MCP Client can´t reach MCP Server on n8n

I am experiencing a persistent issue with the MCP (Model Context Protocol) nodes in n8n.
After switching the SSE endpoint in Claude Desktop’s MCP config from my n8n Cloud instance to a Self-hosted instance (Elestio), and then switching back, the MCP Client node is unable to connect to the MCP Server Trigger in both environments.
The issue remains even after restarts, deleting/recreating all workflows, and setting up fresh nodes with current versions.

Another user has reported a nearly identical issue here.


What is the error message (if any)?

Error in sub-node ‘MCP Client’: Could not connect to your MCP server
Server Logs:
Mcp Client Tool: Failed to connect to MCP Server
No transport found for session [SessionID]
Last session crashed

Even though the SSE endpoint returns a valid response in the browser, the MCP Client node always fails to connect.
---

## Please share your workflow

*Minimal example:*

j
![Fail|690x403](upload://4UzQOT5xxhIT8DojDJXoUQ6adaF.png)


{
“nodes”: [
{
“parameters”: {
“path”: “example-path”
},
“id”: “1”,
“name”: “MCP Server Trigger”,
“type”: “@n8n/n8n-nodes-langchain.mcpTrigger”,
“typeVersion”: 1,
“position”: [240, 280]
},
{
“parameters”: {
“sseEndpoint”: “https://your-n8n-domain.tld/mcp/example-path/sse
},
“id”: “2”,
“name”: “MCP Client”,
“type”: “@n8n/n8n-nodes-langchain.mcpClientTool”,
“typeVersion”: 1,
“position”: [480, 280]
}
],
“connections”: {}
}


The issue occurs even in  minimal setup (just MCP Server Trigger and MCP Client).

---

## Share the output returned by the last node

There is no output – the workflow fails with the above error message as soon as the MCP Client node tries to connect.

---

## Information on your n8n setup

* **n8n version:**
  * 1.91.3 (Self-hosted, Elestio)
  * 1.90.2 (n8n Cloud)
* **Database:** PostgreSQL (Elestio), (default for n8n Cloud)
* **n8n EXECUTIONS_PROCESS setting:** default
* **Running n8n via:** Self-hosted (Elestio,, n8n Cloud
* **Operating system:** Linux (Self-hosted: Elestio), n8n Cloud

---

**Additional info:**
The problem started **immediately after changing the SSE endpoint in Claude Desktop’s MCP configuration** (from n8n Cloud to Elestio and back).
Even after complete resets, creating new workflows, and re-adding the nodes, the connection error persists on both instances.
The SSE endpoint is reachable in the browser and returns the expected event/response, but the MCP Client node fails with the session/transport error.

---

If you need more logs or details, I can provide them.
![Fail|690x403](upload://4UzQOT5xxhIT8DojDJXoUQ6adaF.png)
  • Restart n8n and Claude Desktop.
  • If possible, check the session ID or token parameters used by the MCP Client node for the connection.
  • Make sure the MCP Server Trigger on the server side is working properly and that the SSE endpoint is continuously live.

Thank you for your suggestions. I have carefully checked all the steps:

I have restarted both n8n (self-hosted) and Claude Desktop multiple times, including full reboots.

  • There are no manually set tokens or session IDs in the workflow; the node uses only the automatically generated parameters.
  • The MCP Server Trigger is activated, the endpoint is accessible in the browser, and it returns the expected “event: endpoint” response.
  • The problem persists even in minimal workflows, regardless of the instance, network, or browser used. It did work without any issues before I tried to change SSE endpoint in Claude.

Hi Florian,

I had a similar issue. I hosted on Hetzner via Coolify-Dashboard. I resolved the problem by unchecking gzip compression. Watch py post here:

Have a great day!

Thanks! That did it for me as well. I send a message to elast.io support and they made the changes because I didnt have the rights.
My message: Is there gzip or HTTP compression enabled for my n8n service (including on custom domain/SSL proxy)? If so, how can I disable it for API and SSE compatibility?

Have a great day as well

1 Like

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.