# AI security

This project designs and implements an architecture that connects the data space with generative AI. Through the Model Context Protocol (MCP) server as a secure middle layer, the language model can indirectly access the DuckDB database. It has the characteristics of modularity, extensibility, and log tracking, laying the foundation for future upgrades to the RAG architecture.

MCP-Scan is a security scanning tool for MCP servers, used to detect common security vulnerabilities such as prompt injection, tool poisoning, and cross-domain escalation.

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.

An AI-based NPM package analysis MCP server that provides real-time security scanning, dependency analysis, performance evaluation, and other functions. It integrates Claude and Anthropic AI technologies to optimize the management of the npm ecosystem.

This project is a collection of resources related to the security of the Model Context Protocol (MCP), including research papers, security vulnerability analyses, tools, articles, and other relevant resources. It aims to help developers and security experts understand and address security challenges in the MCP protocol.

The Enkrypt AI MCP Server is a tool integrating red team testing, prompt auditing, and AI security analysis, supporting integration with MCP - compatible clients.

The Enkrypt AI MCP Server is a server for integrating red team testing, prompt auditing, and AI security analysis. It is compatible with the Model Context Protocol (MCP) standard and supports functions such as real - time prompt risk analysis and adversarial prompt generation. It can be seamlessly integrated with MCP clients such as Claude Desktop and Cursor IDE.