Skip to main content

MCP

MCP is Experimental​

The Model Context Protocol is still under development. These features are provided as-is and are subject to change.

Provide feedback and raise issues for this MCP feature by completing this ~2-minute survey

The Model Context Protocol (MCP) is an open-source standard for connecting AI applications to external systems, introduced by Anthropic. Many modern AI applications (e.g. ChatGPT, Claude, Cursor) leverage MCP to empower their users.

Qualtrics MCP Servers​

Qualtrics provides an MCP server for each of the APIs documented in this API reference. Each server exposes an MCP tool for an API endpoint documented here. Every JSON-based API is supported. Non-JSON APIs are not supported at this time.

Qualtrics users can authenticate their MCP clients with the MCP server using the standard login flow, leveraging OAuth tokens.

Configuring your MCP client​

Various AI tools can act as an MCP client. A list of compatible MCP clients is maintained by the MCP maintainers and is available here.

MCP clients configure their connected servers in slightly different ways. Typically, servers follow the format defined by Anthropic.

// mcp.json
{
  "mcpServers": {
    "qualtrics-surveys": {
      "url": "https://my-brand.iad1.qualtrics.com/API/mcp/survey-crud"
    }
  }
}

We provide a guide on forming the MCP URL to use in your MCP configuration.

In order for your MCP client to connect to Qualtrics, you will need to register an OAuth Client inside Qualtrics for it to use. Ask your Brand Administrator to follow the instructions here to create a new OAuth Client with the Authorization Code grant type. Ask your Brand Administrator to provide you with the Client ID and Client Secret created by that process. Review the documentation for your AI tool's MCP configuration for the right way to securely input your OAuth Client ID and secret.

MCP Scopes​

Your Brand Administrator will specify the maximum set of OAuth scopes that can be authorized using each Client Credential. Make sure that the Client Credential includes the openid scope as well as the scopes for any endpoint you will need to access using MCP. Each user that accesses MCP using this Client Credential will need to separately authorize their AI tool to use these scopes for their user account. For more information on OAuth Scopes in the Qualtrics Public API, including which scopes are necessary for each API, review this page.

An MCP client will need to support logins using OAuth 2, as defined in the MCP specification. This will allow you to log in to your Qualtrics account using your MCP client and authorize your agent to access your Qualtrics account.

Client Credentials Flow​

If your MCP client does not support OAuth 2 logins using the authorization_code flow, you can generate a client_credentials access token using the method outlined here and provide it as a Bearer token in the Authorization header. Note that these tokens will expire after one hour and will need to be rotated manually.

// mcp.json
{
  "mcpServers": {
    "server-name": {
      "url": "https://my-brand.iad1.qualtrics.com/API/mcp/survey-crud",
      "headers": {
        "Authorization": "Bearer <oauth token>"
      }
    }
  }
}