Other Considerations for Using the MCP Server

Learn how to use the MCP server for Marketing Cloud Engagement safely and effectively, and find answers to common questions.

Before you let the AI assistant execute a request in your Marketing Cloud Engagement account, it’s a good idea to ask what steps it plans to take. Ask the assistant if it can perform a dry run, show you the code it wants to execute, or state how many records the code impacts. It’s important to keep a human in the loop, especially if the AI assistant has permissions that allow it to rewrite or delete data.

The scopes you select determine the actions that the AI assistant can perform. When you create the installed package that the MCP uses to connect to Marketing Cloud Engagement, select the permission scopes carefully. Assign only the permission scopes that are necessary for the AI assistant to complete the tasks you assign it.

If you aren’t sure which permission scopes to use, take a conservative approach. You can always add permission scopes, but you might not be able to undo changes caused by granted the assistant permissions that are too broad.

You can use the MCP server for Marketing Cloud Engagement for no additional cost. However, when you use the MCP server, your AI assistant consumes tokens. The cost for consuming these tokens varies depending on which assistant you use and your organization’s subscription or contract with the AI provider.

The MCP server works by issuing requests to the Marketing Cloud APIs. There’s a limit on the number of API calls that your account can issue each year. The limit for your account depends on which edition of Marketing Cloud Engagement you use. See API Limits and Guidelines.

Marketing Cloud Engagement returns specific HTTP response codes when it encounters an error. The MCP server passes these response codes back to you. AI assistants can typically help you interpret these codes. You’re most likely to encounter these codes when using the MCP server:

• 401: You aren’t authenticated against the MCP server. Confirm the details of the MCP connection. Make sure that the client secret for your connection isn’t expired. If it is, generate a new secret, and then update your MCP connection to use the new secret.
• 403: You’re authenticated against the MCP server, but the installed package that you use to connect to the MCP server doesn’t include the permissions required to complete your request. Check the permission scopes for the installed package and add permissions if necessary.
• 500: Marketing Cloud Engagement couldn’t process the request. Make sure that the Tenant ID and Client ID values in the MCP configuration are correct in your AI assistant.