Skip to main content

MCPServerStart

Controlled node

Overview

The MCPServerStart node is an event trigger that fires when there is a request to start a new MCP (Model Context Protocol) server. This node serves as the entry point for workflows that define or configure MCP servers, allowing you to execute logic when a server initialization is requested.

Unlike the standard Start node which fires when a workflow begins execution, MCPServerStart specifically listens for internal MCP server start events (__internal__mcp-server-start). This makes it essential for building workflows that act as MCP server definitions.

Inputs

This node has no inputs. It triggers automatically when an MCP server start request is received.

Outputs

OutputTypeDescription
firedEventFires when an MCP server start request is received.

Runtime Behavior and Defaults

When the workflow initializes, the MCPServerStart node automatically registers a connection to the internal event __internal__mcp-server-start. When this event is triggered by the system (indicating an MCP server should start), the node executes and fires its fired output event.

The node has no configurable properties or data inputs. It acts purely as a signal that an MCP server initialization has been requested.

Example Usage

Basic MCP Server Workflow

Connect MCPServerStart to nodes that define your MCP server configuration:

[MCPServerStart] --fired--> [MCPServer] --output--> [SetMCPOutput]

In this example:

  1. MCPServerStart fires when an MCP server start is requested
  2. MCPServer node defines the server name, version, and available tools
  3. SetMCPOutput returns the server configuration to the requesting client

Handling Tool Calls

You can also use MCPServerStart to trigger workflows that handle specific MCP tool calls:

[MCPServerStart] --fired--> [MCPServerToolCallEvent] --fired--> [Your Tool Logic]
note

This node is specifically designed for MCP (Model Context Protocol) integrations. It will only fire when the platform receives an MCP server start request, not during regular workflow execution.