Articles on: Integrations
This article is also available in:

How to connect Trustmary to an AI assistant with MCP

Quick steps


  1. In the AI assistant, add a custom MCP server with the address https://mcp.trustmary.io/mcp.
  2. Sign in to Trustmary when asked, choose the organisation on the Authorize access screen, and click Approve.
  3. In a chat, pick the Trustmary connector and mention Trustmary in the request.


About Trustmary's MCP server


Trustmary's MCP server lets an AI assistant such as Claude, ChatGPT or Microsoft Copilot work inside the Trustmary account from a chat: read answers and reviews, build surveys and reports, check who received a survey. MCP (Model Context Protocol) is the open standard these assistants use to reach outside services. The server is hosted by Trustmary, so nothing is installed on the computer: the assistant connects over the internet and signs in with a Trustmary user account. The feature is in beta.


The assistant can do only what the connecting user can do in Trustmary, and only with the tools the organisation's owner has allowed. A new organisation starts with read-only tools.


Add the server in the AI assistant


The server address is the same in every client, and sign-in happens in the browser, so no API key is needed. In a company workspace, custom connections usually have to be allowed by the organisation's admin first, so if the option to add one is missing, ask the admin:


https://mcp.trustmary.io/mcp


Claude (claude.ai and Claude Desktop)


  1. Open Settings from the profile menu, then Connectors, and click + Add → Add custom connector.
  2. Enter a name, such as Trustmary, and the server address, then click Continue. Keep the detected authentication settings, choose Sign in now, and click Continue again.
  3. Click Add.


On a Team or Enterprise plan only an owner can add the connector, under Organization settings → Connectors → Add → Custom → Web; each member then clicks Connect next to Trustmary under Settings → Connectors. Claude Code takes the same address on the command line, followed by /mcp to sign in:


claude mcp add --transport http trustmary https://mcp.trustmary.io/mcp


ChatGPT (web)


  1. Open the Plugins page in ChatGPT, click +, then Add custom MCP server.
  2. Enter a name, paste the server address under Server URL, and choose OAuth as the authentication.
  3. Confirm the warning with I understand and want to continue, click Create as a plugin, and install the plugin.


In a Business, Enterprise or Edu workspace an admin may have to allow custom servers first. OpenAI's own guide is Custom MCP servers in ChatGPT.


Microsoft 365 Copilot


In Microsoft 365 an MCP server is added to an agent in Copilot Studio by an admin or an agent builder, not by each user:


  1. On the agent's Tools page, click Add a tool → New tool → Model Context Protocol.
  2. Enter a server name, a description and the server address. Choose OAuth 2.0 with the type Dynamic discovery, click Create, then Next.
  3. Click Create a new connection, sign in to Trustmary, and click Add to agent. Then publish the agent to Microsoft 365 Copilot.


Any other assistant that supports remote MCP servers with browser sign-in, such as Cursor or GitHub Copilot in VS Code, takes the same address.


Approve the connection


Signing in opens the Authorize access screen in Trustmary, which names the application asking and the tools it gets. Choose the organisation and click Approve to return to the assistant.


The organisation selector and the tool groups the connection gets.


A connection is tied to one organisation. To use another one, disconnect the client on the Connected clients page and connect again.


Use the connector in a chat


Pick the Trustmary connector in the chat: in Claude under + → Connectors, in ChatGPT by typing @ and choosing the plugin, in Copilot by opening the agent. Then mention Trustmary in the request, so the assistant uses the connector rather than answering from memory:


  • Trustmary: which survey has the most answers this month, and what do people complain about?
  • Build a Trustmary report with one card per question of the onboarding survey, and give me a share link.


Before a write, the assistant asks for confirmation; deleting a survey or report always requires one. Every change to a survey keeps the previous version, and the assistant can restore one.


Managing what an assistant may do


Two pages behind the gear icon in Trustmary control the connections:


  • MCP tool access — tools come in the groups Read, Create, Edit, Delete and Manage access. Under My tools each user unticks the tools their own assistants should not use and clicks Save. Under Organization an owner sets the default for everyone with Edit default, or a different set for one user with Edit.
  • Connected clients — every approved connection, with the user, the date it was connected and when it was last used. Disconnect revokes it at once.



Please note: removing a tool takes effect on the assistant's next call, but a tool that has been added shows up only after the client is disconnected and connected again, because a client learns its tool list when it connects.


Wrapping up


The assistant reads what the user can see in Trustmary, and nothing more. Contacts and contact lists, the email sent with a survey, and question-by-question editing of the default surveys stay in the app for now. The tool descriptions on the MCP tool access page are the right thing to forward to an IT department asking what the connection can do.


Questions and feedback go to [email protected].

Updated on: 05/11/2026

Was this article helpful?

Share your feedback

Cancel

Thank you!