MCP Connection to Claude

[mcdougaj]

Bug Description

MCP workflows do not trigger, and no communication occurs between n8n and Claude.

To Reproduce

Open Claude
Active MCP Server Supergateway fails

Expected behavior

Until the latest upgrade - n8n and Clause would communicate

Operating System

windows

n8n Version

1.94.1

Node.js Version

v22.12.0

Database

PostgreSQL

Execution mode

main (default)

[Joffcom]

Hey @mcdougaj,

We have created an internal ticket to look into this which we will be tracking as "GHC-2237"

[jeanpaul]

Hey, it seems that supergateway has some issues (see also here)

I’ve now switched to mcp-remote and that resolved my issues. The MCP config in Claude Desktop now looks like this:

{
  "mcpServers": {
    "n8n": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "<URL>",
        "--header",
        "Authorization: Bearer ${AUTH_TOKEN}"
      ],
      "env": {
        "AUTH_TOKEN": "<TOKEN>"
      }
    },
  }
}

Please let me know if this resolves your issues!

[mcdougaj]

TRIED THIS PS C:\Users\mcdou> npx mcp-remote https://mcdou.app.n8n.cloud/mcp/claude/sse --header "Authorization: Bearer?????" [5608] Using automatically selected callback port: 25129 [5608] Using custom headers: {"Authorization":" Bearer ????"} [5608] [5608] Connecting to remote server: https://mcdou.app.n8n.cloud/mcp/claude/sse [5608] Using transport strategy: http-first [5608] Received error: Error POSTing to endpoint (HTTP 404): {"code":404,"message":"This webhook is not registered for POST requests. Did you mean to make a GET request?"} [5608] Recursively reconnecting for reason: falling-back-to-alternate-transport [5608] [5608] Connecting to remote server: https://mcdou.app.n8n.cloud/mcp/claude/sse [5608] Using transport strategy: sse-only [5608] Connection error: SseError: SSE error: Non-200 status code (403) at _eventSource.onerror (file:///C:/Users/mcdou/AppData/Local/npm-cache/_npx/705d23756ff7dacc/node_modules/mcp-remote/dist/chunk-ZLWDMQIP.js:6215:23) at EventSource.failConnection_fn (file:///C:/Users/mcdou/AppData/Local/npm-cache/_npx/705d23756ff7dacc/node_modules/mcp-remote/dist/chunk-ZLWDMQIP.js:6118:53) at file:///C:/Users/mcdou/AppData/Local/npm-cache/_npx/705d23756ff7dacc/node_modules/mcp-remote/dist/chunk-ZLWDMQIP.js:5960:74 at process.processTicksAndRejections (node:internal/process/task_queues:105:5) { code: 403, event: { type: 'error', message: 'Non-200 status code (403)', code: 403, defaultPrevented: false, cancelable: false, timeStamp: 989.1135 } } [5608] Fatal error: SseError: SSE error: Non-200 status code (403) at _eventSource.onerror (file:///C:/Users/mcdou/AppData/Local/npm-cache/_npx/705d23756ff7dacc/node_modules/mcp-remote/dist/chunk-ZLWDMQIP.js:6215:23) at EventSource.failConnection_fn (file:///C:/Users/mcdou/AppData/Local/npm-cache/_npx/705d23756ff7dacc/node_modules/mcp-remote/dist/chunk-ZLWDMQIP.js:6118:53) at file:///C:/Users/mcdou/AppData/Local/npm-cache/_npx/705d23756ff7dacc/node_modules/mcp-remote/dist/chunk-ZLWDMQIP.js:5960:74 at process.processTicksAndRejections (node:internal/process/task_queues:105:5) { code: 403, event: { type: 'error', message: 'Non-200 status code (403)', code: 403, defaultPrevented: false, cancelable: false, timeStamp: 989.1135 } } PS C:\Users\mcdou> Error

…

On Thu, May 29, 2025 at 4:42 PM jeanpaul ***@***.***> wrote: *jeanpaul* left a comment (n8n-io/n8n#15831) <#15831 (comment)> Hey, it seems that supergateway has some issues (see also here <supercorp-ai/supergateway#67>) I’ve now switched to mcp-remote and that resolved my issues. The MCP config in Claude Desktop now looks like this: { "mcpServers": { "n8n": { "command": "npx", "args": [ "mcp-remote", "<URL>", "--header", "Authorization: Bearer ${AUTH_TOKEN}" ], "env": { "AUTH_TOKEN": "<TOKEN>" } }, } } Please let me know if this resolves your issues! — Reply to this email directly, view it on GitHub <#15831 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BIIOQXBZ27HT5VKLUKW2KZL3A5WLTAVCNFSM6AAAAAB6FHGJB6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDSMRQGUZTONRZG4> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[jeanpaul]

Hi, this sounds like an issue with the authentication (403 means "forbidden"). Please check all the parameters to see that there are no spaces in places they shouldn't be.

[mcdougaj]

I have tried all I know to try to get this working, nothing is working.

…

On Fri, May 30, 2025 at 8:06 AM jeanpaul ***@***.***> wrote: *jeanpaul* left a comment (n8n-io/n8n#15831) <#15831 (comment)> Hi, this sounds like an issue with the authentication (403 means "forbidden"). Please check all the parameters to see that there are no spaces in places they shouldn't be. — Reply to this email directly, view it on GitHub <#15831 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BIIOQXHRNFK3U2LNCCRZ6533BBCTBAVCNFSM6AAAAAB6FHGJB6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDSMRSGIYTGMJUHA> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[marcel-more]

Same Issue here: worked perfectly the week before but suddenly throwing errors. Tried update to n8n Version 1.94.1 but no luck at all.

I tracked down all components, but it comes down to the simple fact that n8n is not responding on the configured MCP server endpoint anymore. Even when CURL by GET directly on the Endpoint URL without any MCP Client involved, there is no valid answer from n8n anymore, instead getting HTTP 530 response.

The logs state that a connection to the MCP endpoint now leads to an error.
Docker log for n8n shows suspicious entries:

Received request for unknown webhook: This webhook is not registered for HEAD requests. Did you mean to make a GET request?

Received request for unknown webhook: The requested webhook "GET 4c206807-d45b-40c6-bd53-043df9d087c5/sse" is not registered.

It looks like the MCP server node behavior changed and webhook registration is broken. The MCP endpoints that were automatically created before are no longer being registered as webhooks or behaving different than expected.

Can anyone confirm this? Has the MCP server implementation changed between versions?

Environment:

• n8n: tested with n8n v1.94.1

When changing back to v1.90.0 the issue is not present!
There might be some breaking change in the newer n8n versions.