Use the MCP Server to Automate Content and Campaign Management
Model Context Protocol (MCP) is an open framework for connecting AI systems to external systems. Use MCP to connect your org to a large language model (LLM). After you make this connection, you can use AI tools such as Claude Code to help manage your marketing content and campaigns.
AI assistants can produce inaccurate or harmful results. Assign the assistant only the permission scopes necessary for it to complete the tasks you ask it to, and review its responses to your requests carefully for accuracy and safety. You assume responsibility for how the outcomes of AI are applied to your organization.
Think of an MCP server as a translator that helps an AI assistant interact with your apps. The MCP server in Marketing Cloud Engagement is a bridge that enables an AI assistant to view your content and contact data securely, and to interact with features such as journeys and automations. The MCP server isn’t the AI model itself, nor is it a storage unit where your data lives permanently. Instead, it’s a temporary interface that only opens when an authorized user accesses it. By using the MCP server in Marketing Cloud Engagement, you can interact with Marketing Cloud Engagement APIs programmatically without writing any code.
Using an MCP server involves three main components:
- The LLM
Think of the LLM as the brain. When you enter a request in your client, the LLM interprets your request, and coordinates with the MCP server to find a way to complete the task. In most client apps, you can select from several LLMs, each with their own specialties and optimizations. For example, in Claude Code, you can choose between the Opus LLM for complex reasoning, the Haiku LLM for fast results, or the Sonnet LLM for a balanced approach.
- The client
The client is the application that you use to interact with the LLM, such as Claude Code or Gemini CLI. It coordinates authentication and communication between the LLM and the MCP server.
- The MCP server
The MCP server receives the request from the LLM and executes the appropriate function in Marketing Cloud Engagement. It returns the results of the operation to the LLM, which shows the results in the client.
To illustrate this relationship, imagine you enter a request in Claude Code (the client), asking to create a marketing email. Claude Code sends the request to Claude Sonnet (the LLM), which generates the HTML code for the email. The LLM then realizes that it can’t create the object in Marketing Cloud Engagement on its own, so it connects to the MCP server. It finds a suitable tool on the MCP server, and then takes the HTML code that the LLM built and uses it to create an email object in Marketing Cloud Engagement.
AI assistants specialize in synthesizing data from multiple sources. By using the MCP server to connect Marketing Cloud Engagement to an AI assistant, you can reduce the amount of time and effort required to perform routine marketing tasks, especially those tasks that require comparing information found in multiple places.
For example, you can ask the assistant to identify which journey had the highest unsubscribe rate over the past week. The assistant automatically retrieves engagement data for all activities in all your journeys that ran over the past week, parses the data, and responds with the correct data, reducing the time required to collect this data from hours to minutes.
As a marketer, you want to make sure that customers engage with your messages, but you also want to avoid inundating them. An AI assistant can provide a high-level view of your campaigns, analyzing the data extensions that your journeys use, and identifying over-targeted customers across all your journeys. The assistant can then remove those customers from lower priority journeys.
AI assistants are also helpful for branding and compliance issues. For example, if you change the URL for your privacy policy, the AI assistant can save you time by finding and replacing all instances of the old link.
Admins can also benefit from using AI assistants with Marketing Cloud Engagement. For example, writing complex SQL for data extensions can be time-consuming and prone to syntax errors. An AI assistant that’s connected to the MCP server can see the schemas for all your data extensions, and then write and test the code for you.
Any third-party AI assistant or LLM that you choose to install and configure with MCP (such as Anthropic’s Claude or Google’s Gemini) is provided directly by that third party, may be subject to applicable terms by that third party, and is not provided, supported, or warranted by Salesforce.