Hermes Install & Usage Guide (2026): Windows One-Click Hermes Agent Setup, Model Connection, and First-Run Pitfalls
TL;DR: Installing Hermes Agent on Windows is three steps. Open PowerShell as administrator and run a one-line install script; close the window, open a new one, and type
hermesto enter the configuration wizard; in the wizard, pick a model provider, paste your API Key, and skip the messaging-platform step. After that you can talk to it and invoke tools straight from the terminal. It takes about 10 minutes if things go smoothly. The place people get stuck most often is not reopening the terminal after installing — the newly written environment variables can't be read, and the symptom is "model not recognized." You don't need to uninstall and reinstall: reopen the terminal, or use/modelto specify the model manually.
Hermes Agent is Nous Research's open-source AI agent. Below is the complete process for getting it installed on Windows, connected to a model, and into daily conversation. For what Hermes is and which model versions exist, see the Hermes tutorials hub; if it's already installed and you want it in WeChat, go straight to the Hermes + WeChat guide.
📚 Table of Contents (click to jump)
- Prerequisites: What You Need Before Installing Hermes Agent on Windows
- Step 1: Run the One-Click Install Script as Administrator
- Step 2: Run the Terminal Configuration Wizard and Connect a Model (Minimax as the Example)
- Step 3: Verify the Install and Start the Hermes Terminal
- Step 4: Your First Conversation and Common Terminal Commands
- Fixing "Model Empty / Not Recognized"
- Hermes Agent on Windows: Pros and Cons
- Who It's For
- FAQ
- Related Guides
- Summary
Prerequisites: What You Need Before Installing Hermes Agent on Windows
Miss any one of these and you'll get stuck later:
- A Windows computer. Windows 10 or 11 both work; leaving a few GB of disk space for dependencies like Python and Node.js is recommended.
- A network that can reach GitHub. The install script and everything it installs are pulled from GitHub. If the network can't reach it, the install can't proceed — this is the most common cause of a stuck install.
- A model API Key. Hermes Agent doesn't ship with a model; it has to be connected to a model service to do anything. If you don't have one yet, register on the relevant platform and create an API Key first — you'll need it in step 2. This guide uses Minimax, a common choice in China, as the example.
All commands below run in PowerShell.
Step 1: Run the One-Click Install Script as Administrator
The official one-click script detects and installs dependencies like Python, Node.js, Git, and ripgrep automatically, so you don't install them one by one yourself.
- Open PowerShell. Right-click the icon and choose "Run as administrator." Without administrator privileges you're likely to hit a permission error partway through.
- Copy and run this command:
irm https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.ps1 | iex
The script detects your environment and downloads and installs the core dependencies, which takes a few minutes depending on your network. There are no choices to make during this — just let it finish.
Once it's done, close the current PowerShell window and open a new one. The environment variables the install script writes only apply to new windows; continuing in the old window will likely fail to find the
hermescommand.
Step 2: Run the Terminal Configuration Wizard and Connect a Model (Minimax as the Example)
In the new PowerShell window, type:
hermesThe first run enters the terminal configuration wizard. Use the arrow keys to select the first item, Quick setup, and press Enter.
If the wizard doesn't appear at this point, the install didn't finish or the environment variables still haven't taken effect. Go back to step 1 and run the install script again — it's safe to run repeatedly.
Configure the Model Provider (Minimax as the Example)
Hermes Agent doesn't ship with a model, so this screen is where you bind it to a model service. Which one you pick depends on which platform's API Key you have. This guide uses Minimax, and switching to another provider works exactly the same way.
- In the provider list the terminal prints, select the platform that matches the Key you have.
- Follow the on-screen prompts to paste the API Key and press Enter to confirm.
The terminal won't display the Key as you paste it — that's normal. Just paste and press Enter. You only need to enter this Key once, so copy it before you start.
Messaging Platform Configuration (WeChat / Feishu Gateway Notes)
Hermes Agent can connect its conversation entry point to third-party messaging apps, but this screen of the wizard has no direct option for Chinese platforms like Feishu, so just choose Skip.
