[TUTORIAL] [Project] Proxmox MCP Full Admin – Manage and monitor Proxmox via AI (Claude Desktop, Cursor, etc.)

rmarquesa

New Member
Nov 4, 2026
1
1
0
Hi everyone,

I’d like to share an open-source project I’ve been developing for the community: Proxmox MCP Full Admin.

It is a Model Context Protocol (MCP) server that allows you to connect Proxmox VE directly to AI interfaces like Claude Desktop, Cursor, or any other MCP-compatible client. The goal is to make managing and troubleshooting Proxmox infrastructure faster and easier using natural language.

Key Features​

  • Full Administration: Monitor node health, storage, and resource usage.
  • VM & LXC Management: Start, stop, reboot, and check real-time status/logs.
  • Natural Language Queries: Ask your AI assistant things like "Which LXC containers are running out of disk space?" or "Restart VM 102 and check its status."
  • Secure API Integration: Uses Proxmox API Tokens for granular control and permissions.

Link & Setup​

You can check out the source code, installation guide, and setup examples here: https://github.com/rmarquesa/proxmox-mcp-full-admin

If you find this tool useful for your lab or workflows, giving the project a star on GitHub would be greatly appreciated and helps keep the development going!

I’d love to get your feedback, suggestions, or bug reports to make this even better for the community.

Thanks!
 
  • Like
Reactions: FrankList80
Hi.

thanks for your work on this, I'd love to make use of it. Do you have step by step instructions on how to set it up, for us non-expert techies?

Thanks!
Tony
 
There is already one of these existing. Search the forum for "proximo".
And it also contains quite a lot of arguments why such a thing isn't a good idea some people think it is. To be fair: Most of them had more to do with the architecture of Proximo and the communication of it's developer and are most likely not valid for this mcp. But there were also some people ( like me ) who have enough reasons to distrust such tools in general.. Your milage may vary though so decide for yourself dear fellow community member:
 
Last edited:
I’d start with a disposable VM or lab node, create a dedicated read-only API token, and test basic queries before allowing restarts or shutdowns. Keep the token limited to only the required paths, enable logging, and have backups plus a manual approval step for anything destructive. For a non-expert setup, a short install guide with screenshots and a safe default config would help a lot.