Failed to connect to MCP Server

Questions
1

Hi,

My MCP server has suddenly stopped working. It was working fine, but for some reason it has stopped working.

I’ve replicated the issue in a simpler workflow, as you can see in the image below:

Captura de pantalla 2025-05-22 095240
Captura de pantalla 2025-05-22 0952402013×1266 160 KB

The Agent enters into a loop.

The configuration is fine. The console output is:

McpClientTool: Failed to connect to MCP Server
Error in sub-node MCP Cal.com
McpClientTool: Failed to connect to MCP Server
Error in sub-node MCP Client
Received request for unknown webhook: The requested webhook "sheet/sse" is not registered.
McpClientTool: Failed to connect to MCP Server
Error in sub-node MCP Client
Received request for unknown webhook: The requested webhook "sheet/sse" is not registered.
Received request for unknown webhook: This webhook is not registered for HEAD requests. Did you mean to make a GET request?

I’ve tried several times to change the test/prod mode, paths, tools, etc. , but with no success.

Help, please!

  • n8n version: 1.93
  • Database (default: SQLite):
  • n8n EXECUTIONS_PROCESS setting (default: own, main):
  • Running n8n via (Docker, npm, n8n cloud, desktop app): Docker
  • Operating system: Ubuntu 22
2

Hello @motoneurona!
Welcome to n8n community!

Your workflows are likely fine, the problem seems to be in how SSE/webhook connections are handled by your hosting environment. Adjusting execution mode and checking reverse proxy behavior should help.

→ Similar issues in the n8n community and Coolify GitHub

What you can try?

  1. Switch execution mode (if possible)

If you’re using EXECUTIONS_MODE=queue, try changing it to:

EXECUTIONS_MODE=own

This allows n8n to manage SSE/webhook registration directly.

  1. Check your reverse proxy

If you’re deploying with Coolify or have NGINX/Traefik in front of n8n:

  • Make sure /sse and /webhook/* routes are not cached, rewritten, or blocked
  • Ensure SSE connections are supported
  1. Try running the same flow locally

To rule out infra issues:

  • Run n8n locally
  • Activate the MCP Server workflow first, then the MCP Client/Agent
  • If it works locally, the issue is definitely with your cloud/proxy config

Let me know!

3

Solved! It was a problem with Traefik. I use EasyPanel and the version 2.16 introduces a bug solved in version 2.17 ( * removed default compress middleware due to a bug in Traefik).

Once upgraded, everything works properly!

Thank you for you time!

4

I’m running n8n in Docker on a Plesk (IONOS) server behind the built-in Nginx reverse proxy. When my vAPI Voice AI instance calls the MCP Server Trigger endpoint, I get:

502 Bad Gateway
failed to connect to MCP server

I suspect GZIP or buffering is interfering with the SSE connection, so I added a custom Nginx block to disable compression, but the error persists.

Configuration:

  • Plesk edition: IONOS
  • Web servers: Apache + Nginx (reverse proxy)
  • Docker container:
    • Image: n8nio/n8n
    • Port mapping: 127.0.0.1:9001→5678
  • Custom Nginx include: /var/www/vhosts/system/n8n.domain.com/conf/vhost_nginx.conf
location ^~ /mcp/ {
    proxy_pass           http://127.0.0.1:5678/mcp/;
    proxy_http_version   1.1;
    proxy_set_header     Connection "";
    proxy_buffering      off;
    proxy_cache          off;
    proxy_read_timeout   3600s;
    proxy_send_timeout   3600s;
    gzip                 off;
}

Request:

What am I missing in my Nginx or Plesk setup to allow the MCP SSE endpoint to connect successfully without triggering a 502? Any pointers on next steps or additional directives to try?