I’ve been using AI tools (Claude and ChatGPT, mainly) more and more to manage client work, and I noticed Cloudron doesn’t have an official MCP server or connector yet. Is that something the Cloudron team has thought about? I think it'd be an awesome feature to include soon with the explosive growth of AI.

In the meantime, I actually built my own. It’s basically an MCP server built on top of the Cloudron REST API schema, with the endpoints exposed as tools so the AI can do things like list apps and their versions, check status and updates, install new apps and trigger backups. So far it works well for me, but I haven’t fully vetted it for bugs or security issues yet, so I’m holding off on sharing it for now. I’m hoping to clean it up and post it here in the future though.

One workflow I really want is “backup first, then change”. For example, when an AI agent is about to make changes inside a WordPress app (through a WordPress MCP), it should create a Cloudron backup of that app first, just in case something goes sideways. Having that built in natively would be a great safety net. Right now it's very much a manual process officially, although my own Cloudron MCP seems to be working to take that off my plate for now.

Curious if anyone else has built something similar, or how you’re using AI to manage your Cloudron instances (creating apps, automating backups, maintenance, etc.). Would love to start that discussion and see where people are at in terms of managing their Cloudron with AI (and totally okay if you refuse to let AI touch Cloudron, that's understandable too).