
A complete beginner guide for installing Hermes Desktop on Windows, completing first setup, connecting a model, and knowing what to do if the agent is not running yet.
Goal: install Hermes with the official Hermes Desktop installer, finish the first setup, connect one model/provider, and confirm one useful chat.
Hermes Desktop, plus the shared Hermes command-line/runtime files used behind the scenes.
Python via uv, Node, PortableGit, ripgrep, ffmpeg, browser-tool packages, and the local Hermes data folder.
Hermes opens, one model/provider works, and a safe test prompt gets a useful answer.
You do not need to prepare a developer environment first. Just confirm these basics:
| Path | Use when | SAGE9AI note |
|---|---|---|
| Hermes Desktop installer | You want the easiest Windows setup. | Recommended. Start here unless a support person tells you otherwise. |
| Windows native PowerShell installer | You are comfortable with Terminal or Desktop download is unavailable. | Supported fallback. It installs to the same native Windows Hermes area. |
| WSL2 | You specifically need Linux-style development, POSIX terminal behavior, or the dashboard embedded terminal pane. | Supported, but not required for normal Desktop/native Windows use. |
Use the official Hermes Quickstart and let the Desktop app do the Windows setup work for you.
The first time Hermes Desktop runs, it may take longer than a normal app because it is preparing the Hermes engine behind the scenes.
| You may see... | What it means | Beginner answer |
|---|---|---|
| Installing/provisioning dependencies | Hermes is setting up tools it needs, such as Python, Node, PortableGit, ripgrep, ffmpeg, or browser packages. | Let it finish. This can take several minutes on the first run. |
| Windows Defender or security scan delay | Windows is scanning a newly downloaded app and new tool files. | Wait. If blocked, verify you downloaded from the official Hermes site before allowing it. |
| Provider/model setup | Hermes needs one selected provider and model before it can answer. | Choose one connection you recognize, complete its authentication, and wait for setup to confirm it. |
| ChatGPT/Codex device-code page | The browser asks for a 9-character code that is displayed in a Terminal window. | Leave the browser open, use Alt+Tab or the taskbar to find Terminal, then enter the code exactly, including the hyphen. The Terminal window may be behind Hermes or the browser. |
| Tool or gateway setup | Hermes can connect extra services such as messaging, browser, voice, or gateway features. | Skip extras until one basic chat works. |
| Open new terminal / PATH note | The command-line hermes command may need a new terminal window to appear. | For Desktop users, continue in the app. Only open a new terminal if using the CLI fallback. |
Hermes must be connected to a provider and a live model before it can answer. Choose one supported connection you intend to use, complete only its official authentication flow, and confirm a simple test prompt returns an answer.
| Choice | Beginner guidance | Why |
|---|---|---|
| Nous Portal | Select this only if Nous Portal is the connection you intend to authorize. | It is a separate subscription/login path; confirm the billing path before continuing. |
| ChatGPT or Codex subscription | Select the OpenAI/Codex subscription option when that is the access you already use. | Hermes uses a device-code flow: the browser requests the code displayed in Terminal. |
| Other providers | Choose a provider you already understand and can authorize. Follow the exact login, OAuth, or API-key method shown for that provider. | First prove one connection works; add alternatives later. |
| API key | Paste only into the setup prompt/settings field, not into a normal chat. | An API key is a secret and may allow paid usage. |
| Model list | Select a currently offered live model for the provider you authorized. | Model tuning can wait. First goal is one working answer. |
| Multiple providers/fallbacks | Skip for now. | Add backups after your first chat works. |
Hermes can do many things, but a beginner should avoid turning on everything during the first install.
| Feature | Beginner answer | Add later when... |
|---|---|---|
| Messaging gateway | Skip. | You want Hermes on Telegram, Discord, Slack, WhatsApp, Signal, Teams, email, etc. |
| Cron/scheduled jobs | Skip. | You have one repeatable task and know when it should run. |
| MCP servers / plugins | Skip. | You know exactly which external app/tool Hermes should connect to. |
| Voice/TTS/STT | Skip. | Text chat works and you want voice as an add-on. |
| Local models/Ollama/LM Studio | Skip unless already installed. | You intentionally want local AI models and understand the hardware limits. |
Open a new chat in Hermes Desktop and ask a safe, practical prompt:
You are done with first install when Hermes answers normally. Do not add gateway, cron, voice, MCP, local models, or messaging until this basic chat works.
| Topic | What to know |
|---|---|
| Where Hermes lives | Native Windows Hermes data normally lives under %LOCALAPPDATA%\hermes. You do not need to open that folder during beginner setup. |
| Desktop and CLI share setup | After first install, Desktop and the PowerShell-installed hermes CLI share the same native Windows install/data area. |
| WSL2 | WSL2 can coexist with native Windows, but it is a separate Linux environment with separate data. Do not mix paths unless you know which side you are on. |
| Dashboard embedded terminal | Most Hermes features run natively on Windows. The dashboard embedded terminal pane is the main feature that needs WSL2/POSIX behavior. |
| Terminal newline | If you use Hermes CLI later, Ctrl+Enter is the Windows-friendly newline shortcut. Alt+Enter may be captured by Windows Terminal. |
| Symptom | Try this first |
|---|---|
| Desktop installer will not run | Close the installer, re-download from https://hermes-agent.nousresearch.com/, and run the newly downloaded installer. Check any Windows security message carefully. |
| Installer seems stuck | Wait several minutes on first launch, especially while dependencies download. If it fails, write down the exact error text or take a screenshot. |
| Provider/model setup is confusing | Choose one supported provider you intend to use and follow only the authentication method Hermes currently shows for it. Skip extra fallback providers until later. |
| API key question appears | Only paste a key into the setup/settings field. Do not paste API keys into normal chats or support messages. |
| Hermes opens but does not answer | Check that a model/provider is selected and authenticated. Try a simple prompt. If billing/key/login failed, fix that before configuring tools. |
| Old instructions say to install WSL first | Treat those as stale for this beginner Desktop path. Native Windows/Desktop is currently supported; WSL2 is optional for specific Linux-style needs. |
Normally, once your agent is operational, the best move is to ask Hermes to help fix Hermes. But during install, Hermes may not be running yet. Use an accessible fallback LLM.
Copy and paste this template into that LLM. Replace the bracketed parts with your details:
If you contact a person for help, send the error text or screenshot, but redact passwords, API keys, tokens, email verification codes, and private files.
Once your basic agent is up and running, it can install, configure, and set up anything else you would like within the agent harness. No need to guess or research how to do it, because your agent already knows. Just ask it. For safety, tell it to explain the plan before making changes, connecting accounts, or sending messages.
Hermes official docs checked: Installation, Platform Support, and Quickstart. Key facts: Desktop is recommended on Windows; native Windows 10/11 on x86_64/aarch64 is Tier 1; the installer provisions dependencies; one provider/model must be verified; ChatGPT/Codex subscription authentication uses a device code.