-
-
Notifications
You must be signed in to change notification settings - Fork 2.5k
Description
Hello,
I'm experiencing issues with n8n-MCP and would appreciate your guidance on the best approach for stable usage with Claude.
Current Situation:
I deployed n8n-MCP in remote mode on my OVH infrastructure (using a subdomain structure like n8n-mcp.core.example.com) because the local mode via npx n8n-mcp on Claude Desktop was causing recurring timeouts on my Mac. The server works perfectly (verified with curl), but requests are being blocked by Anthropic's proxy with a 403 Forbidden error and the header x-deny-reason: host_not_allowed.
Why I Switched to Remote:
Initially, I was using n8n-MCP in local stdio mode on my Mac, but Claude Desktop wasn't stable - sessions would constantly timeout while using MCP tools. So I decided to deploy remotely to have a 24/7 available server accessible from all my devices (Mac/iPad/iPhone), but now I'm hitting Anthropic's proxy blocking my custom domain pattern (*.core.example.com).
My Questions:
Anthropic Proxy: Do you encounter this issue often with custom domains? Is there a procedure to request adding a domain to Anthropic's proxy allowlist?
Claude Code vs Claude Desktop: Would it be more stable to use Claude Code CLI rather than Claude Desktop for working with n8n-MCP remotely? Does Claude Code use the same restrictive proxy?
Overall Stability: Honestly, I find Claude Desktop quite unstable - issues locally (timeouts) AND remotely (proxy blocking). Did I miss something during installation or is this a known problem?
Context: I'm developing complex n8n workflows (multilingual Telegram bot with AI Agent) and really need stable access to n8n-MCP tools for creation, validation, and debugging.
For now, I've temporarily reverted to local stdio mode despite the timeouts, but this isn't a viable long-term solution.
Thanks for any advice on the best configuration to adopt!