Creating an MCP server in MCP Hub from scratch

In the management console, select the folder for which your account has the serverless.mcpGateways.editor role or higher.

In the list of services, select AI Studio.

In the left-hand panel, select MCP servers and click Create MCP server. In the window that opens, do the following:

1. Under Add Method, select Create.

2. Under Tools, select the type of tool you want to add to the MCP server: HTTPS request, Cloud Functions, or Workflows:

   HTTPS request

   Cloud Functions

   Workflows

   1. In the Tool name field, specify a name for the tool you are creating. Follow these naming requirements:

      • It must be from 2 to 63 characters long.
      • It can only contain lowercase Latin letters, numbers, and hyphens.
      • It must start with a letter and cannot end with a hyphen.

   2. In the Instructions for agent field, enter the required text description for the tool. This description must define conditions for the AI agent to use this tool. Here is an example:

      A tool for creating tickets in a corporate task processing system. Use this 
      tool to log customer complaints, support requests, and 
      error reports. When using this tool, set the `queue`, `type`, `priority`, and `summary` parameters. 
      In the `queue` parameter (queue name), always provide the SUPPORT constant. In the `type` parameter (request 
      type), provide one of the following values: bug (error report), complaint (customer complaint) or 
      support-request (request for consultation). In the `priority` parameter, 
      provide one of these values: low, medium, or high. In the 
      `summary` parameter, provide a text describing the gist of the user request in their own words.
      

   3. In the URL field, specify the endpoint the HTTPS request will be sent to.

      • You can use jq templates to add path parameters to the endpoint value. Here is an example:

        https://console.yandex.cloud/folders/\(.folder-id)
        

        In this example, the folder-id field must also be added to the Tool parameters section of the MCP server settings.

      • To send query parameters of the endpoint request:

        • Add relevantly named fields to the Tool parameters section of the new MCP server settings, e.g., country, city, and name.
        • Add relevant query parameters to the Tool parameters section of the new MCP server settings. Specify jq templates with the names of the fields mentioned above as values for the new query parameters, e.g., \(.country), \(.city), and \(.name).

   4. In the Method field, select the HTTP request method: GET, POST, DELETE, PATCH, OPTIONS, or HEAD.

   5. Expand the Advanced section and specify the authentication data to provide in HTTPS requests:

      • Service account: To provide the IAM token issued for the service account attached to the MCP server.
      • Access token: To provide the specified access token.
      • Without authorization: To skip authentication.

   6. Under Tool parameters, specify the names, types, and text descriptions of parameters the tool will use.

      You can add the parameters one by one using the Add button on the Form tab or provide them on the JSON schema tab using the JSON structure below.

      JSON schema:

      {
        properties?: { [key: string]: object };
        required?: string[];
        type: "object";
      }
      

      JSON struture example

      {
        "type": "object",
        "properties": {
          "location": {
            "type": "string",
            "description": "City name or zip code"
          }
        },
        "required": ["location"]
      }
      

   7. Under HTTPS request settings, configure the HTTPS request settings:

      • Under Request headers, specify the names and values ​​of the headers that will be provided in requests.

      • Under Query parameters, specify the names and values ​​of the parameters that will be provided in the request.

      • Under Request body, review an example of the request you got and adjust it if required.

        By default, an empty request body is generated with a flat JSON object consisting of pairs of tool parameters and their values. For example:

        {
          "parameter-a": <parameter-a_value>,
          "parameter-b": <parameter-b_value>,
          "parameter-c": <parameter-c_value>
        }
        

      All HTTPS request parameters support templating: their values can be generated dynamically. The templating language is jq. For more information, see the jq guide.

      For example, the \(.city) value of the HTTPS request parameter will be taken from the city tool parameter. In addition, you can use the syntax Bearer \(.token) for the Authorization header to set up authorization via a token provided in the tool parameter.

   1. In the Tool name field, specify a name for the tool you are creating. Follow these naming requirements:

      • It must be from 2 to 63 characters long.
      • It can only contain lowercase Latin letters, numbers, and hyphens.
      • It must start with a letter and cannot end with a hyphen.

   2. In the Instructions for agent field, enter the required text description for the tool. This description must define conditions for the AI agent to use this tool. Here is an example:

      This tool recognizes text in an image transmitted in base64 encoding. In the recognized 
      text, the tool extracts "article":"quantity" pairs and returns them in JSON structure format.
      

   3. In the Function field, select a function from Cloud Functions that will handle the requests and its version.

   4. Under Tool parameters, specify the names, types, and text descriptions of parameters the tool will use.

      You can add the parameters one by one using the Add button on the Form tab or provide them on the JSON schema tab using the JSON structure below.

      JSON schema:

      {
        properties?: { [key: string]: object };
        required?: string[];
        type: "object";
      }
      

      JSON struture example

      {
        "type": "object",
        "properties": {
          "location": {
            "type": "string",
            "description": "City name or zip code"
          }
        },
        "required": ["location"]
      }
      

   By default, an empty request body is generated with a flat JSON object consisting of pairs of tool parameters and their values. For example:

   {
     "parameter-a": <parameter-a_value>,
     "parameter-b": <parameter-b_value>,
     "parameter-c": <parameter-c_value>
   }
   

   1. In the Tool name field, specify a name for the tool you are creating. Follow these naming requirements:

      • It must be from 2 to 63 characters long.
      • It can only contain lowercase Latin letters, numbers, and hyphens.
      • It must start with a letter and cannot end with a hyphen.

   2. In the Instructions for agent field, enter the required text description for the tool. This description must define conditions for the AI agent to use this tool. Here is an example:

      This tool invokes a workflow to automatically summarize long text. Provides 
      the source text, maximum summation length, and language.
      

   3. In the Workflow field, select the Workflows workflow that will handle requests.

   4. Under Tool parameters, specify the names, types, and text descriptions of parameters the tool will use.

      You can add the parameters one by one using the Add button on the Form tab or provide them on the JSON schema tab using the JSON structure below.

      JSON schema:

      {
        properties?: { [key: string]: object };
        required?: string[];
        type: "object";
      }
      

      JSON struture example

      {
        "type": "object",
        "properties": {
          "location": {
            "type": "string",
            "description": "City name or zip code"
          }
        },
        "required": ["location"]
      }
      

   By default, an empty request body is generated with a flat JSON object consisting of pairs of tool parameters and their values. For example:

   {
     "parameter-a": <parameter-a_value>,
     "parameter-b": <parameter-b_value>,
     "parameter-c": <parameter-c_value>
   }
   

3. To add an additional instrument to the MCP server, click Add instrument.

   Note

   One MCP server can contain up to 50 tools.

4. Under Server parameters:

   1. In the Name field, enter a name for the new MCP server. Follow these naming requirements:

      • It must be from 2 to 63 characters long.
      • It can only contain lowercase Latin letters, numbers, and hyphens.
      • It must start with a letter and cannot end with a hyphen.

   2. Optionally, add a description and labels for the server you are creating by using the corresponding buttons.

   3. In the Access field, select the server type:

      • Private: To build a private MCP server with the authentication-based access.
      • Public: To build a public MCP server that requires no authentication.

   4. In the Service account field, select the service account your MCP server will use to access Yandex Cloud services and resources. The service account should get the roles sufficient to access these resources and services.

   5. Optionally, enable Specify network to specify a cloud network to host your MCP server instances.

   6. Optionally, turn on the Enable logging option and configure the logging settings to keep a log of the MCP server you are creating.

5. Click Save.