One Assistant Across HubSpot and Every Recorded Call

What you can do together

• Ask AI Intelligence Hub for an account's history before a call and get it assembled from HubSpot's records and the recorded meetings in the library, each source cited.
• Have a follow-up drafted from what was said in the recorded meeting, approved by the rep before it goes back to HubSpot.
• Qualify and route an inbound lead against your own criteria in a workflow, with an approval node before any step that writes back or sends something.
• Serve campaign and product video inside HubSpot pages from Nexus, and capture leads with a HubSpot form inside the video itself.

How it connects

HubSpot publishes an official MCP server, and AI Intelligence Hub connects to it as an MCP client. The server is configured once with a name, URL, transport and the headers that carry its credential; its tools are discovered when the connection is bound. In a workflow, an MCP node runs one named tool with arguments fixed at design time. For an agent, an MCP tool hands over the server's tool list, and the agent chooses a tool and fills the arguments as it works. The connection carries the credential it is configured with, so what an agent can read in HubSpot is bounded by what that credential is granted there. Retrieval from the library alongside it still runs under the asking user's own identity.

Where a step needs something the MCP server does not expose, the HTTP Request node calls HubSpot's REST API. In the other direction, the platform's REST API and a workflow's HTTP trigger are the inbound paths, so automation on the HubSpot side can start a workflow.

The JavaScript widget places player and library views inside HubSpot-hosted pages, restricted to approved domains, with the platform's access control still applied. A HubSpot form placed at a point in a video captures a lead, and the submission routes into HubSpot, not the portal.