Windows 11 WSL2 Hermes Agent Setup — Linux AI Agent on Windows¶

Q: How do I prevent WSL2 from shutting down and stopping Hermes Agent?

Create a Windows Task Scheduler task that runs 'wsl.exe -d Ubuntu-24.04 -e sleep infinity' at system startup, or ensure the Hermes Agent systemd service has 'Restart=always' configured.

Run Hermes Agent on Windows 11 with full Linux compatibility via WSL2 — no dual-booting required. Get NVIDIA GPU acceleration for local models through CUDA passthrough, native Linux filesystem performance, and all Hermes Agent features. This Windows WSL2 setup guide covers everything from installation to persistent 24/7 operation.

Overview¶

WSL2 provides a real Linux kernel inside Windows, enabling Hermes Agent to run exactly as it would on a native Linux machine. GPU passthrough means your NVIDIA card accelerates Ollama models at full CUDA speed. The setup is free if you already own a Windows PC, making it the easiest entry point for Windows users who want the full Hermes Agent setup experience.

How It Works¶

Feature	WSL2 Benefit
Linux kernel	Real Linux, not emulation
GPU passthrough	NVIDIA CUDA works inside WSL2
Filesystem	ext4 inside, accessible from Windows via `\\wsl$`
Networking	Shared with Windows — no VM networking hassle
Systemd	Supported — run Hermes Agent as a service

Step-by-Step Installation¶

Step 1: Install WSL2¶

Open PowerShell as Administrator:

wsl --install -d Ubuntu-24.04
# Set up Linux username/password after reboot
wsl --list --verbose

Step 2: Configure WSL2 Resources¶

Create %USERPROFILE%\.wslconfig:

[wsl2]
memory=8GB
processors=4
swap=4GB
networkingMode=mirrored

Apply: wsl --shutdown in PowerShell, then reopen WSL.

Step 3: GPU Passthrough (NVIDIA)¶

# In PowerShell — install NVIDIA WSL2 driver
winget install --id=Nvidia.CUDA -e

Inside WSL:

nvidia-smi  # Should show your GPU
sudo apt install nvidia-utils-550  # If not found

Step 4: Install Dependencies¶

sudo apt update
sudo apt install -y python3 python3-pip python3-venv nodejs npm git curl ffmpeg
curl -fsSL https://deb.nodesource.com/setup_20.x | sudo -E bash -
sudo apt install -y nodejs

Step 5: Install Hermes Agent¶

pip install hermes-agent
hermes profile create wsl-agent
hermes profile use wsl-agent
hermes config set providers.openrouter.api_key ***
hermes config set model.default openrouter/anthropic/claude-sonnet-4

Step 6: Local Models with Ollama¶

curl -fsSL https://ollama.com/install.sh | sh
ollama pull llama3.2
ollama pull nomic-embed-text
ollama pull qwen2.5:14b  # If 12GB+ VRAM
ollama run llama3.2 --verbose | grep -i cuda  # Verify GPU

Step 7: Filesystem Strategy¶

Location	Speed	Best For
`/home/` (Linux ext4)	Fast	Hermes config, skills, crons
`/mnt/c/` (Windows NTFS)	Slower	Shared files, backups

Rule: Keep active Hermes data in /home/ for performance. For model selection and memory, WSL2's ext4 filesystem is significantly faster than NTFS.

Step 8: Systemd Service¶

sudo tee /etc/systemd/system/hermes-gateway.service << 'EOF'
[Unit]
Description=Hermes Agent Gateway
After=network-online.target

[Service]
Type=simple
User=your-username
ExecStart=/home/your-username/.local/bin/hermes gateway run
Restart=always
RestartSec=10
Environment=HOME=/home/your-username

[Install]
WantedBy=multi-user.target
EOF

sudo systemctl daemon-reload
sudo systemctl enable --now hermes-gateway

Step 9: Keep WSL2 Running¶

WSL2 shuts down when idle. Prevent this:

# Task Scheduler → Create Task → Trigger: At startup
# Action: wsl.exe -d Ubuntu-24.04 -e sleep infinity

Step 10: Browser Automation¶

pip install playwright
playwright install chromium
playwright install-deps chromium

For headed browsing, install a Windows X server like VcXsrv and export DISPLAY=:0.

Benefits¶

Free: No additional hardware needed beyond your Windows PC
Full GPU acceleration: NVIDIA CUDA works natively inside WSL2
No dual-boot: Run Linux Hermes Agent alongside Windows apps
Local models: Ollama at full speed via GPU passthrough
Persistent: Systemd service with auto-restart keeps Hermes alive

FAQ¶

Does Ollama GPU acceleration work in WSL2?¶

Yes. NVIDIA's WSL2 driver enables full CUDA passthrough. Ollama detects the GPU automatically and uses it for inference — same speed as native Linux.

Why not just use Windows native Hermes Agent?¶

WSL2 provides a real Linux environment with systemd, ext4 filesystem, and native package management. Many MCP servers and developer tools expect Linux. WSL2 delivers this without dual-booting.

How do I prevent WSL2 from shutting down?¶

Create a Windows Task Scheduler task that runs wsl.exe -d Ubuntu-24.04 -e sleep infinity at startup, or ensure the Hermes systemd service has Restart=always.

Hermes Agent Setup Overview — All platform options
Gaming PC Setup — Native Linux for max GPU performance
Docker Setup — Alternative Windows deployment
Model Selection Guide — GPU model sizing
Troubleshooting Guide — WSL2-specific issues

Next: Gaming PC Setup · Docker Setup · Troubleshooting