We asked Claude a relationship question and let Goodword supply the answer.
The request was to find whether the user knew somebody at a particular company. Other searches had not produced the contact. The Goodword tool found a relevant professional connection and returned enough context to continue the research.
The connection used Model Context Protocol, or MCP. It exposed a defined set of Goodword actions to the assistant rather than giving Claude unrestricted access to the product or its database.
The first actions were straightforward. Search for a person, view a profile, add or retrieve notes, and use relationship context already stored in Goodword.
The test mattered because the work began in a conversation. A person researching a company did not have to stop, open Goodword, repeat the search, and manually carry the result back into Claude.
The assistant still needed to show which source produced the answer. A contact returned by Goodword was not the same as a result from the public web. The user needed to know that the information came from their authorized relationship data.
The tool call worked. Goodword's MCP connection answered a relationship question from within Claude after the other available searches came up empty.
The test also exposed follow-up work. Search results needed consistent profile links, the connection state had to survive a new conversation, and the available actions needed names that made sense without internal product knowledge.
The basic path worked: ask a relationship question in Claude, use Goodword as the authorized source, and receive a result that could support the next step.