The €0 AI Stack Playbook
The €200 stack was never one bill — it's death by €20 a month.
ChatGPT Plus (~$20/mo), Midjourney (~$10–30/mo), the automation lane on top — each subscription "starts at $19" until the list crosses €50, €100, and the €200/month the title quotes. None of it is yours, and your client data lives in their cloud while you rent the capability.
The playbook is the migration: 8 full replacements with honest setup times, real price math, and a 3-year cost trace that shows exactly when owning beats renting on your hardware.
Stop renting AI. Own it.
Every AI subscription is a lease on capability you could own. This playbook shows the exact migration: what to pull, what to install, how long it really takes, and what it actually costs versus the cloud.
What's inside
📄 The Playbook (22-page PDF)
- The Math: honest 3-year TCO — cloud rent vs. own-it-once (used RTX 3090 ~€800, ~€6,200 saved over 3 years)
- 8 Replacement Recipes: ChatGPT->Ollama, Midjourney->ComfyUI+FLUX, Notion->AnythingLLM, Zapier->n8n, and 4 more, each with the exact stack + setup step
- Hardware Tiers: what runs on 8GB / 16GB / 24GB VRAM (honest — no overselling)
- Migration SOP: week-by-week, no business disruption
- Troubleshooting: Ollama truncation, quantization, Windows vs WSL, ComfyUI conflicts
📊 The Calculator (XLSX, 5 tabs) — list YOUR subscriptions, see your real break-even date. Includes used-GPU price table.
⚙️ A Working Modelfile — operator-writer, a tuned 8B writer. Three commands and you own it: ollama pull qwen3:8b -> ollama create operator-writer -f Modelfile -> ollama run operator-writer.
🖨️ Cheatsheet (2-page PDF) — printable quick-ref.
€29 once, not a €50-plus month
One payment, no monthly anything. The named stack this replaces — ChatGPT Plus (~$20/mo), Midjourney (~$10–30/mo), Zapier Professional ($19.99/mo) — clears €50 a month on its own, and the full line the playbook maps reaches the €200/month in its title. €29 once; the stack then runs on hardware you already own at €0/month.
Your data stays on your machine
Every recipe is local — Ollama + Open WebUI, ComfyUI + FLUX, self-hosted n8n, and local RAG. Client work and conversations stay on your box, not on a vendor cloud.
One-time price · no subscription · editable · unwatermarked
€29 once. The playbook, the calculator and the Modelfile are plain files you own. Nothing is paywalled, nothing is watermarked.
Who this is for
Freelancers and small agencies spending €80-plus a month on AI subscriptions who want to cut costs without cutting capability. You'll get most out of it if you already own (or will buy a used) GPU.
Who it's NOT for
People without a GPU or the willingness to learn basic self-hosting; those who need 100% uptime SLAs.
Why trust this?
We run Operator Co on exactly this: Ollama on an RTX 3090, Open WebUI, ComfyUI + FLUX, self-hosted n8n, local RAG. Monthly AI bill: €0. Prices verified July 2026. No affiliate links. Where cloud wins, we say so.