Autarch/windows_manual.md

AUTARCH — Windows User Manual

Autonomous Tactical Agent for Reconnaissance, Counterintelligence, and Hacking By darkHal Security Group and Setec Security Labs

---

Table of Contents

1. Windows Overview
2. System Requirements
3. Installation
4. Running AUTARCH on Windows
5. Web Dashboard
6. LLM Setup on Windows
7. Metasploit on Windows
8. Nmap on Windows
9. Hardware & Device Tools
10. WireGuard VPN
11. Known Limitations on Windows
12. Troubleshooting
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.

What works on Windows:

• Web dashboard (full UI)
• AI chat (all LLM backends except GPU quantization)
• OSINT tools
• Android/iPhone device management via ADB (USB or WebUSB)
• Hardware ESP32 flashing
• Metasploit RPC client (MSF must be started separately)
• Reverse shell management
• Targets & Settings
• Agent Hal (autonomous AI agent)

What has reduced functionality on Windows:

• Packet capture (tshark/pyshark) — requires Npcap
• WireGuard — requires Windows WireGuard app
• Linux service manager (--service flag) — use Task Scheduler instead
• Metasploit auto-start — must start MSF manually

---

2. System Requirements

Component	Minimum	Recommended
OS	Windows 10 (64-bit)	Windows 11
RAM	4 GB	16 GB (for local AI models)
Storage	2 GB free	20 GB (for AI models)
Python	3.10	3.11 or 3.12
Browser	Chrome / Edge	Chrome (required for WebUSB)
GPU (AI)	None needed	NVIDIA GPU (for GPU-accelerated models)

---

3. Installation

Step 1 — Install Python

Download Python from python.org. 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
python autarch.py -m chat	Start AI chat directly
python autarch.py --setup	Re-run the setup wizard
python autarch.py --skip-setup	Skip AI setup
python autarch.py --show-config	Show current settings
python autarch.py --mcp stdio	Start MCP server
python autarch.py -l	List all available modules

Running as a Background Service on Windows

AUTARCH's --service commands use Linux systemd. On Windows, use Task Scheduler instead:

1. Open Task Scheduler (taskschd.msc)
2. Create Basic Task → name it "AUTARCH"
3. Trigger: "When the computer starts"
4. Action: Start a program
   • Program: python
   • Arguments: C:\she\autarch\autarch.py --web
   • Start in: C:\she\autarch
5. Run whether user is logged on or not

Alternatively, use NSSM (Non-Sucking Service Manager) for a proper Windows service:

nssm install AUTARCH "python" "C:\she\autarch\autarch.py --web"
nssm start AUTARCH

---

5. Web Dashboard

The dashboard runs on port 8080 by default. Access it at http://localhost:8080.

Login

Default credentials are set during first-run setup. Change them in Settings → Password.

Sidebar Navigation

Section	What's There
Dashboard	System overview, tool status
Targets	Pentest scope and target management
Defense	System hardening, firewall checks
Offense	Metasploit modules, port scanning
Counter	Threat hunting, detection
Analyze	File forensics, malware analysis
OSINT	Intelligence gathering
Simulate	Attack scenarios, Legendary Creator
Wireshark	Packet analysis (needs Npcap)
Hardware	Android/iPhone/ESP32 management
Android Exploit	Android-specific testing
iPhone Exploit	iPhone forensics
Shield	Anti-stalkerware scanner
Reverse Shell	Remote device management
Archon	Android companion app
UPnP	Port forwarding
WireGuard	VPN management
MSF Console	Metasploit terminal
Settings	All configuration

HAL Chat Button

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
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
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 — 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 and get the CUDA version
   • CPU only: pip install torch --index-url https://download.pytorch.org/whl/cpu
2. Install remaining dependencies:

   pip install transformers accelerate
   

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
2. Run the installer — it installs to C:\metasploit-framework by default
3. After install, start the MSF RPC daemon:

   C:\metasploit-framework\bin\msfrpcd.bat -P yourpassword -S -f
   

   Or use msfconsole directly and enable RPC from within it.

Connecting AUTARCH to MSF

1. Go to Settings in AUTARCH
2. Set MSF RPC host: 127.0.0.1, port: 55553
3. Enter your RPC password
4. In the web dashboard, go to MSF Console and click Reconnect

Using MSF in AUTARCH

• MSF Console page (/msf) — terminal-style console, type commands directly
• 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 — 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 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 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 — 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
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.

