astafford8488/agentaegis-mcp

What it does

AgentAegis is a security and trust layer designed for AI agents and MCP (Model Context Protocol) environments. It provides pre-installation scanning capabilities for MCP servers and AI skills, identifying potential security threats before they are integrated into your system. The tool flags critical security issues including data exfiltration attempts, prompt injection sinks, dangerous capabilities, install hooks, and code obfuscation patterns.

How to set up

AgentAegis is available as an MCP server that can be integrated into Claude, ChatGPT, and other AI platforms. Installation involves adding the MCP server configuration to your environment and enabling the scanning capabilities. Once configured, you can run scans on any MCP plugin or skill before installation using the dedicated scanning functions.

Key features

• Scan MCP plugins and skills for security vulnerabilities
• Detect exfiltration attempts and data leakage risks
• Identify prompt injection attack vectors
• Flag dangerous capabilities and permissions
• Detect install hooks and malicious code patterns
• Identify obfuscated or suspicious code
• Provide PROCEED/CAUTION/BLOCK recommendations
• Compatible with multiple AI platforms and MCP servers