From Homelab to Revenue: The Self-Hosted AI Stack I Packaged for Sale

From Homelab to Revenue: The Self-Hosted AI Stack I Packaged for Sale
Affiliate disclosure: This article contains affiliate links. If you click and purchase through one, we may earn a small commission at no additional cost to you.

AI assistance: Drafted with AI assistance and edited by Auburn AI editorial.

If you’ve been running n8n on a Raspberry Pi or a Proxmox VM and wondering “could I actually turn this into something,” here’s my honest answer from 4 months of doing exactly that.

Running n8n self-hosted on a Hostinger VPS alongside Claude API calls (Sonnet and Haiku), we built out automation covering six WordPress blogs, a Canadian finance data site, a KDP colouring book pipeline, and a BTC/CAD trading signal engine – the whole stack sitting comfortably under 2GB of RAM and costing less per month than a Netflix subscription. From our experience, that kind of resource efficiency surprises people who assume serious automation workloads require serious infrastructure spend. It doesn’t, and this post walks through exactly how the stack is structured and what it took to package it for sale.

Two weeks ago I started packaging the pieces that were solid enough to hand off, and launched them on Gumroad. If you’re the kind of homelab person who would rather deploy someone else’s working n8n flow than spend a weekend building one, these might save you some time.

The full stack — $47

Four complete n8n workflow JSONs: content generator, SEO metadata builder, internal linker, and a category auto-assigner. Plus the Claude prompt library, WordPress REST API wiring, and a Google Sheets keyword queue template. Imports cleanly into any n8n instance (self-hosted, n8n.cloud, docker — I test on all three).

Pairs with Claude Sonnet at ~$5-15/month in API costs at my pace (3 posts/day across several sites). You can downgrade to Haiku and drop it to $1-3/month if you’re cost sensitive.

n8n + Claude Blog Automation Stack on Gumroad

The research agent — $9

A standalone n8n workflow I built to replace my “open 30 Chrome tabs researching a new idea” habit. Takes a topic, returns a structured Notion doc with competitor survey, pricing comparison, keyword gaps. Three depth modes so you can trade compute time against thoroughness.

$9 because it’s a single-purpose utility. Good gateway into the bigger stack if you want to test the approach before committing.

AI Research Agent on Gumroad

The dashboard — $27

Notion-based ops dashboard specifically for solo operators running multiple product lines. Pulls from Gumroad, KDP, AdSense, affiliate networks via free Zapier tier. The important part: it doesn’t have any “team” or “delegation” fields, so you’re not scrolling past garbage to see the five numbers that matter.

Solopreneur Ops Dashboard on Gumroad

The keyword pack — $27

500 pre-researched blog post titles with buyer intent, across 10 niches (SaaS, AI, no-code, PM, email, CRM, analytics, content, automation, dev tools). If you’re trying to spin up a niche site and don’t want to spend a weekend in Ahrefs, this is most of that work done.

500 Buyer-Intent Blog Titles on Gumroad

The coloring book — $5

Outlier product. I publish coloring books on Amazon KDP and this is a 22-page AI-themed one. Listed the digital PDF separately on Gumroad at $5 because Gumroad buyers and Amazon buyers are different audiences. Mostly included for completeness — if you’re a homelab person, the first three are more relevant to you.

AI Agent Coloring Book on Gumroad

Why I’m telling homelab people specifically

Most blog automation products are sold to marketers who will never touch a YAML file or a systemd unit. If you already run n8n in a docker container on a mini PC, you’re going to find the barrier to using these products approximately zero. The workflows import as JSON, the prompts are text files, the credentials go in n8n’s credential store.

If you’d rather build your own from scratch — honestly, go for it. You’ll learn more. But if you’d rather skip 40 hours of trial-and-error and start from a working setup, here’s the shortcut.

Questions or feedback welcome — I read every DM and respond within a day. The next version of whichever product gets real traction will be shaped by actual buyers, not my guesses.

— Auburn AI

— Auburn AI editorial, Calgary AB

What the sales pitch won’t tell you

I’ve been running some version of this stack for four years now. The moment you move from homelab to “product you’re selling,” a few hard truths surface that affiliate posts skip over.

First: your customers will have networking setups you didn’t anticipate. I spent three weeks debugging a client’s Docker networking issue that only happened because they had a Ubiquiti Dream Machine with specific VLAN rules. The stack works great in a standard home environment. It breaks at the edges. Budget for that support burden or build in better documentation than you think you need.

Second, the hardware cost floor is real. You can run this on a single NUC or used laptop for hobbyist work. The moment someone depends on it for their business—even part-time—you’re recommending redundancy, backup power, proper thermals. That’s two or three times the initial cost. Many customers balk. Some should, because the ROI doesn’t justify it yet for them.

Third: licensing and compliance creep in. If you’re self-hosting LLMs or automation tools for others, you need to understand what you’re actually licensing. I had to walk back an initial offering because one library had restrictions I’d missed. Small problem at homelab scale. Real one at product scale.

  • Start with documentation that assumes the buyer is less technical than you. Then assume they’re less technical than that.
  • Keep a separate “production checklist” that’s honest about monitoring, backup, and failure modes.
  • Price your support time explicitly, even if you bundle it initially. You’ll know faster whether the business works.

The stack itself is solid. The business case depends entirely on whether you can support it at the margin you’re charging.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top