---

11. Known Limitations on Windows

Feature	Status	Notes
Web dashboard	Full	Works perfectly
AI chat (cloud APIs)	Full	Claude, OpenAI, HuggingFace all work
AI chat (local GGUF)	Full (CPU)	Slow without GPU
GPU quantization (4-bit/8-bit)	Partial	Needs CUDA + bitsandbytes
Nmap scanning	Full	Needs Nmap + Npcap installed
Packet capture	Partial	Needs Npcap + admin rights
Metasploit	Full (manual start)	MSF must be started manually
ADB (server mode)	Full	Bundled ADB binary works
ADB (WebUSB/Direct)	Full	Chrome/Edge only, needs WinUSB driver
ESP32 flashing	Full	COM port instead of /dev/ttyUSB
WireGuard	Partial	Needs Windows WireGuard app
Background service	Via Task Scheduler	--service flag doesn't work
System uptime	N/A	Shows "N/A" (uses /proc/uptime)
mDNS discovery	Partial	May require Bonjour

---

12. Troubleshooting

"Python not found" or command not recognized

Python is not in your PATH. Either:

• Reinstall Python and check "Add to PATH"
• Or run: py autarch.py instead of python autarch.py

Web dashboard won't start — "Port already in use"

Another process is on port 8080. Use a different port:

python autarch.py --web --port 8090

Or find and kill the conflicting process:

netstat -ano | findstr :8080
taskkill /PID <PID> /F

bitsandbytes install error

ERROR: Could not find a version that satisfies the requirement bitsandbytes

This is normal on Windows without CUDA. Either:

• Install with pip install bitsandbytes --prefer-binary for a best-effort install
• Or skip it — AUTARCH detects absence and disables quantization automatically

LLM won't load — "No module named llama_cpp"

Install llama-cpp-python:

pip install llama-cpp-python

If you have an NVIDIA GPU and want GPU acceleration:

set CMAKE_ARGS="-DLLAMA_CUBLAS=on"
pip install llama-cpp-python --force-reinstall --no-cache-dir

ADB device not detected

1. Enable USB Debugging on your phone (Settings → Developer Options → USB Debugging)
2. When prompted on the phone, tap "Allow"
3. Check if Windows recognizes the device: android\adb.exe devices
4. Install the correct USB driver for your phone manufacturer

Nmap not found

AUTARCH reports "nmap not found" in the Dashboard. Fix it:

1. Install Nmap from nmap.org
2. Add C:\Program Files (x86)\Nmap to your Windows PATH
3. Or configure the path in AUTARCH Settings → Tool Paths

Metasploit can't connect

1. Verify MSF RPC daemon is running: netstat -ano | findstr :55553
2. If not running, start it: msfrpcd -P yourpassword -S -f
3. Check password matches what's in AUTARCH Settings
4. Try clicking Reconnect in the MSF Console page

Firewall blocking the dashboard

Windows Firewall may block port 8080. Allow it:

1. Windows Defender Firewall → Advanced Settings
2. Inbound Rules → New Rule
3. Port → TCP → 8080 → Allow

Or from Command Prompt (as Administrator):

netsh advfirewall firewall add rule name="AUTARCH" dir=in action=allow protocol=TCP localport=8080

"Permission denied" errors

Run Command Prompt as Administrator. Right-click Command Prompt → Run as Administrator.

---

13. Quick Reference

Startup Commands

# Start menu
python autarch.py

# Start web dashboard
python autarch.py --web

# Different port
python autarch.py --web --port 9090

# List all modules
python autarch.py -l

# Run AI chat
python autarch.py -m chat

# Reset configuration
python autarch.py --setup

Key URLs

URL	What It Is
http://localhost:8080	Main web dashboard
http://localhost:8080/targets	Target management
http://localhost:8080/settings/llm	LLM configuration
http://localhost:8080/msf	MSF Console terminal
http://localhost:8080/manual	Full user manual

Important Paths

Path	What It Contains
autarch_settings.conf	All configuration
data/targets.json	Saved targets
data/sessions/	Saved sessions
data/dossiers/	OSINT dossiers
android/adb.exe	Bundled ADB binary
tools/	Bundled tools

Common Tool Locations (Windows Defaults)

Tool	Default Path
Nmap	C:\Program Files (x86)\Nmap\nmap.exe
Metasploit	C:\metasploit-framework\bin\
WireGuard	C:\Program Files\WireGuard\
Wireshark	C:\Program Files\Wireshark\
Python	C:\Python311\ or C:\Users\<you>\AppData\Local\Programs\Python\

---

AUTARCH is for authorized security testing and research only. Always obtain written permission before testing systems you do not own.