Turn Any USB Drive Into a Portable AI Agent — Hermes AI

💡 The Concept: Portable Hermes Agent 0:00

The Hermes desktop agent can run entirely from a USB drive — 100% locally with no cloud AI, no API key, and no monthly subscription. Everything stays on the drive: code, files, prompts, and the AI model itself.

Works on Linux, Windows, and Mac — same drive, same setup, all three platforms
Also works from an external SSD or internal SSD for better performance
Supports both local and online models — use local for full privacy, online if hardware is limited
No installation required on the host machine — truly portable

          Key idea: One portable drive holds your complete AI agent environment — plug it into any computer and start working with Hermes immediately.
        

🪟 Windows Setup: Format & Download 1:27

Start by formatting the USB drive as exFAT — the only filesystem that works natively across Windows, macOS, and Linux without additional drivers.

Download the local-hermes-portable project from GitHub (zip or clone)
Extract to the USB drive setup location
Three launcher files included: windows.bat, mac.sh, linux.sh
Double-click windows.bat — first run downloads required files and checks system

          Why exFAT? It supports large files (needed for AI models), has no 4 GB file-size limit like FAT32, and is read/write compatible across all three operating systems.
        

📦 Download & Load Local Model 2:05

After running the launcher, an interactive menu appears. The setup guides you through downloading and configuring the local AI model.

Press 1, enter Y to open the local setup page
Setup page recommends models based on your system specs
Recommended model: Gemma 3 4B — download directly from browser
Move the model file into the project's models/ folder
Re-run windows.bat — launcher auto-detects the model
Select model, press enter to start the local AI server
Test in browser web app to confirm the model works

          Auto-detection: The launcher scans the models/ folder automatically — no manual path configuration needed. Just drop the file in and go.
        

🛠️ Install Hermes Desktop & CLI 3:14

With the model server running, install the Hermes agent itself. Keep the model server terminal open and start a new instance of the launcher.

Run windows.bat again in a separate terminal
Press 3 to install Hermes agent (first setup: 5–8 minutes)
Hermes auto-detects the local Gemma model — no manual configuration needed
Press 1 for Hermes CLI, press 2 for Hermes Desktop
Both interfaces work with the same local model using the same portable setup

          Two interfaces: Hermes CLI gives you a powerful terminal-based experience, while Hermes Desktop provides a graphical interface — both connect to the same local model and share the same portable environment.
        

Note: The first response may be slower depending on system hardware — subsequent responses are faster as the model warms up.

🍎 macOS Setup 5:01

The same USB drive works on macOS — just use the Mac launcher script instead.

Open terminal in the project folder on the USB drive
Run bash mac.sh
First run downloads required macOS-specific files
Press 1 to start local model server (auto-detects Gemma)
Open a new terminal window, run bash mac.sh again
Select option 3 to install Hermes
Choose CLI or Desktop interface

          No re-download needed: The model file you already placed in models/ on Windows is still there — no need to re-download or reconfigure anything on macOS.
        

🐧 Linux Setup 6:21

Linux follows the exact same pattern — same process, different launcher script.

Open terminal, navigate to the project folder
Run bash linux.sh
Option 1 for model server, option 3 for Hermes
Works without any extra configuration
Same portable setup across all three platforms

          Truly cross-platform: The same folder structure, same model, same scripts — three launcher files are the only platform-specific components.
        

🎯 What You Get 6:46

One USB drive (or SSD) with a complete, portable AI agent setup that works everywhere.

Cross-platform: Windows, macOS, and Linux from the same drive
Both interfaces: Hermes CLI and Hermes Desktop included
Local model: Gemma 3 4B runs offline with full privacy
Online option: Connect to online models if hardware is limited
Remote access: Can connect to Telegram and other platforms for remote interaction
Full privacy: Everything stays on the drive — no data leaves your device

✅ Key Takeaways

Run Hermes AI Agent from a USB drive, external SSD, or internal SSD
100% local: no cloud, no API key, no subscription
Cross-platform: same USB works on Windows, macOS, and Linux
Format drive as exFAT for cross-OS compatibility
Three launcher scripts: windows.bat, mac.sh, linux.sh
Gemma 3 4B recommended for local use
Auto-detects model in models/ folder — no manual config
Both Hermes CLI and Hermes Desktop included
Also supports online models if hardware is limited
Can connect to Telegram for remote interaction
First setup takes 5–8 minutes per OS, then instant