Skip to main content

Create MCP

This section introduces how to create, configure, and use MCP in AI Asset of SERVICEME.
Through MCP, users can introduce external capabilities for Agents, test tool availability, publish personal/organizational Agents as MCPs, and invoke these functional plugins in real business scenarios.

Creation Steps​

  1. In the left navigation bar of the AI Asset page, go to the personal MCP page;
  2. Click the "Create" button in the upper-right corner of the page to enter the MCP creation interface;
  3. Fill in the basic information and complete the configuration.

Basic Information Description:

  • MCP Name: Set a unique name for this MCP for identification purposes (limited to 50 characters).
  • MCP Avatar: You can choose a default avatar provided by the system (custom upload is not supported at this time).
  • MCP Description: Briefly describe the functions, application scope, and purpose of this MCP (limited to 200 characters).
  • MCP Category: Select the category to which this MCP belongs for easier management and lookup later.
  • MCP Service Configuration:
    By filling in JSON format code, you can register external services (such as APIs or system interfaces), thereby adding external invocation capabilities to the agent.
    For example, services such as weather, translation, and reporting systems can be connected to give the Agent richer functionality.

💡 Tip: Be sure to save after completing the configuration, otherwise the MCP will not take effect during testing or invocation.

Test MCP​

After the MCP service configuration is completed, you can verify the availability of the MCP through the built-in test function:

  1. On the MCP configuration page, click the "Test" button;
  2. The system will automatically execute the MCP Tools test;
  3. After the test is completed, the "MCP Tools" area on the right will display the list of available tools provided by this MCP.

Testing Recommendations​

  • It is recommended to test each tool one by one to ensure the interface responds properly;
  • If the test fails, check whether the JSON configuration is correct and whether the interface is accessible.

Publish Agent as MCP​

In addition to introducing external resources, you can also publish an Agent you created as an MCP, so that other members within the organization can also use the capabilities of this agent.

Steps​

  1. Create a "Personal Agent" and complete the basic configuration;
  2. Go to the configuration page of that Agent;
  3. In the settings, select "Publish as MCP";
  4. After confirming the publication, the Agent will be displayed on the personal MCP page.

Application Notes:

  • If an organizational MCP is needed, you must first make the personal Agent public as an organizational agent, then convert it into an MCP. After publication, the MCP can be directly referenced by other members within the organization in their own Agents.
  • This method is especially suitable for consolidating and sharing mature agent capabilities (such as automatic report generation, knowledge retrieval, process approval, etc.) within the team, improving overall collaboration efficiency.

Find and Use MCP​

In an already created Agent, you can select and enable existing MCP services for it.

Steps​

  1. On the Agent detailed configuration page, find the "MCP Service" section;
  2. According to business needs, select and enable the corresponding MCP from the list (such as the add function in the calculator plugin);
  3. After adding it, click "Publish" in the upper-right corner.

After successful publication, the Agent can use the capabilities provided by the selected MCP.

For example:

  • Enter "1+1=" in the chat interface,
  • The Agent will perform the calculation through the calculator MCP and display the answer with an MCP invocation identifier.

Practical Suggestions​

  • Quick Start: For business users, it is recommended to first try already published organizational-level MCPs, which can be directly invoked without manual configuration.
  • Flexible Expansion: Users with some technical background can connect external systems through JSON to expand enterprise-level AI capabilities.
  • Collaboration and Sharing: Publishing frequently used Agents as MCPs can improve team reuse efficiency.

Through the above steps, you can flexibly introduce, test, share, and use MCP in SERVICEME, allowing agents to truly integrate into business systems and improving the practicality and scalability of AI automation.