Self-hosted automation and why it belongs on a VPS

Tools like n8n, Open WebUI and Docker make it easy to build your own automation and AI stack. Where that stack actually lives matters more than most people expect — a laptop or home PC quietly undermines the whole point of automating something in the first place.

Why "always on" actually matters

Automation tools like n8n work by reacting to things: a schedule, a webhook from another app, an incoming email. Every one of those triggers assumes the thing listening for them is actually running at the time. A laptop that's closed, asleep, or just turned off for the evening breaks that assumption — the workflow doesn't fail loudly, it just silently never runs.

The reachability problem

Anything that needs to call into your setup — a webhook from a SaaS tool triggering an n8n workflow, or a chat client hitting your Open WebUI instance — needs a stable, public address to call. Most home and business broadband doesn't hand out a fixed public IP by default (many connections sit behind CGNAT, where the address is shared and not really yours to expose), so this often doesn't work at all without extra setup, and can break again the next time the router resets. A VPS has a real, static public IP as standard, so this isn't something you have to solve yourself.

Docker keeps the stack tidy — the VPS keeps it available

Docker itself doesn't care where it runs; its job is keeping n8n, Open WebUI and anything else you add as separate, self-contained containers that don't interfere with each other and can be updated or rolled back independently. What the VPS adds on top is a machine dedicated to running that stack reliably, rather than one also used for browsing, video calls or whatever else a home PC does — with nothing else competing for its resources or uptime.

Where a model like Nous Hermes fits — and where it doesn't

For most automation use cases, though, the model doesn't need to run on the same box at all. Routing model calls through something like OpenRouter — which proxies requests out to hosted LLM providers — means n8n and Open WebUI just need to run reliably and make outbound API calls, not host any model weights themselves. That's a genuinely light workload a standard VPS handles comfortably. If self-hosting the model itself is the actual goal, that's a different, GPU-backed hosting requirement — worth knowing which of the two you actually need before choosing where to run it.

Which setup actually fits

A laptop or home PC is genuinely fine for trying things out — testing a workflow, poking at Open WebUI, seeing whether an idea is worth pursuing further. The moment something else needs to reliably call into that workflow, or you want it running whether or not your laptop is open, that fragility becomes the actual problem, not the tools themselves.