guardfive 0.1.0

Security scanner for MCP (Model Context Protocol) servers

GuardFive 🛡️

The security scanner for AI agent tools (MCP servers)

GuardFive scans the apps/tools that AI agents use to make sure they're safe and not secretly malicious.

What It Does

1. SCAN → Find all MCP servers your AI uses
2. ANALYZE → Check for hidden threats
3. MONITOR → Watch for changes over time
4. ALERT → Tell you when something's wrong
5. REPORT → Give you proof for compliance

Threats We Detect

Threat	Description
🎭 Tool Poisoning	Hidden instructions in tool descriptions
🔄 Rug Pull	Tools that change after you trust them
👯 Shadowing	Malicious tools pretending to be legitimate ones
💉 Command Injection	Vulnerable code that hackers can exploit

Quick Start

# Install
pip install -e .

# Scan your MCP config
guardfive scan ~/.cursor/mcp.json

# Or scan a specific server
guardfive scan --server "npx -y @modelcontextprotocol/server-filesystem /"

Installation

# Clone the repo
git clone https://github.com/yourusername/guardfive.git
cd guardfive

# Create virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -e .

Project Structure

guardfive/
├── guardfive/
│   ├── cli.py              # Command line interface
│   ├── scanner.py          # Main scanning logic
│   ├── models.py           # Data structures
│   ├── detectors/          # Threat detection modules
│   │   ├── tool_poisoning.py
│   │   ├── rug_pull.py
│   │   └── shadowing.py
│   └── connectors/
│       └── mcp_client.py   # Connect to MCP servers
└── tests/

License

MIT