counts as a successful build before enabling Auto-Publish on anything client-facing. Keep the toggle off for any project where a bad screen reaching a live URL, even briefly, carries real cost.
Manus described Auto-Publish in a post on its company blog at manus.im, published July 14, 2026.
