Install Hermes Desktop
Install on the host you chose in 02. If you’re OK with prompts processed on a remote server, enable the free cloud model now. (Switch to local later.)
You need
- The machine from lesson 02 — dedicated hardware by default (Windows, macOS, or Linux Desktop as offered)
- If you took an exception: VPS ready, or personal PC with a Docker isolation plan
- The API key you saved in lesson 02, ready to paste
Get the installer
Open hermes-agent.nousresearch.com ↗ on the host you chose in lesson 02 and take the Desktop app download for that OS. There is a terminal one-liner too, but the graphical installer is the safer route — lesson 10 explains why piping a remote script into a shell is a habit worth not forming.
Install Desktop (on the chosen host)
Later modules assume this cockpit: chat, profiles, tools, approvals, gateway, and cron — on the host you selected, not on a random daily-driver “for now.”
- Confirm install target is the dedicated host from lesson 02 (or your written exception).
- Run the installer and let it finish — it downloads dependencies and configures the machine once.
- Exception only (personal PC path): enable Docker / isolated terminal backends before aggressive tools — or use a VPS instead.
- Launch Hermes when setup completes.
Connect your free model
On first launch Hermes asks for a model provider. Nous Portal is offered as the recommended one-click option — it is a paid subscription, and you do not need it to finish this course. Choose I have an API key and paste the key you saved in lesson 02.
Choosing I have an API key opens the provider list. Find the provider you signed up with in lesson 02 — OpenRouter, or OpenCode Zen if you went that way — and paste your key into its row. Each provider has its own row, so paste into the one you actually created the key for; a key in the wrong row simply will not work. One is enough to start, and you can add more whenever you want.
You can come back to this any time under Settings → Providers → API keys, which is also how you check a key landed: providers with a key show a filled dot and a masked value, the rest still say “Paste … key”.
sk-…); empty ones invite a paste. Only the provider you actually use needs one.With the key in place, pick the model. Type free in the model search and Hermes filters to the routes that cost nothing — OpenCode Zen and OpenRouter both publish several. Take one and get a reply. OpenRouter and OpenCode are not exclusive: add both if you like, and the free models from each show up side by side in this list. More routes means more to fall back on when one is busy.
First contact
Say hello and get one reply — that is the proof. Then open Settings once (the gear, top right) to learn the layout, and continue.
- Providers / models — where your keys live and which model answers
- Gateway / messaging — lesson 05; leave closed for now
Where your files live
Everything Hermes keeps is plain files, in this profile’s Hermes home ↗ on that host — often ~/.hermes, or under your user AppData on Windows. You do not need to touch it today; lesson 04 sends you there to write the personality file.
Optional companions
Nothing here is required. Hermes ships a pet — an animated mascot that floats over the app and reacts to what the agent is doing — under Settings → Appearance. Pick one if you like it; it changes nothing about the harness. IDE-shaped coding agents are optional too: add them only after Hermes works, and only under the same isolation rules as the host.
Other ways to deploy
Other deploy shapes (reference)
If you need a layout beyond dedicated Desktop / VPS / Docker isolation, map it deliberately or ask for help.
- Nous Portal subscription — models + tool gateway under official Portal setup
- Own infra — local/self-hosted models; Hermes on the same machine or a LAN endpoint
- Cloud providers — keys in Providers (OpenRouter, OpenCode, and others listed in docs)
- Multi-machine — split Desktop UI, always-on gateway, and model host across boxes