Let’s get ready
Pick where you’ll run Hermes (your laptop is fine). Sign up for one free AI account — we’ll link the exact page. Copy the long string the site calls an “API key.” Treat it like a password.
Your machine
This course assumes you install Hermes on a machine that exists for the agent — a spare PC, mini-PC, old laptop wiped for this purpose, or similar. Not the computer that holds banking, work SSO, password vaults, and family photos.
Why: once tools are on, the agent can act like a capable operator with a shell on that OS. Treat the host as something you can wipe and rebuild without wrecking daily life.
- Windows, macOS, or Linux Desktop as offered for your build
- Hermes home is a folder on that host — everything the agent keeps is plain files
- Next lesson (03) installs Desktop on this box
Default host for this course?
Pick one — the answer is revealed straight away.
Using free cloud models means…
Pick one — the answer is revealed straight away.
Where the brain runs (day one)
Hermes on your dedicated box is the cockpit, but the brain runs elsewhere on day one. Pick one free route — OpenRouter ↗ or OpenCode ↗ — and create the account now, so install in lesson 03 is copy-paste instead of a detour.
- Create a free account on either OpenRouter or OpenCode. One is enough — you can add the other later.
- Generate an API key in that account. OpenRouter hands you one during signup; on OpenCode, open API Keys and create one (screenshot below).
- Copy it somewhere you can paste from in lesson 03 — a password manager, not a sticky note or a chat window.
- Your prompts, replies, and often tool context leave the host on these routes. Useful for learning — not private.
- Do not paste secrets, customer data, or keys into cloud chats “just to try”
You are ready for install when
Next: lesson 03 — install Hermes on that host and get one chat reply.
Other ways to deploy
Curated alternatives
Day one is Hermes Desktop on a spare machine with a free cloud model, and that is enough to finish the course. Everything below is for when you want a different shape — and each one needs setup work this course does not walk through.
| If you want… | We point you to | Extra steps this course does not cover |
|---|---|---|
| The agent running locally | Open weights via Hugging Face, run on a GPU box you own | GPU sizing, downloading weights, running an inference server |
| Docker on a personal PC | Official Hermes Docker / terminal backends — never a bare host shell | Container setup, volume mapping, which mounted folders to back up |
| A VPS, alive 24/7 | Infomaniak (Swiss jurisdiction) for the host, Nous Portal for models | Renting and hardening a server, SSH keys, firewall, jurisdiction choice |
| Anonymity or privacy | Venice.ai as a provider, or local weights if you can run them | Provider policy, payment and identity hygiene, what still leaks |