[Agent 101 #4] How to install Hermes Agent Desktop: your first AI assistant, no terminal

The short version: three posts of theory, now we boot it up

The first three posts covered the ideas — how an assistant differs from a chat box, why you use an existing framework instead of wiring it yourself, and the combo we're using: a ChatGPT brain inside a Hermes body. The concepts are done. This post is where you actually install it and turn it on.

The good news: Hermes ships a desktop app, and the process is easier than you'd expect — about the same as installing any normal program. It handles the annoying part — all the dependencies — for you. You just need to download, double-click, and sign in with ChatGPT.

Follow along and you'll have your first assistant by the end.

Before you start: two things you need

From the previous post, you need two things on hand: your own computer, and a paid ChatGPT account. Grab both and let's go.

The whole thing is four steps: download → install → connect the brain → chat.

Step 1: Download Hermes Desktop

Get the installer from the official download page: hermes-agent.nousresearch.com/desktop (or grab it from the latest GitHub release).

• Mac: download the .dmg
• Windows: download the .exe

Just pick the build for your OS.

Step 2: Install — it sets itself up

Double-click the file you just downloaded:

• Mac: open the .dmg and drag Hermes into your Applications folder
• Windows: run the .exe and click through the installer

Open it and you'll get the Hermes welcome screen. Hit Install Hermes to begin:

Here's the nicest part of the desktop app: it installs all the dependencies for you, step by step — Python, Node.js, Git, ffmpeg, things you may never have heard of — in roughly 11 automatic steps. You don't pre-install anything, and you never touch a terminal. Just watch the progress bar finish.

When it's done you'll see a big HERMES IS READY. Hit Launch Hermes:

It opens into a point-and-click setup wizard that walks you through the rest. No mystery config files to edit.

Step 3: Connect the brain — sign in with ChatGPT

The wizard asks which brain to use (the AI model that thinks for your assistant, from the last post). The screen defaults to recommending Nous Portal (Nous's own all-in-one service), but this series uses the ChatGPT you're already paying for, so don't pick that one:

Expand "Other providers" at the bottom and choose "OpenAI OAuth (ChatGPT)".

The ChatGPT in the name is the tell — this option means "sign in with your ChatGPT account", and it's the officially supported way to use ChatGPT as the brain. (If an older guide you've seen calls it "OpenAI Codex", don't let it throw you — just pick the one that says OAuth / ChatGPT.)

Once you select it:

1. A browser window opens and asks you to log in
2. Sign in with your ChatGPT account and click authorize
3. Done. The credentials are saved to ~/.hermes/auth.json on your own machine — you're not handing your password to anyone

(One heads-up: Nous's docs don't spell out exactly which ChatGPT plans are supported, so if your plan logs in but won't run, that's not a mistake on your end — it's a plan-support thing. Just double-check and move on.)

Once it's connected you'll see OpenAI OAuth (ChatGPT) connected ✓ and the default model showing gpt-5.5 — the ChatGPT you were already paying for is now your assistant's brain.

Step 4: Send your first message

After setup you land on the chat screen, with your chosen model shown up top (usually OpenAI's current flagship, e.g. gpt-5.5). The input box is right there and it feels just like using ChatGPT — except this one uses tools, remembers you, and can later run on its own.

Start with something simple to confirm it replies. Then try a small task that makes it actually use a tool, so you can feel the difference from a plain chat box — for example:

• "Check today's weather in Taipei and tell me whether I should bring an umbrella."
• "Turn this into three bullet points: …"

Send it and watch it actually go look something up / do something before answering. The first time you see it act on its own, that's the moment it clicks.

A few commands you'll want later (type / in the box and they pop up):

• /help: list the commands
• /tools: see which tools it currently has
• /model: swap the brain later

One important tip: get chat solid before adding features

Nous's docs say this plainly, and it's worth repeating:

If basic conversation isn't even working smoothly yet, don't rush to add LINE, scheduling, or voice.

The classic beginner mistake is installing it and immediately wiring up every shiny feature — and then when something breaks, you have no idea where. First confirm you have an assistant that chats reliably and does small things for you. Once that base is solid, adding things on top won't turn into chaos.

Play with it for a few days. Let it look things up, tidy things, remember a few things. Once "having an assistant around" feels normal —

Next: talk to it from your phone

You now have an assistant that talks to you on your computer. The next move is to reach it from your phone — and the easiest way is Telegram. The next post (Part 5) walks you through it: create a bot with one command, grab a key, hand it to Hermes, and you're messaging your assistant from your phone — no public address, no tunnel, because Hermes fetches your messages itself.

---

This series — AI Agent Life, from Zero:

• Part 1: AI assistant vs ChatGPT
• Part 2: What is an agent framework
• Part 3: The ChatGPT-brain + Hermes-body combo
• Part 4: Install Hermes Desktop and talk to your first assistant (this post)
• Part 5: Connect Hermes to Telegram
• Part 6: Let your assistant run on its own
• Part 7: Give your AI assistant eyes and ears