🚀 Hermes Windows Native

🇨🇳 中文 | English

[]

🖥️ AI Agent Running Natively on Your Windows — No Docker · No WSL2

Hermes Agent + WebUI. Native Windows. Zero overhead.

---

🎯 Why Windows Native?

The official Hermes stack tells you to install Docker or WSL2 on Windows.

That's ~500MB–2GB of memory gone before you even start.

This project removes all that:

	Docker (official)	WSL2 (official)	Hermes Windows Native ⭐
Memory overhead	~500 MB	~1–2 GB	~50 MB ✅
Setup complexity	Medium	High	Low ✅
Virtual machine?	Yes (container)	Yes (Linux)	No ✅
Works on 8GB RAM?	Barely	Struggling	Comfortably ✅
Install time	10–30 min	20–60 min	< 5 min ✅

What you get back

• 🚀 ~250MB RAM saved vs Docker — run other things alongside Hermes
• 💾 ~750MB–1.9GB RAM saved vs WSL2 — use your machine, not a VM inside it
• ⚡ Faster cold start — no container/VM boot delay
• 🔧 Easier debugging — native processes, native tools, native everything

---

⚡ Quick Start — 3 Commands, 3 Minutes

# 1. Clone
git clone https://github.com/markwang2658/hermes-windows-native.git
cd hermes-windows-native

# 2. Install (auto-detects Python, creates venv, installs deps)
.\install.ps1

# 3. Start
.\start.ps1

Then open http://127.0.0.1:8787 in your browser.

🎉 That's it. No Docker. No WSL2. No configuration hell.

Figure: Clone → Install → Start — up and running in under 3 minutes

---

🏗️ Architecture

Figure: Hermes automatically routes input to the best AI model based on type (chat/vision/audio/text)

---

✨ Features

🧠 AI Agent Core (from NousResearch/hermes-agent)

• Persistent Memory — remembers conversations across sessions, builds context over time
• Cron Jobs — scheduled automations while you're offline
• Multi-Provider LLMs — OpenAI, Anthropic, Google Gemini, DeepSeek, Kimi, Ollama (local), Groq...
• Self-Improving Skills — creates, tests, and refines its own tools automatically
• Tool Ecosystem — browser automation, terminal access, file editing, web search, code execution...

Figure: Hermes Agent terminal startup — auto-loading skills, connecting to model gateway, ready to serve

🌐 Web UI (from nesquena/hermes-webui)

• Full Chat Interface — conversation history, streaming responses, markdown rendering
• Workspace Browser — explore files, edit code, git integration built-in
• Session Management — multiple sessions, search, export/import
• 7 Color Themes — Dark, Light, Nord, Monokai, OLED, Solarized...
• Settings Panel — model selection, provider config, theme switcher

Figure: Hermes WebUI three-panel layout — session list, chat area, workspace browser

🖥️ Windows Native Advantages (this project)

• One-Click Install — install.ps1 handles everything (no manual venv, no pip fighting)
• One-Click Start — start.ps1 auto-detects agent, sets env vars, launches server
• Native Performance — no virtualization layer between you and the AI
• Low Resource Usage — ~330MB total on 8GB machine (vs ~1080MB with WSL2)
• PowerShell Native — all scripts in .ps1, no bash/shell hacks

Figure: Hermes Agent live chat — streaming responses, markdown rendering, multi-turn context

---

📁 Project Structure

hermes-windows-native/
├── hermes-agent/          # AI agent core (memory, skills, cron, gateway)
├── hermes-webui/          # Browser UI (chat, workspace, sessions, themes)
├── install.ps1            # One-click installer
├── start.ps1              # One-click launcher
├── README.md              # This file
└── LICENSE                # MIT License

---

💾 Memory Comparison

Running on a machine with 8GB RAM:

Component	Docker	WSL2	Native (this)
Base overhead	~300 MB	~800 MB	~50 MB
Agent process	~200 MB	~200 MB	~200 MB
WebUI server	~80 MB	~80 MB	~80 MB
Total	~580 MB	~1080 MB	~330 MB

You save ~250MB vs Docker, ~750MB vs WSL2. On a 8GB machine, that matters.

---

🤖 Local Model Inference

Figure: LM Studio interface — search, download, and run local AI models (GLM-4.6V for vision/text)

Privacy-first: Images, audio, and files are processed locally on your machine. Data never leaves.

---

🔧 Requirements

• Windows 10 (1809+) or Windows 11
• Python 3.10+ (download)
  • During installation: ✅ Check "Add Python to PATH"
• ~500MB free disk space (for dependencies)

---

🛠️ Troubleshooting

Problem	Fix
python: command not found	Install Python 3.10+, check "Add to PATH"
Port 8787 already in use	.\start.ps1 -Port 8788
Module import errors	Re-run .\install.ps1
Agent not detected	Set $env:HERMES_WEBUI_AGENT_DIR manually in start.ps1
Push fails with timeout	Check your network proxy settings, or try again later

---

🗺️ Roadmap

• Initial release: unified monorepo + PowerShell scripts
• Windows adapter fixes — POSIX paths → Windows paths, fcntl locks, Unix sockets
• Bootstrap bypass — skip WSL detection in bootstrap.py, go straight to native
• CI/CD — GitHub Actions on Windows runner, auto-test on push
• Release builds — portable zip bundle, one-click download & run
• Chinese README (README.zh-CN.md) for Chinese users

---

🤝 Contributing

Found a bug? Have an idea?

• 🐛 Open an Issue — bug reports, feature requests
• 💻 Pull Requests — code contributions welcome!
• 💬 Discussions — questions, ideas, chat

See CONTRIBUTING.md for guidelines.

---

📜 License

MIT License — see LICENSE.

Original work by:

• NousResearch — hermes-agent
• nesquena — hermes-webui

This is a Windows-native fork/adaptation of their work, distributed under the same MIT terms.