itsalissonsilva/ModelSafetyMCP

What it does

ModelSafetyMCP is an MCP server designed to enhance the security of machine learning workflows by scanning model artifacts for potential vulnerabilities and threats. It detects unsafe serialization methods, identifies malicious model patterns, flags risky packaging practices, and performs URL-based and directory-level security triage.

How to set up

Install the server as an MCP server in your environment. Configure it to integrate with your Claude or compatible AI assistant. Once deployed, it can scan model artifacts through direct file paths, directories, or URLs provided by users.

Key features

• Unsafe serialization detection using PickleScan
• Malicious model pattern recognition
• Risky packaging identification
• URL-based artifact scanning capabilities
• Directory-level triage and bulk analysis
• ModelScan integration for comprehensive threat detection
• Heuristic-based security analysis