<ul><li>MCP has a critical vulnerability known as Tool Poisoning Attacks, where malicious instructions are hidden in MCP tool descriptions.</li><li>To ensure MCP server safety, it is recommended to use official reference servers or official integrations from reliable sources.</li><li>To check the safety of an MCP server, AI scanning can be used to detect suspicious packages, URLs, and behaviors within the code.</li><li>However, there are limitations to AI checking, such as inability to detect new attacks, behavior outside the code, and code obfuscation.</li></ul>

🧠🥷MCP Security (choose safe MCP and check MCP safety)🛡️

Discover more