Stop rebuilding the same CMS plumbing

Evidence to request: Through the MCP Agent Bridge, an external agent reads the installed theme, page-type and layout catalogue, and typed validation rejects invented components before mutation. That bridge invokes no model; the separate optional AIOrchestrator can call configured providers and carries its own credential and cost boundary.

Alternative fit: For a five-page brochure this is more structure than the job needs. It earns its keep once the same shapes repeat across dozens of pages and several hands touch them.

Inspect this behaviour: The HTML-versus-blocks choice is a real per-page toggle editors own, not a site-wide setting a developer has to change, so editors get tidy block editing by default and hand-written HTML exactly where editors need it. Because page types share one model, the fields sit where editors expect and the daily loop stays predictable.

A different tool may fit when: On a tiny fixed site editors would never feel this, because there is nothing much to keep consistent. The payoff shows up as the site and the number of page types grow.

The demonstration should show: A new page subject, an Event or Product backed by a real model, registers through a typed call rather than a core edit, so agency teams extend a client's content model in a package agency teams reuse on the next one. That single shared model makes a year-three revisit billable work, not an excavation.

Scope check: A hosted CMS offloads that maintenance entirely, at the cost of ever really owning the thing. If a client wants none of the upside of ownership, the hosted route is the honest recommendation.