One Video Library Behind Every Front End

What you can do together

• Model video as a field on an entry and render it in every front end that consumes the delivery API, from one library rather than one upload per channel.
• Name the front ends allowed to embed, and refuse the rest, while signed URLs stop a content address from being shared onward.
• Let an agent work against the content model itself through Contentful's server, so an entry can be found, checked or drafted without a connector being built for it.
• Read engagement on embedded media back per surface, so a video's performance on the marketing site is distinguishable from the same video in the help centre.

How it connects

Video becomes another content type. A field on an entry holds the reference to the item in Nexus, modeled and edited in Contentful like any other field, and the media is never copied in. Editors keep working in Contentful, and developers keep consuming the delivery API.

Rendering is the JavaScript widget, placed by the front end wherever that field is filled. The documented set covers video, audio, image, document, collection and playlist players, media library gallery and detail views, insights, transcription, maps, carousel and collage, quiz and survey players and a mashup filter, so a front end can render a browsable library or a search view from the same field. This is the path the shipped WordPress and SharePoint integrations already run on, and it works in any front end that renders HTML.

The domain list matters more here than on a single site. Embedding is restricted to approved domains, so every front end that should render video is named and anything else refused, and content URLs are signed and expire. Each embed resolves against the viewer's identity, so an entry restricted in Nexus renders without its video for someone who may not see it, whichever application asked for it.

Separately, Contentful publishes an MCP server and AI Intelligence Hub connects to MCP servers as a client. The connection is configured once with a name, the server URL, the transport and the headers carrying its credential, and the tool list is fetched when it binds rather than coded in. That credential is the boundary: what an agent can read or change in Contentful is what the credential is granted there.