PenForge — Security Tool Orchestrator

PenForge is a cybersecurity dashboard and API bridge that connects AI models to penetration testing tools via the Model Context Protocol (MCP). It features a built-in AI Chat interface with real-time tool orchestration, multi-provider support, and dynamic tool discovery.

📸 Screenshots

Dashboard	AI Chat

🚀 Features

• 🤖 AI Chat Dashboard — Chat with an AI agent that can orchestrate security tools in real-time directly from your browser
• 🔌 MCP Protocol — Standard MCP server for external clients (Claude Desktop, 5ire, Cursor, etc.)
• 🔍 Dynamic Tool Discovery — AI auto-discovers available tools from the MCP server; add a tool once, use it everywhere
• 📊 Agent Activity Visibility — See what the AI is doing: Thinking → Orchestrating → Analyzing
• 🧠 Multi-Provider AI — Supports Google Gemini, OpenAI, Anthropic, and Ollama (local)
• 🔐 Client-Side Key Storage — API keys stored in localStorage, never sent to the server
• ⚡ Tool Caching — Cached MCP tool definitions for fast response times

Security Tools

🏗️ Architecture

┌─────────────────────────────────────────────────┐
│                  Web Dashboard                   │
│  ┌──────────┐  ┌──────────┐  ┌───────────────┐  │
│  │ AI Chat  │  │  Tools   │  │ Health Check  │  │
│  └────┬─────┘  └──────────┘  └───────────────┘  │
│       │                                          │
│  ┌────▼─────────────────────────────────────┐    │
│  │       Flask API (kali_server.py)         │    │
│  │  ┌─────────────┐  ┌──────────────────┐   │    │
│  │  │ AI Providers │  │   MCP Client     │   │    │
│  │  │ (Gemini,GPT) │  │ (Tool Discovery) │   │    │
│  │  └─────────────┘  └───────┬──────────┘   │    │
│  └────────────────────────────┼──────────────┘    │
│                               │                   │
│  ┌────────────────────────────▼──────────────┐    │
│  │       MCP Server (mcp_server.py)          │    │
│  │  nmap │ sqlmap │ hydra │ nikto │ ...      │    │
│  └───────────────────────────────────────────┘    │
└─────────────────────────────────────────────────┘

Two ways to use the tools:

1. Dashboard AI Chat — The browser-based chat uses the MCP client internally for fast, cached tool orchestration
2. External MCP Clients — Claude Desktop, 5ire, Cursor, etc. connect directly to the MCP server

🛠️ Installation

Quick Start

git clone https://github.com/prasannashindeit/PenForge-Tool
cd PenForge
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
python3 kali_server.py

Open http://127.0.0.1:5000 in your browser to access the dashboard.

Command Line Options

python3 kali_server.py                              # localhost:5000 (default)
python3 kali_server.py --ip 0.0.0.0                 # all interfaces (⚠️ use caution)
python3 kali_server.py --ip 192.168.1.100 --port 8080  # specific IP + port
python3 kali_server.py --debug                      # verbose logging

🤖 Using the AI Chat

1. Click AI Chat in the sidebar
2. Click the ⚙️ Settings button in the top bar
3. Select your AI Provider (Gemini, OpenAI, Anthropic, or Ollama)
4. Enter your API Key (stored locally, never sent to the server)
5. Choose a Model or type a custom model name
6. Start chatting! Try:
   • "Scan 192.168.1.1 with nmap for open ports"
   • "Check example.com for SQL injection vulnerabilities"
   • "Run a directory brute force on http://target.com"

Discover Tools

Click 🔍 Discover Tools in the chat header to see all available MCP tools in real-time.

🔌 External MCP Client Setup

Local (same machine)

python3 mcp_server.py --server http://127.0.0.1:5000

Remote (via SSH tunnel — recommended)

# Terminal 1 — SSH tunnel
ssh -L 5000:localhost:5000 user@SERVER_IP

# Terminal 2 — MCP client
python3 mcp_server.py --server http://127.0.0.1:5000

Claude Desktop Config

Edit your config file:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

See mcp-kali-server.json for an example configuration.

5ire Desktop

Add an MCP with command: python3 /path/to/mcp_server.py http://SERVER_IP:5000

📁 Project Structure

PenForge/
├── kali_server.py       # Flask API + MCP Client + Dashboard
├── mcp_server.py        # MCP Server (tool definitions)
├── ai_providers.py      # AI provider abstraction (Gemini, OpenAI, Anthropic, Ollama)
├── requirements.txt     # Python dependencies
├── templates/
│   └── index.html       # Dashboard UI
├── static/
│   ├── css/style.css    # Dashboard styles
│   └── js/app.js        # Dashboard logic
└── mcp-kali-server.json # Claude Desktop config example

🔮 Other Possibilities

The AI agent can execute arbitrary commands, enabling tasks beyond the built-in tools:

• Memory forensics with Volatility — process enumeration, DLL injection checks
• Disk forensics with SleuthKit — timeline generation, file carving
• Network analysis with tcpdump/Wireshark — packet capture and analysis
• CTF solving — automated recon and exploitation in real-time

⚠️ Disclaimer

This project is intended solely for educational and ethical testing purposes. Any misuse of the information or tools provided — including unauthorized access, exploitation, or malicious activity — is strictly prohibited.

The author assumes no responsibility for misuse.