11. [Known Limitations on Windows](#11-known-limitations-on-windows)
12. [Troubleshooting](#12-troubleshooting)
13. [Quick Reference](#13-quick-reference)
---
## 1. Windows Overview
AUTARCH runs on Windows with most features fully functional. A few Linux-specific tools (like `tshark` packet capture and WireGuard kernel integration) have limited support, but the web dashboard, AI chat, OSINT tools, hardware management, and Metasploit all work on Windows.
Download Python from [python.org](https://www.python.org/downloads/). During installation:
- **Check "Add Python to PATH"** (critical — do this before clicking Install Now)
- Check "Install pip"
- Use the default installation path
Verify the install by opening Command Prompt and typing:
```
python --version
pip --version
```
Both should print version numbers without errors.
### Step 2 — Get AUTARCH
If you received AUTARCH as a ZIP file, extract it to a folder like `C:\she\autarch`.
### Step 3 — Install Dependencies
Open Command Prompt, navigate to your AUTARCH folder, and run:
```
cd C:\she\autarch
pip install -r requirements.txt
```
This installs Flask, requests, and other core libraries. It may take a few minutes.
**Note on bitsandbytes:** The `requirements.txt` includes `bitsandbytes` for GPU model quantization. This package requires Linux/CUDA to fully work. On Windows without CUDA:
```
pip install bitsandbytes --prefer-binary
```
If it fails, you can skip it — AUTARCH will detect its absence and load models without quantization automatically. No other features are affected.
### Step 4 — First Run
```
python autarch.py
```
A setup wizard will appear asking you to configure an AI backend. If you don't have an AI model yet, choose **Skip Setup** — you can configure it later in Settings → LLM Config.
---
## 4. Running AUTARCH on Windows
### Starting the Terminal Menu
```
python autarch.py
```
Navigate with number keys. Type `0` to go back. Type `99` for Settings.
### Starting the Web Dashboard
```
python autarch.py --web
```
Then open your browser to: `http://localhost:8080`
> **Tip:** Use `http://YOUR-IP:8080` (find your IP with `ipconfig`) to access from other devices on your network.
### Useful Startup Flags
| Command | What It Does |
|---------|-------------|
| `python autarch.py` | Start the interactive menu |
| `python autarch.py --web` | Start the web dashboard |
| `python autarch.py --web --port 9090` | Use a different port |
The **HAL** button in the bottom-right corner opens a persistent AI chat panel. It works on every page and uses whatever LLM backend you have configured.
### Debug Console
The **DBG** button (bottom-right, appears after first action) opens a live debug panel showing system logs. Use it to diagnose errors, see LLM load status, and trace tool activity.
---
## 6. LLM Setup on Windows
AUTARCH supports four AI backends. All work on Windows — but local GPU models have some limitations.
### Option A — Claude API (Easiest, Recommended for Windows)
No local GPU needed. Uses Anthropic's cloud API.
1. Get an API key from [console.anthropic.com](https://console.anthropic.com)
2. In AUTARCH: Settings → LLM Config → Claude tab
3. Paste your API key
4. Select a model (e.g. `claude-sonnet-4-6`)
5. Click **Save & Activate Claude**
6. Click **Load Model** — status dot turns green
### Option B — OpenAI API
Works the same way as Claude. Also supports local LLM servers (Ollama, vLLM, LM Studio).
1. Get an API key from [platform.openai.com](https://platform.openai.com)
2. In AUTARCH: Settings → LLM Config → OpenAI tab
3. Paste your API key and select a model
4. Click **Save & Activate OpenAI** then **Load Model**
**Using with a local server (Ollama, LM Studio):**
- Set Base URL to your local server (e.g. `http://localhost:11434/v1` for Ollama)
- Leave API key blank or use `ollama`
- Set model to the name you pulled (e.g. `llama3`)
### Option C — Local GGUF Model (CPU, No Internet)
Runs on CPU — no GPU required, but slow on large models.
1. Install llama-cpp-python:
```
pip install llama-cpp-python
```
2. Download a GGUF model file (e.g. from [HuggingFace](https://huggingface.co) — look for Q4_K_M quantized files, they're the best balance of speed and quality)
3. In AUTARCH: Settings → LLM Config → Local Model tab
4. Set Models Folder path (e.g. `C:\models`)
5. Select your GGUF file
6. Click **Save & Activate Local**
7. Click **Load Model** — first load takes 10–60 seconds
**Recommended GGUF models for Windows (CPU):**
-`mistral-7b-instruct-v0.3.Q4_K_M.gguf` — good for most tasks
-`phi-3-mini-4k-instruct.Q4_K_M.gguf` — fast, good on low-RAM systems
-`llama-3.2-3b-instruct.Q4_K_M.gguf` — very fast, lightweight
### Option D — HuggingFace Transformers (GPU Recommended)
For NVIDIA GPU users. On Windows without CUDA, models will load on CPU (very slow for large models).
