Publish and manage MCP Tools

Add a tool

1. Select the application and application version, and configure basic information such as the tool name and its parent MCP Server.

   

   Parameter	Description
   Application and application version	Important Relationship between a tool and an RPA application: You can create MCP Tools only from published applications. An application has a one-to-one relationship with a tool. Similarly, an application version has a one-to-one relationship with a tool version.
   Tool name	The tool name can be up to 100 characters long. Chinese characters are not supported. We recommend using underscores (_) to separate words. Important We recommend that you do not use the application name as the tool name. In scenarios that involve on-premises robots, users can also filter and run applications by name in the robot client. The large model uses the tool name and tool description to find the required tool. Therefore, the tool name should be easily understood by the large model.
   MCP Server	You can select only a business MCP Server.

2. Provide the tool description and input parameter information for this tool version.

   

   Parameter	Description
   Tool description	The product provides a default description template. You can use this template as a starting point and modify it as needed.
   Input parameters	Input parameters fall into two categories: automation process input parameters and system input parameters. System input parameters cannot be modified or deleted. You define the name, type, and required status of an automation process input parameter during the process development stage. Important In this step, the system validates the parameter name and type. If validation fails, you cannot proceed. Follow the error message to modify the parameters in the editor's Parameter Panel, republish the application as a new version, and then add a tool for the new version. You must define or modify the parameter name, type, and required status in the editor's Parameter Panel, as shown in the following figure.

   • Automation process input parameters

     Item	Details
     Parameter name	Define and edit in the editor's Parameter Panel. Format requirements: Case-sensitive letters, digits, or underscores. The name must start with a letter and cannot end with an underscore.
     Parameter type	Editor Parameter Panel Note When you develop an MCP Tool with RPA, we recommend that you use the "Text Box" type for all parameters, as shown in the preceding figure. Perform format conversions as needed within the automation process. This approach minimizes issues caused by format conversion during data transmission. Tool parameter type The MCP protocol uses JSON-RPC to describe tool definitions. Therefore, when you add tool input parameter information in this step, use only basic JSON types and JSON Schema. For detailed recommendations on how to use specific parameter types, see the following sections.
     Required	Define and edit in the editor's Parameter Panel.
     Parameter description	Provide a detailed description to help the large model understand the purpose of this input parameter and any additional requirements, such as format constraints.

     Parameters can be categorized into three types based on their purpose: basic types, structured strings, and files.

     Basic types

     Supported basic types include string, integer, date, and datetime. You can also set validation conditions for these types, such as default values and enumerated values (enums).

     Structured strings

     Automation processes often need to operate on a set of data, such as sending different messages to multiple people. In such cases, an input parameter must have a specific data structure. When you develop an MCP Tool with RPA, keep input parameter structures simple to help the large model generate tool inputs.

     For structured input parameters, select the JSON structure type and follow the instructions to define it.

     We recommend that the parameter structure does not exceed the complexity of the following examples:

     Type	Example
     Simple array	["1","2","3"]
     Two-dimensional array	[ ["1","2","3"], ["a","b","c"] ]
     Key-value structure	{ "name":"Tom", "age":21, "address":"Beijing" }
     Array of key-value objects	[ { "name": "Tom", "age": 21, "address": "Beijing" }, { "name": "Jerry", "age": 19, "address": "Hangzhou" } ]

     Generate parameter definitions. You can use the features on the page to automatically generate parameter definitions.

     • Select JSON structure as the parameter type to open the JSON definition window. If you are familiar with JSON Schema, you can define the data structure directly in the window or import existing JSON Schema content.

     • 

     • You can also use the Quick JSON Data Import feature to automatically generate the corresponding JSON Schema from sample JSON data.

     Note

     • We recommend that you define a schema for structured input parameters to help the large model better understand and construct the required parameters.

     • To generate the definition, we recommend using the Quick JSON Data Import feature to automatically generate the schema from your data.

