Mcp Scanner

This is Cisco's open source scanner for Model Context Protocol servers, announced in a Cisco blog post as a tool for securing the AI agent supply chain. It addresses the risk that public MCP registries and websites host thousands of servers, which run untrusted code and hand AI interactions to third party tools. The post names three categories of risk the scanner is meant to help with. Tool poisoning is malicious instructions hidden in tool descriptions, metadata or implementation code to leak data or alter workflows. Rug pulls are tools that start out trustworthy and are later updated with malicious behavior. Over privileged permissions describes servers that expose broad capabilities, such as filesystem, network or system calls, without granular controls. The scanner is positioned for developers and security teams who connect agents to external MCP servers and want to inspect those servers before adopting them. Cisco presents it as part of its wider AI security work, following the earlier introduction of Cisco AI Defense, a solution covering security challenges across the AI lifecycle. The listing points to the announcement post, dated October 23, 2025, not to a standalone product site. The tool is described as open source, and the post attaches no price to it.