1. Install PyTorch for your platform:
- **With NVIDIA GPU:** Visit [pytorch.org](https://pytorch.org/get-started/locally/) and get the CUDA version
3. Optionally install bitsandbytes for quantization (CUDA required):
```
pip install bitsandbytes --prefer-binary
```
4. In AUTARCH: Settings → LLM Config → Local Model tab → enable "Use HuggingFace Transformers"
5. Enter a model ID (e.g. `microsoft/Phi-3-mini-4k-instruct`)
> **Windows note:** If bitsandbytes is not installed or doesn't work, AUTARCH will automatically disable 4-bit/8-bit quantization and load the model in full precision. You'll see a warning in the debug log — this is normal and expected.
### LLM Load Button
On the LLM Config page, after saving settings, always click **Load Model** to initialize the backend. The status indicator shows:
| Color | Meaning |
|-------|---------|
| Grey | Not loaded |
| Amber | Loading... |
| Green | Ready |
| Red | Error — check Debug Log |
Click **Debug Log** to open the live debug console and see exactly what went wrong.
---
## 7. Metasploit on Windows
Metasploit Framework runs on Windows via the official Windows installer.
### Installing Metasploit
1. Download the Windows installer from [metasploit.com](https://www.metasploit.com/download)
2. Run the installer — it installs to `C:\metasploit-framework` by default
- **Offense → Run Module** — quick-launch SSH scanners, port scanners, OS detection with live output
- **Offense → Agent Hal** — tell the AI to run operations autonomously
> **Note:** AUTARCH cannot auto-start/stop the MSF daemon on Windows (that uses Linux `pgrep`/`pkill`). Start MSF manually before connecting.
---
## 8. Nmap on Windows
Nmap is used by many AUTARCH scanning modules.
### Installing Nmap
1. Download from [nmap.org](https://nmap.org/download.html) — use the Windows installer
2. During install, **also install Npcap** (required for raw socket scanning)
3. Nmap installs to `C:\Program Files (x86)\Nmap` by default
### Configuring Path
If AUTARCH can't find nmap, add it to Settings → Tool Paths, or add `C:\Program Files (x86)\Nmap` to your Windows PATH:
1. Search for "Environment Variables" in Start
2. Edit System Environment Variables → Path
3. Add `C:\Program Files (x86)\Nmap`
---
## 9. Hardware & Device Tools
### ADB / Android Devices
AUTARCH includes bundled ADB binaries in `android/`. No separate install needed.
**USB Device Access:** Windows handles USB permissions automatically for most devices. Enable USB Debugging on your Android phone first (Settings → Developer Options → USB Debugging).
**WebUSB Mode (Direct Connection):**
AUTARCH supports WebUSB for direct ADB access from your browser without a server connection. This requires:
- **Chromium-based browser** (Chrome or Edge) — Firefox does not support WebUSB
- Install the [Android ADB driver](https://developer.android.com/studio/run/win-usb) for your device manufacturer
- Go to Hardware page → click the connection mode toggle → select "Direct (WebUSB)"
> **Note:** WinUSB driver is needed for WebUSB. If your device is recognized by standard ADB but not WebUSB, use [Zadig](https://zadig.akeo.ie/) to install the WinUSB driver.
### ESP32 Flashing
Fully supported on Windows. Connect your ESP32 via USB-serial adapter:
1. Install the CP210x or CH340 USB-serial driver for your adapter
2. Windows will assign it a COM port (e.g. `COM3`)
3. In AUTARCH Hardware page → ESP32 tab → select your COM port
4. Flash or interact normally
### Wireshark / Packet Capture
Requires Npcap (installed with Nmap or Wireshark):
1. Install [Wireshark for Windows](https://www.wireshark.org/download.html) — it includes Npcap
2. After install, `tshark` will be available in `C:\Program Files\Wireshark\`
3. Run AUTARCH as Administrator for raw packet capture permissions
---
## 10. WireGuard VPN
### Installing WireGuard
1. Download from [wireguard.com](https://www.wireguard.com/install/)
2. Install the Windows app
### Using with AUTARCH
AUTARCH's WireGuard page generates and manages config files. On Windows, apply the config manually:
1. Generate your config in AUTARCH → WireGuard
2. Copy the config
3. Open the WireGuard Windows app
4. Click "Add Tunnel" → Import from clipboard or file
5. Click Activate
> **Note:** Automatic WireGuard tunnel management (via `wg` CLI) requires WireGuard to be in your PATH or configured in AUTARCH Settings.
