A company's own AI agent, working in an owner's or admin's browser, can now publish the company's declared fields without a person approving each change. It is on by default and can be turned off in Settings, Workspace.
What changed
The profile editor now offers a fifth tool, trooth_profile_publish, while the workspace's publishing permission is on. Trooth checks the permission three times: before the tool is offered, before each call, and again when the change is written. The agent can publish only the fields a company declares. It never changes what Trooth witnessed, the page address, artwork or signed cards. Every publish is credited to the agent in the workspace's activity, and a change that reads as instructions to an AI model, or carries hidden characters, is held for a person instead of going live.
Why it changed
Trooth is built for agents. A company that keeps its record current through its own agent should not have to stop for a click on every change.
Who is affected
- Affected: Company owners and admins who use an AI assistant in their browser. Editors' agents still cannot publish, because editors cannot. Buyers see published changes labeled declared, as before.
- Availability: Every company workspace, in a browser whose assistant supports WebMCP.
- Release stage: Available.
- Areas: Trust Profiles, Company workspace.
What to do
No action required. Nothing is required. To keep publishing a person's decision, an owner or admin turns off Let our AI agent publish changes in Settings, Workspace.
Dates
- Announced: October 4, 2026, the day this entry was written. Entries are never backdated.
- Released: October 4, 2026.
What this does not fix
An agent outside the browser, signing its own requests, cannot publish yet. That path is specified and not built.