Create an MCP server in the Hub MCP Designer

Learn how to design, set up, and deploy an SAP MCP server using the Hub MCP Designer, including connecting your SAP system as a defined remote system, defining back-end logic, and structuring requests and responses within the SAP Integration Hub framework.

Prerequisites

  • You have defined your SAP system as a remote system of type Neptune DXP - SAP Integration Hub in the Remote Systems tool.

  • You have added an authentication method for the remote system.

    Neptune recommends that you enable principal propagation to pass the SAP user identity to the remote system, so authorizations are applied correctly.

Procedure

Create MCP server

  1. In the Cockpit, select the Hub MCP Designer.

    Result: A list of all MCP servers from all configured remote systems of type Neptune DXP - SAP Integration Hub displays.

  2. Select Create.

    Result: The Create new MCP Server dialog opens.

  3. In Remote System, select a configured remote system (only remote systems that support the Hub MCP Designer will be available).

  4. In MCP Server name, enter a meaningful name for your MCP server. This will also be part of the path.

    The server name is sanitized automatically as you type. Spaces are replaced with -, and any character outside the set A–Z, a–z, 0–9, _, and -, is removed.

  5. In Data Provider Class, select the DPC class associated with the remote system, as it provides the back-end ABAP class logic that handles MCP tool requests for the MCP server. The class must implement the interface /NEPTUNE/IF_DXP_MCP_DPC. See Data provider class.

  6. Select Create.

    Result: You are navigated to the General tab of your MCP server.

  7. In Title and Version, enter values for these mandatory fields.

  8. Specify the description and instructions to let MCP clients understand the usage of this MCP server.

    The description and instructions are part of what the AI agent receives when it connects to the server. Write them to give the model a clear understanding of the server’s purpose and when it should use the tools it provides.

  9. In Package, assign your MCP server to a development package that enables transport to another deployment stage system.

In the section Metadata SAP and Metadata External System, any metadata from your connected remote SAP system/external system, if available, are prefilled in the displayed fields.

Configure MCP tools

  1. Go to the Tools tab to see all public methods of the DPC listed as tools. By default, tools are disabled unless explicitly enabled.

  2. Select a tool to open its detail view.

  3. On the General tab, set the tool to Active status and describe what the the tool does and when the AI should use it.

  4. On the Request tab, you can add additional descriptions and specifications on how the parameters are to be used, as needed.

  5. On the Response tab, you can add additional descriptions and specifications on how the AI can use the response data.

Save MCP server, prepare for transport, and validate

  1. When you have configured the MCP server and tools, select Save.

    Result: You are prompted to assign the MCP server to a transport request to transport the MCP server configuration across deployment stage environments.

  2. Select the appropriate transport request from the selection.

  3. In the Inspector tab, use the integrated MCP Inspector user interface to validate your MCP server configuration.

    1. Select List tools to verify that only the tools you have enabled are returned, and that their names and descriptions appear as expected.

    2. Select a tool, fill in the input parameters, and select Run tool to execute a live call against your back-end DPC class. Verify that the response contains the expected data and that the output is valid against the defined schema.

    3. Review the call history to inspect the full request and response payloads for any call made in the session.

    4. Repeat for each tool you want to validate.

Results

  • You have fully defined and validated the MCP sever, linking it to the back-end SAP system, so that it correctly processes requests and returns responses according to your configured tools.

  • The MCP server is transportable across SAP environments and ready for productive use, allowing AI tools to reliably interact with your back-end logic.

Related topics