MCP Response

Marks the end of an MCP tool flow and returns the result back to the AI client.

Inputs

payload any

The result of your tool execution. This becomes the tool's return value to the AI.

ch.mcp.request object

Internal tracking from MCP Tool (automatically passed through). Contains toolNodeId and requestId.

error object | string

When present (for example from a Catch node), the AI receives an error result with this message.

Outputs

This node has no outputs—it sends the result back to the AI client through the MCP Tool node.

Details

Every flow started by an MCP Tool must end with an MCP Response node. This node:

1. Captures the final msg.payload as the tool result
2. Signals completion to the originating MCP Tool
3. Returns the result to the AI client

Return Values

The AI receives whatever is in msg.payload:

• String: AI sees it as text
• Object / Array: sent as JSON the AI can read
• Failure: a message from a ConnectHealth node that failed (payload.success is false) or carrying msg.error is sent as an error result

If the call already ended—it timed out, or the workflow was redeployed—the response is not sent and the node shows a warning.

Flow Pattern

[MCP Tool] → [Your Logic] → [MCP Response]

Examples

Simple Result:

msg.payload = "Patient SYN-123 has 2 recorded allergies";
return msg;  // Goes to MCP Response

Structured Result:

msg.payload = {
  found: true,
  allergies: ["synthetic-allergen-a", "synthetic-allergen-b"],
  count: 2
};
return msg;

Best Practices

• Always end MCP tool flows with MCP Response
• Return meaningful data in msg.payload
• Wire a Catch node here so failures are returned as errors
• Keep responses concise—AI has context limits