Install Hermes Agent Without Special Networking: A Practical Windows Guide

A step-by-step Windows guide for installing Node.js, Codex, WSL, Ubuntu, Hermes Agent, and finally connecting the Weixin gateway.

This guide is written for people who want a practical Windows path to Hermes Agent without relying on “special networking” tricks. The sequence matters: prepare the basics, install Codex, prepare Ubuntu, install Hermes, then wire the gateway.

1. Prepare the base environment

Before touching Hermes itself, download the offline package and make sure the machine is ready for Node.js, npm, Codex, a Base URL, and an API key.

Download the offline package

2. Install Node.js

Use the official Node.js site, install the Windows package, and verify both node and npm in PowerShell.

node -v
npm -v

3. Install Codex first so troubleshooting is easier

Codex is not required to install Hermes, but it is extremely useful when you hit dependency or command errors later.

Create a token in SuperAIAPI first and keep the Base URL plus API key ready for the later Hermes model setup.

4. Prepare WSL and Ubuntu

This Windows route depends on Ubuntu running under WSL. Enable the virtualization feature, restart, then install Ubuntu and create the Linux account.

dism.exe /online /enable-feature /featurename:VirtualMachinePlatform /all /norestart
wsl --install -d ubuntu

5. Install Hermes Agent itself

Run the Hermes installation script, wait for it to finish, and then open Ubuntu from the Start menu for the real runtime configuration.

6. Configure the Hermes model endpoint

Start Hermes, choose custom endpoint, enter the prepared Base URL and API key, then select gpt-5.4-mini and confirm that the TUI can actually answer.

Base URL: https://superaiapi.com/v1
Model: gpt-5.4-mini

7. Connect the Weixin gateway

Exit the TUI, run gateway setup, choose Weixin, open the link in a browser, scan with WeChat, and allow the required permissions until you receive a reply inside WeChat.

8. Check the real cost

After everything is working, open the usage details and watch the actual spending. This setup is usually manageable for day-to-day usage, especially with gpt-5.4-mini.

Final takeaway

The safest path is: Node.js and npm first, Codex next, WSL and Ubuntu after that, then Hermes model setup, and finally the Weixin gateway. Keeping that order makes failures much easier to diagnose.