SFMC Tips #300 : The Future of Marketing Cloud Engagement Changed by MCP Servers

In the Summer ’26 new feature release for Marketing Cloud Engagement, MCP Servers will finally become available.

• Previously, July 2026 was explicitly mentioned in the help documentation as the start of a rolling release deployment. However, since that information is no longer displayed, the release timing may currently be undecided.

MCP (Model Context Protocol) is a common standard proposed by Anthropic that allows AI to safely operate external systems. Simply put, it is a mechanism that acts as a “bridge” enabling AI to directly operate external tools.

Going forward, we are entering an era where users can give natural language instructions to Claude or Gemini and operate Marketing Cloud Engagement through MCP Servers.

For example, operations such as the following:

• Create a new Data Extension
• Identify the journey with the highest unsubscribe rate in the past week
• Investigate journeys and email assets
• For a list of things AI will be able to do by utilizing MCP, see here

Marketing Cloud Engagement did not have an Agentforce running on the core platform like Marketing Cloud Next.

However, by connecting Claude or Gemini through MCP Servers, there is a possibility that we are entering an era where external AI can be used like an “operator for Marketing Cloud Engagement”.

In this article, I will actually configure the MCP Server and review the flow up to the start of usage.

Prerequisites

Before starting the steps in this article, install and configure AI assistants that support MCP, such as the following.

• Claude Code
• Gemini CLI

Claude’s free plan can be used to chat with Claude, but Claude Code itself cannot be used. Therefore, in this implementation, I am using Gemini CLI, which is available for free.

Create an Installed Package

To use MCP Servers, create an Installed Package on the Marketing Cloud Engagement side so that AI assistants can connect to it.

1. From the “Setup” screen in Marketing Cloud Engagement, navigate to the following.

Platform Tools ＞ Apps ＞ Installed Packages

2. Click “New”.

3. Enter a package name (example: Marketing Cloud MCP for Gemini) and save it.

• The help documentation instructs you to select “Other” for the category, but currently category selection is not available. 💡

Create an API Integration Component

1. Scroll down within the created Installed Package and click “Add Component”.

2. Select “API Integration” as the component type.

3. Select “Public App” as the integration type and proceed to the next step.

Configure Public App Properties

1. Next, a Redirect URI is required, but it cannot yet be finalized at this stage, so temporarily enter the following value. It will later be replaced with the official URI.

https://salesforce.com

2. Next, configure the permission scopes that will be granted to AI through the MCP Server.

⚠️ Warning: Do not grant more permissions to AI than necessary.

For example, if you want to allow the creation and modification of Data Extensions, grant scopes such as the following.

• Data ｜ Data Extensions ｜ Read
• Data ｜ Data Extensions ｜ Write

Check the Client ID and Tenant ID

After creating the Installed Package, confirm and make note of the following.

Client ID

This is the displayed 24-character alphanumeric string.

Tenant ID

This is the 28-character alphanumeric string included in the Authentication Base URI and similar values.

• For example, if the Authentication Base URI is as follows: https://mcphchq9d5b8mlzeyc2v1example.auth.marketingcloudapis.com/
• Then the Tenant ID is the following value. mcphchq9d5b8mlzeyc2v1example

Configure the Official Redirect URI

Next, click the “Edit” button in the API Integration section.

Using the Client ID and Tenant ID you noted earlier, configure the following official Redirect URI.

• Replace the previously entered “https://salesforce.com” value.
• Replace {tenantId} and {clientId} with the actual values.

① When using the US-hosted MCP Server

https://mai-mce-mcp-cdp1.**sfdc-yfeipo**.svc.sfdcfc.net/t/{tenantId}/c/{clientId}/api/mcp/oauth/callback

② When using the EU-hosted MCP Server

https://mai-mce-mcp-cdp1.**sfdc-yzvdd4**.svc.sfdcfc.net/t/{tenantId}/c/{clientId}/api/mcp/oauth/callback

**In the case of Japan, it is likely acceptable to use the US-hosted MCP Server from a distance perspective. This difference mainly affects connection latency.*

Use MCP Servers with Gemini CLI

Add the MCP Server in Gemini CLI.

gemini mcp add mce -t http https://xxxxx.svc.sfdcfc.net/t/{tenantId}/c/{clientId}

① When using the US-hosted MCP Server

https://mai-mce-mcp-cdp1.sfdc-yfeipo.svc.sfdcfc.net/t/{tenantId}/c/{clientId}/api/mcp

② When using the EU-hosted MCP Server

https://mai-mce-mcp-cdp1.sfdc-yzvdd4.svc.sfdcfc.net/t/{tenantId}/c/{clientId}/api/mcp

Replace {tenantId} and {clientId}* here as well.

• After configuration, when executing the MCP connection from Gemini CLI, the OAuth authentication screen for Marketing Cloud Engagement will open in the browser.
• Once authentication is completed, Gemini CLI will be able to access Marketing Cloud Engagement through the MCP Server.
• As a result, it will become possible to execute operations in Marketing Cloud Engagement while giving natural language instructions to Gemini.

Incidentally, I tried implementing this for testing purposes at the current point in time. Although the MCP registration itself to Gemini CLI succeeded, it remained in a Disconnected state, and I could not connect from a general environment at this time.

Most likely, since I could not select the “Other” category during Installed Package creation, I believe this indicates that it is still before release. 🤔

Verification with Gemini

It should be possible to request the creation of a Data Extension such as the following.

Please create a Data Extension named TestSubscribers in Marketing Cloud Engagement.

Please create the following fields.

• SubscriberKey Text Length 100 Primary Key
• EmailAddress EmailAddress Length 254
• FirstName Text Length 50
• LastName Text Length 50
• CreatedDate Date

Please configure it as a Sendable Data Extension. Please map SubscriberKey to Subscriber Key.

*I plan to test this after July 2026.

Considerations

About Usage Fees

• The MCP Server itself for Marketing Cloud Engagement can be used free of charge.
• However, when using MCP Servers, tokens are consumed on the AI assistant side, such as Claude or Gemini. Usage fees associated with token consumption vary depending on the type of AI being used and the contract between the organization and the AI provider.
• In addition, MCP Servers internally operate by calling Marketing Cloud Engagement APIs. Therefore, the number of API calls on the Marketing Cloud Engagement side is also consumed.
• In Marketing Cloud Engagement, there is a limit on the number of API calls that can be executed per year, and the limit differs depending on the contracted edition.

How was it?

With this MCP Server support, Marketing Cloud Engagement may be entering an era where it can be operated while directly conversing with AI.

In particular, areas that Marketing Cloud Engagement previously lacked, such as:

• Natural language operations
• SQL validation and updates
• Automation of administrative tasks
• Acceleration of analysis

will likely be supplemented by Claude or Gemini.

This may be considered the beginning of an era where AI works “side by side” with Marketing Cloud Engagement administrators and operators.

I believe this update has the potential to significantly change the operational methods of Marketing Cloud Engagement itself going forward.

It may be interesting to start experimenting with small use cases first and explore what kinds of things become possible.

That is all for this time.

Stay tuned for more Salesforce Marketing Cloud tips! 😎

Nobuyuki Watanabe

https://www.linkedin.com/in/nobuyuki-watanabe/ (+Follow me)