Slack MCP node broken after recent n8n Cloud update — HTTP Streamable returns invalid_union JSON-RPC error
Describe the problem/error/question
My Slack MCP node stopped working overnight with no changes made on my end. It had been working reliably for over a month.
What is the error message (if any)?
{ “code”: “invalid_union”, “errors”: […], “path”: , “message”: “Invalid input” }
The error includes "Unrecognized keys: \"ok\", \"error\"" — meaning the Slack MCP server is returning a Slack API-format response instead of JSON-RPC 2.0.
Node config:
-
Endpoint:
https://mcp.slack.com/mcp -
Server Transport: HTTP Streamable
-
Authentication: Bearer Auth (xoxp token)
What I’ve ruled out:
-
Token is valid — a direct curl test returns a clean JSON-RPC 2.0
tools/listresponse -
Tried SSE as a workaround and SSE transport to
https://mcp.slack.com/ssefails with “Could not connect to MCP server” (Slack only supports HTTP Streamable)
My theory: The recent n8n core update that added a GET handler for Streamable HTTP spec compliance (#28787) appears to have broken compatibility with Slack’s MCP server. The GET handshake n8n now initiates is being rejected by Slack’s server, which responds in Slack API format (ok/error) rather than JSON-RPC 2.0.
Environment:
-
n8n Cloud (pendoio.app.n8n.cloud)
-
Slack MCP node as a tool inside an AI Agent node
Has anyone else hit this? Is there a workaround while a patch is in progress?