Workflow MCP Server

The Workflow MCP Server allows published Flowgear Workflows to be exposed as MCP tools.

This works well for two broad categories of scenarios:

• Utility Workflows - perform a specific task such as approving a purchase order, listing service tickets or updating the status of a record.
• Exploratory Workflows - allow an AI agent to progressively find the data it needs by creating three MCP tool workflows - one to list available entities (e.g. tables), one to obtain the structure of an entity (e.g. fields) and one to run a query.

To use the Workflow MCP Server tools, register the Workflow MCP Server at https://your-environment-hostname.flowgear.net/mcp, where your-environment-hostname is the subdomain of the Environment you want to target.

Steps

Create an MCP Tool Workflow

1. Create a new v2 Workflow.

2. Click + to add a Step, choose the MCP Node, and select the McpReceive method. Enter the tool name in the ToolName property.

3. Click + to add a Step, choose the MCP Node, and select the McpRespond method.

4. Add the Steps that perform the work between the receive and respond Steps.

5. Click ... on each MCP Step, choose Edit Properties, and define the input and output Properties.

6. Save the Workflow.

Example

Paste this summarize-request example into the Workflow Code view. It accepts typed values and returns them with a confirmation message.

Parameters: {}
Steps:
  - Name: "receiveTool"
    Description: "Receives a request for the summary."
    Node: "v2.Mcp@1.0.0.0"
    Method: "McpReceive"
    Parameters:
      ToolName:
        Type: "string"
        Sample: "summarize-request"
        Value: "summarize-request"
    Returns:
      Arguments:
        Type: "object"
        Properties:
          requestId:
            Type: "string"
            Sample: "REQ-001"
            Value: "REQ-001"
          quantity:
            Type: "number"
            Sample: 123
            Value: 123
          approved:
            Type: "boolean"
            Sample: true
            Value: true
          metadata:
            Type: "object"
            Properties:
              region:
                Type: "string"
                Sample: "ZA"
                Value: "ZA"
              priority:
                Type: "number"
                Sample: 2
                Value: 2
  - Name: "respondTool"
    Description: "Returns the request summary."
    Node: "v2.Mcp@1.0.0.0"
    Method: "McpRespond"
    Parameters:
      StructuredContent:
        Type: "object"
        Properties:
          requestId:
            Type: "string"
            Expression: "{receiveTool.Arguments.requestId}"
          quantity:
            Type: "number"
            Expression: "{receiveTool.Arguments.quantity}"
          approved:
            Type: "boolean"
            Expression: "{receiveTool.Arguments.approved}"
          metadata:
            Type: "object"
            Expression: "{receiveTool.Arguments.metadata}"
            Properties:
              region:
                Type: "string"
                Expression: "{receiveTool.Arguments.metadata.region}"
              priority:
                Type: "number"
                Expression: "{receiveTool.Arguments.metadata.priority}"
          message:
            Type: "string"
            Value: "Request received."
      IsError:
        Type: "boolean"
        Value: false
Returns: {}

Enable the Workflow and Configure Access

1. After saving the Workflow (and publishing it to the correct Environment if you're using Release Management), click <- to go to the landing pane. Choose Workflow Status and tick the checkbox against the Environment you want to enable the Workflow for (e.g. Test).

2. Create a Delegated MCP User Key by going to API Keys, and then New API Key. Choose Delegated MCP User for API Key Type.

3. Provide a name for the key, choose the correct Environment, choose the list of users who should be able to run the Workflow from the Permitted Users list, and choose the list of Workflows (i.e. the one you just created) from the Assigned Workflows list.

4. Click Save.

5. Follow the steps at Connect Flowgear MCP to Claude or Connect Flowgear MCP to ChatGPT to start using the MCP Tool.

See also

MCP Server
Connect Flowgear MCP to Claude
Connect Flowgear MCP to ChatGPT