- Free guide, no sign-up
- 31 step-by-step guides
- Linux or Mac mini
- Français · English · 中文FR · EN · 中文
A mini PC, a coding agent, and everything I learned making them work together.
On my desk, a small, silent machine runs about thirty sites and apps. This guide shows you, step by step, how to set it up, how to connect it to Claude Code, OpenCode or Codex, and above all how to work well with the agent once it's there. 31 guides, checked on a real machine, in French, English and Chinese.
I've been running at Ulrich's for 213 days. He wrote everything down so you can build me too.
The hybrid setup: local for everyday work, cloud for the big jobs. Your call, guide by guide.
One weekend
to set up the machine
From the USB stick to Linux installed, the agent running and the private network connected. Without ever having used a terminal before.
One command
and the agent sets up the rest
You install Claude Code, OpenCode or Codex first. After that, every step of the guide is done by asking it, and by reviewing its plan.
31 services
run on mine
Everything written here was done, broken and fixed on a real machine. Every guide shows the date it was last checked.
The path: 31 guides in five parts. In order, or straight to where you're at.
About 7 hours of reading · from Easy to Advanced
Part 1 · Before you start, guide by guide
Every guide: level, time, “do this first”, date checked- 1 Why a mini-machine (already read)
- 2 Spotting your needs: what to automate, what to hand to an agent (already read)
- 3 Project ideas (already read)
- 4 Choosing the hardware (already read)
- 5 The machines we tested: which mini-PC for AI? (already read)
- 6 What about a Mac mini? (already read)
- 7 Frontier models on a subscription: Claude, ChatGPT, Grok, Gemini (already read)
- 8 Chinese LLMs: MiMo, Qwen, DeepSeek, GLM, Kimi (already read)
Part 2 · Set up the machine, guide by guide
Every guide: level, time, “do this first”, date checkedPart 3 · Open it to the world, guide by guide
Every guide: level, time, “do this first”, date checkedPart 4 · Agents and local AI, guide by guide
Every guide: level, time, “do this first”, date checked- 19 Setting up your agents (already read)
- 20 Agents in the service of a project (already read)
- 21 Agents running on the mini-PC (already read)
- 22 Hermes & OpenClaw: your in-house agents (already read)
- 23 Running several agents together (already read)
- 24 Choosing and sizing your model (already read)
- 25 Ollama & local models (already read)
- 26 Going hybrid: cloud + local (already read)
Part 5 · Working well, guide by guide
Every guide: level, time, “do this first”, date checkedFraming a project with an LLM
How to write a brief the agent understands, review its plan before saying yes, and slice the work so it doesn't go off the rails. The guide to read even without a machine. 18 min, checked on October 1, 2026.
Read the guideIntermediate · 18 min
What's running on my machine right now
A Strix Halo, 128 GB of unified memory, Ubuntu 26.04, 31 services online. Five projects born on it, with the agent.
Vignette Synced sticky notes, signed binary (external site, new tab)
Kiwee Vegan scanner, iOS & Android (external site, new tab)
Quelle IA AI model rankings (external site, new tab)
Pippin Apple prices, 41 countries (external site, new tab)
Saillant Multiplayer 3D game (external site, new tab) + 26 more services online on the machine