Troy AI has launched an AI Security enhancement solution based on MCP... real-time blocking of intelligent agent Hacker attacks.

The artificial intelligence security platform startup TrojAI has launched a new runtime defense solution “TrojAI Defend for MCP” aimed at protecting AI workflows for intelligent agents. This solution focuses on proactively detecting and blocking potential security risks in the rapidly emerging wave of AI applications driven by the recent rise of MCP (Model Context Protocol) technology.

MCP is an open standard designed to enable secure connections between various AI agents and external data, tools, and services. As enterprise-level AI applications become more widespread, the adoption rate of MCP continues to rise, but new threats such as the creation of unverified MCP servers, the execution of malicious code by unauthorized external tools, and sensitive data leaks have emerged. TrojAI defines these risks as typical security blind spots in the “operational real-time layer” and announces that it will respond through a new product that supports unified visualization, policy analysis, and runtime control functions.

“TrojAI Defend for MCP” extends the protection scope of existing security solutions to the MCP layer, allowing for full traffic monitoring between MCP servers, agents, and tools, and real-time blocking of communication with unauthorized servers or tools. Its “MCP Server Registration and Tool Authentication” feature strengthens organizational governance by eliminating shadow servers. Furthermore, the “MCP Traffic Visualization” module can track each request and response item, enabling early detection of signs of tool tampering, command pollution, and data theft.

The solution also features tracking tools to define changes to prevent unauthorized modifications or attempts at abuse, and enables real-time automatic execution of security rules through a dedicated MCP policy engine. TrojAI CEO Lee Weiner emphasized: “By protecting agent AI workflows in real-time operational environments, we help enterprises achieve proactive threat response rather than passive reaction.”

TrojAI has completed five rounds of financing to date, raising a total of $11.4 million (approximately 16.4 billion KRW). Major investors include Flyfish Ventures, Build Ventures, TechStars Central, Flybridge Capital Partners, and Alteryx Inc. Although the new product is still in the early stages of commercialization, its solution for vulnerabilities in the MCP architecture AI infrastructure is making it a highly regarded security platform in the rapidly changing agent AI market.