MCP
Receives MCP tool calls in a Workflow and writes MCP responses back to the caller.
Revision History
0.0.0.13 - Current release.
Setup Notes
Use this Node in a published v2 Workflow that is exposed through the Workflow MCP Server.
Place McpReceive at the start of the Workflow and McpRespond before the Workflow ends.
Provide a stable ToolName on McpReceive so MCP clients can register and call the tool consistently.
Methods
The MCP Node provides one trigger method for receiving MCP tool calls and one connector method for returning the response.
McpReceive
Marks the workflow as being triggered by an MCP request. A workflow will be started when an MCP request is received matching the ToolName.
| Parameter | Type | Description |
|---|---|---|
ToolName |
String | The MCP tool name that the Workflow should expose. |
| Return | Type | Description |
|---|---|---|
Method |
String | The HTTP method used for the MCP request. |
Uri |
String | The full request URI. |
Headers |
Object | The request headers. |
Parameters |
Object | The route and query parameters supplied with the request. |
Body |
Stream | The raw request body stream. |
Tool |
String | The MCP tool name associated with the request. |
Arguments |
Object | The MCP tool arguments parsed from the request body. |
McpRespond
Writes the MCP response for the active request.
| Parameter | Type | Description |
|---|---|---|
StatusCode |
Integer | The status code to return to the MCP caller. |
Headers |
Object | Optional response headers. |
Body |
Object | The response body. This can be a stream, string, JSON object, or another value that Flowgear can serialize to text. |
This method does not return an output value.
Usage Notes
Argumentsis populated from a JSON body property namedargumentsorArguments.- If the request body is empty or cannot be parsed as a JSON object,
Argumentsis returned as an empty object. McpRespondconverts strings and JSON values to a UTF-8 response stream automatically.- This Node relies on the MCP runtime context provided by Flowgear. If the Workflow is run outside that context, the Node fails.