Skipping doesn't affect anything — once installed you can talk to Hermes directly in the terminal. If you later want to connect the agent to WeChat or Feishu, run hermes gateway setup separately to configure the gateway; the full steps are in the Hermes + WeChat guide.
For the remaining confirmation prompts, just answer Y (Yes) and press Enter.
Step 3: Verify the Install and Start the Hermes Terminal
When this welcome screen appears in the terminal, the main install and basic configuration are complete:
Type /model in the input line and press Enter, then check whether the model name the terminal echoes back is the provider you just configured. If it echoes correctly, you're ready to go.
Step 4: Your First Conversation and Common Terminal Commands
Once configured, you can interact with and automate through Hermes Agent directly from the command line.
1. Try your first question and agent tool call
At the prompt at the bottom, type your task or question directly, for example:
Write me a Python script that scans every .txt file in the current directory and counts the total number of wordsHermes plans and generates code using the bound model, and once you authorize it, calls the command-line and file tools in your environment to carry out the task.
2. Common terminal slash commands
In daily use, these core commands manage your session state:
| Command | What it does | When you'd use it |
|---|---|---|
/model | View the current model or switch models manually | Checking whether the model is mounted, switching provider or model version |
/help | List every supported command and feature | Forgetting syntax or checking what's new |
/clear | Clear the current session's context memory | Starting a fresh task without old conversation eating tokens and attention |
/tools | View or manage the system tools available to the agent | Checking whether Hermes has file, web search, or terminal permissions |
/exit or Ctrl+C | Exit the Hermes Agent terminal session | Ending the session and returning to a plain PowerShell prompt |
Fixing "Model Empty / Not Recognized"
After finishing the wizard and reopening Hermes, sometimes the status bar or terminal reports an empty model, or throws an error outright:
It looks alarming, but you don't need to uninstall and reinstall.
Step one: close the terminal, reopen it, and type hermes again.
The install script writes environment variables into the system, but the window you were in was opened before the install finished, so it can't read the new variables. Reopening the window resolves most cases.
Step two: if that still doesn't work, use /model to specify the model once manually.
Type /model in the Hermes dialog or input line, then select or enter the model name as prompted.
When the terminal correctly echoes the model name you configured, the model is mounted. The name you enter here should follow the candidates the terminal lists. Different providers and tiers offer different models — don't just copy a name out of someone else's guide.
Hermes Agent on Windows: Pros and Cons
Advantages of this install approach
- Python, Node.js, Git, and ripgrep are all detected and installed by the script, so there's no manual environment setup
- Only two commands to type by hand; everything else is selecting from prompts
- Model and messaging platform settings live in the wizard, so you can rerun it to change them anytime
Things to note
- The script and its dependencies are all pulled from GitHub, so an unreachable network stops the install
- Not reopening the window after installing means the environment variables aren't read, producing "model not recognized"
- The wizard has no option for Chinese platforms like Feishu yet, so those are configured separately afterward
Who It's For
- First-time Hermes Agent users on Windows who want to get running fast with the official one-click script
- Users who already installed it but are stuck on "model not recognized"
- Users who already have a Minimax or other platform API Key and want the agent working soon
FAQ
The configuration wizard didn't appear automatically after installing. What now?
I configured the model but the terminal says it's empty or unrecognized. Do I need to uninstall and reinstall?
The install script is stuck downloading dependencies.
Do I need to configure the messaging platform in the wizard now?
Do I have to use Minimax?
Roughly how long does the whole install take?
Related Guides
- Hermes + WeChat guide: configure the Gateway for a WeChat group AI assistant — once the terminal works, move it into your daily chat
- Hermes tutorials hub — every Nous Research Hermes configuration and advanced guide on the site
Summary
Installing Hermes Agent on Windows is two commands: one install script and one hermes launch. Run the install script as administrator, and close the window and reopen it when it finishes — skip either step and the symptom is "model not recognized." If you do hit it, reopen the terminal first, then use /model to specify the model manually once.