     Files

     Automation processes often need to handle files, such as reading data from an Excel file to fill out a web form. In such cases, you need to consider how to pass file content to the automation process.

     • Method 1: If the file content is small and has a clear structure, such as an Excel or CSV file, consider parsing the file content and passing it as a structured input parameter, as described earlier.

     • Method 2: If the file is large, we recommend passing a link to the file. The input parameter for the automation process would be a file URL, for example, a file stored in OSS. The process then downloads and parses the file to complete subsequent operations. For more information, see Download and upload files by using OSS.

   • System input parameters

     Parameter name	Type	Required	Description
     RobotId	string	Yes	Specifies the robot that runs the automation task. You can invoke both on-premises robots and service robots by using MCP. For more information about robot allocation strategies, see AI agent integration recommendations.
     TaskPriority	integer	No	The priority of the automation task. This is an integer from 0 to 100. The default value is 50. This priority applies only to the robot's task queue. A higher number indicates a higher priority.

3. (Optional) Configure callback settings.

   1. After you configure a callback, the system triggers a callback notification when the status of a task created by this MCP Tool changes. The callback notification details depend on the robot type. For a service robot, see CreateTask. For an on-premises robot, see createServiceTask.

   2. When you configure a callback URL, the system sends an HTTP POST request to the URL to verify connectivity.

      Important

      The validation method sends an HTTP POST request with an empty body to the URL over the internet. The test passes if the HTTP response status is 200.

4. (Optional) Publish to an MCP gateway instance. A tool must be published to an MCP gateway instance before you can invoke it. You can publish the defined tool directly in this step.

5. (Optional) Use a large model to evaluate the MCP Tool. This helps you refine the tool's information.

Evaluate a tool

Before a large model invokes an MCP Tool, you can use the AI-assisted evaluation feature. This feature assesses, from the model's perspective, whether the large model can accurately understand and use the tool's information. You can then decide whether to make changes based on the evaluation results.

• The evaluation covers the tool name, description, and input schema of this MCP Tool version.

• The evaluation process is static and does not trigger an actual MCP Tool invocation.

• The following figure shows an example of an evaluation result:

Manage tool versions

Testing and production in development iterations

After you publish a tool, continuous development and iteration may involve the following types of changes:

• Modifications to the tool's information, such as its description.

• Adjustments to the automation process, which results in a new version of the corresponding RPA application.

In this situation, you need to test the tool before you deploy it to production. Similar to how you handle RPA applications (for more information, see 2. Test and publish an application to production), you can create an MCP Server and MCP Tool specifically for testing. The following table uses get_city_weather as an example.

Other operations

• Modify a specific tool version: You can modify the tool description and input information. After you make changes, you must publish the tool to the MCP gateway instance for the changes to take effect.

• Modify the tool name: You must take all versions of the tool offline before you can modify its name.

• Delete a tool version: You must take the version offline before you can delete it.

• Delete a tool: After you delete all versions of a tool, the system automatically deletes the tool.

Authentication and permission scope

All MCP requests require authentication. Anonymous invocations are not supported. You can manage credentials in the MCP Server - credential management section.

The permission scope for an MCP invocation depends on the robot type. For more information about robot types, see Choose a robot type.

• on-premises robot

  • You must request and receive approval for the scope of applications that a robot can run for a specific user account. For more information, see Publish and manage applications. Users can trigger applications by methods such as manual execution in the client or API calls.

  • If you use MCP to invoke this type of robot, the MCP Client can still retrieve the list of all tools on the MCP gateway instance. However, tool execution may fail if the token does not have permission to run the required application.

• service robot

  • You can trigger a service robot by using API calls. The API can run all published applications, and your upper-layer system is responsible for controlling the scope of execution.

  • Therefore, when you use MCP to invoke this type of robot, the MCP Client can invoke all tools.

Get MCP connection information

In the MCP Server menu, you can use "Get MCP Server Connection Information" to obtain the invocation URL and connection details for a specific MCP gateway instance.