🚀 Looking for Feedback: Building a Self-Hosted n8n + LLM Setup with MCP (without Claude)

Hey everyone 👋

I’m currently exploring ways to build a fully automated, self-hosted setup where I can describe workflows in natural language and have them created in n8n via MCP (Model Context Protocol). I know there are great Claude integrations out there, but I’d like to avoid Claude and focus on other options – like GPT (ChatGPT Plus), Gemini CLI, or even OpenAI Codex.

My goals:
• Generate n8n workflows via natural language (e.g. “Every morning at 7, get all new AI tools → save to Airtable → send email → publish blog post”)
• Keep the stack low-cost or free (max. 20 CHF/month)
• Have it running 24/7 and accessible from anywhere (not just on local Wi-Fi)
• Use tools like ChatGPT Plus, Gemini CLI, and the SuperAssistant MCP (already tested)

I’m still unsure about the infrastructure:
• Oracle Cloud Free Tier, a cheap VPS, or maybe just a dedicated laptop?
• Has anyone tried this with Hostinger or other shared hosting services?
• What are the gotchas when running n8n and an MCP server together?

I’m also comparing GitHub projects like:
• czlonkowski/n8n-mcp
• salacoste/mcp-n8n-workflow-builder
• Others that let you generate or update workflows directly via an LLM

If you’ve built something similar, or have strong opinions about the best way to structure such a setup, I’d love to hear your thoughts:
• What tools are you using?
• Where do you host it?
• What’s your favorite combo of LLM + n8n + MCP?

Any advice or shared setups would be incredibly helpful. 🙏