Hardware probe agent: SSH-based system analysis with AI upgrade recommendations

I run a small home lab — a desktop workstation, a couple of VMs on Proxmox, and an elderly NUC that I keep meaning to retire but never quite do — and I found myself in a situation that's probably familiar to anyone who accumulates hardware over time: I had no clear picture of what was actually in each machine, what was upgradeable, and where my money would be best spent if I wanted to improve performance for local AI workloads. I could SSH into each box and run a series of commands manually, but that's tedious, easy to forget steps, and doesn't give you the kind of comparative overview you need to make smart upgrade decisions. So I built the Hardware Probe Agent: a Python tool that connects to any machine via SSH and generates a comprehensive hardware report with AI-focused upgrade recommendations.

What it probes and why

Point it at any system accessible via SSH — either by IP address or SSH config alias — and it runs through everything: CPU model and core count, GPU (including whether it's actually usable for AI workloads), RAM capacity and type, motherboard model and expansion slots, storage devices and their speeds, networking interfaces, USB devices, security status including TPM and firewall configuration, and even virtualization detection so it handles VMs and bare-metal differently. It outputs both raw JSON data for programmatic consumption and a formatted markdown report that's actually pleasant to read. The JSON is useful if you want to build dashboards or comparison tools on top of it; the markdown is what I actually look at most of the time.

The AI workload angle

What makes this more than a generic hardware inventory tool is its focus on AI suitability. I wasn't just asking "what hardware do I have?" — I was asking "which of my machines could run a 7B parameter model locally, and what's the cheapest way to make the others capable of it?" The generated reports include GPU assessments specifically for AI workloads (not just gaming benchmarks), cost-effective upgrade pathways with component compatibility analysis so you don't accidentally recommend RAM that the motherboard can't take, actual part numbers and estimated pricing where possible, and software optimisation suggestions covering drivers, CUDA setup, and containerisation options. The result is a report that tells you not just what you have, but what you could have for a given budget — which is the question I actually needed answered.

Setup uses a shell wrapper script and works best with SSH key-based authentication and sudo access on the target systems for deeper hardware detection, though it handles permission denials gracefully rather than crashing. You can probe individual machines or batch-process your whole LAN. The tool uses paramiko for SSH connectivity and requires only standard Linux utilities on the target machines — no agents or special software to install on the boxes you're scanning. The full project with setup scripts is on GitHub.