MCP - Shield is a security tool for scanning MCP server vulnerabilities, which can detect security risks such as tool poisoning attacks, data leakage channels, and cross - domain violations.

GUARDRAIL is a security framework specifically designed for the large language model (LLM) application ecosystem, providing comprehensive protection especially for the Model Context Protocol (MCP). It uses a multi - layer architecture, including an information gateway layer, a context verification layer, a request control layer, an execution isolation layer, and an audit monitoring layer, aiming to prevent data leakage, unauthorized access, and resource abuse. This framework emphasizes progressive adoption, allowing developers to gradually enhance security while maintaining high performance and ease of use.

The Maccy Clipboard MCP Server is a service tool that exposes Maccy's clipboard history to AI assistants such as Claude. It supports searching, viewing, and managing clipboard content, including image support and data statistics functions. However, be aware of the risk of sensitive data leakage.

This project demonstrates MCP poisoning attacks against AI workflows, including verification of multiple attack scenarios such as code generation pollution and financial data leakage, for educational research purposes.

CyberMCP is a network security testing server based on the MCP protocol, focusing on detecting security vulnerabilities in backend APIs and providing various security tools and resources such as authentication testing, injection testing, and data leakage detection.

Shrike MCP is an MCP server that provides real - time security scanning for AI agents. It integrates 9 security tools and scans prompts, responses, SQL queries, file writing, CLI commands, and web searches through a multi - stage detection pipeline to prevent prompt injection, jailbreak, PII leakage, and data theft.