安全扫描器与MCP服务器

BugMentor CAI: Universal Security Intelligence & MCP Server

This is a powerful, open-source BugMentor project that containerizes the CAI (Cybersecurity AI) agent with Ollama and the phi model. It provides both a generic security scanner for any project and a fully-featured MCP (Model Context Protocol) server with automated intelligence reporting.

• Developer: Matías J. Magni, CEO @ BugMentor
• Original Framework: Based on the CAI (Cybersecurity AI) framework by Alias Robotics.
• License: MIT

🚀 Key Features

1. Universal Security Scanner (cai_security_scanner.py)

A cross-platform CLI tool that performs deep security audits on any directory.

• Recursive Scan: Identifies sensitive files (.go, .py, .js, .env, etc.).
• AI-Powered: Uses BugMentor AI (Phi-2) in Docker for high-fidelity auditing.
• Instant Reports: Generates a premium, responsive HTML report and opens it automatically in your browser.

2. Automated MCP Intelligence

The MCP server automatically generates a professional HTML report after every interaction.

• cai_text Tool: Integrates directly with Cursor, VS Code, or Gemini Desktop.
• Persistence: Every response is archived as a styled HTML report inside the container.

🛠 Setup Instructions

1. Build the AI Engine

Ensure Docker is running, then build the image:

docker build -t cai-agent:latest .

2. Launch the MCP Server

Launch the container and the MCP service:

Windows (PowerShell):

.\run_mcp_http.ps1

Linux/macOS (Bash):

chmod +x *.sh
./run_mcp_http.sh

3. Configure Your IDE (MCP)

Add the following to your MCP configuration (e.g., mcp_config.json):

{
  "mcpServers": {
    "cai-ollama-phi": {
      "command": "docker",
      "args": [
        "exec",
        "-i",
        "cai-agent-mcp-http",
        "/home/caiuser/cai_env/bin/python",
        "-u",
        "/home/caiuser/cai_mcp_server.py"
      ]
    }
  }
}

📖 Usage Guide

🔍 Running a Security Audit

Scan any directory on your host:

# Windows
python cai_security_scanner.py "C:\path\to\your\project"

# Linux/macOS
python3 cai_security_scanner.py "/path/to/your/project"

The script will audit the code and automatically open the security_report.html when finished.

🧠 Using the MCP Agent

In your IDE (Cursor/VS Code/Gemini Desktop), simply ask:

"Use cai_text to analyze the security of this feature..."

📄 Viewing MCP Reports

To see the beautifully formatted HTML report from your last MCP interaction:

Windows (PowerShell):

.\get_mcp_report.ps1

Linux/macOS (Bash):

./get_mcp_report.sh

Note: A placeholder "Welcome" report is generated automatically on server startup. Send your first cai_text prompt to update it with real intelligence.

🧹 Project Structure

• cai_security_scanner.py: Host-side CLI for directory auditing (Auto-Open feature).
• cai_scanner_core.py: Container-side audit worker.
• cai_mcp_server.py: MCP server with automated reporting logic.
• run_mcp_http.ps1/sh: Container lifecycle management for Windows and Linux.
• get_mcp_report.ps1/sh: Automation scripts for viewing reports.

💰 Funding & Support

BugMentor is dedicated to building privacy-focused, vendor-lock-in-free alternatives for cybersecurity and agile tools.

Building and maintaining enterprise-grade security tools takes significant resources. Your support directly funds server costs, development hours, and the maintenance of our open-source infrastructure.

⚡ Direct Support (Wise)

If you prefer to support the lead developer directly with lower fees (or for one-off contributions), you can use the link below.

Send a Direct Contribution via Wise

💼 Commercial Support & Training

Need help implementing CAI or other BugMentor tools in your company? BugMentor offers:

• Enterprise Installation & Hosting Support
• Cybersecurity & QA Training
• Custom AI Feature Development

Contact us for commercial inquiries.

🧪 Testing & Quality Assurance

This project maintains 100% code and test coverage to ensure reliable security auditing.

Running Tests

The test suite is built with pytest and covers unit, integration, and E2E scenarios.

Windows (PowerShell):

$Env:PYTHONPATH = "."; pytest -v tests/

Linux/macOS:

export PYTHONPATH="."; pytest -v tests/

Testing Levels:

• Unit Tests: Logic validation for cai_scanner_core.py and cai_mcp_server.py.
• Integration Tests: Subprocess and file-system interaction validation for the scanner.
• E2E Smoke Tests: Connectivity checks for the live MCP server.

🤝 Contributing

BugMentor welcomes community contributions. Fork this repository, open PRs, or file issues at BugMentor.

Safe hunting! 🛡️