cisco-ai-defense/mcp-scanner is a Python tool that inspects MCP servers and tools to find potential threats and security issues. The latest release in this repository is 4.8.7.
MCP (Model Context Protocol) is a specification that defines how models use external tools. This tool checks such servers and tool descriptions for prompt injection and risk factors, and the inspection targets are limited to MCP servers and tools. The README states that it detects malicious MCP tools by using the Cisco AI Defense inspect API, YARA rules, and LLM-based analysis, either together or separately. In form it offers a CLI, a REST API server, and an SDK. The language is Python, the license is Apache-2.0, the repository was created on September 24, 2025, and the star count is shown as 1096.
The given material does not include the changelog or release timing for 4.8.7, so what changed in this version cannot be confirmed. No proprietary benchmarks or measurements that would gauge detection accuracy are provided either. The material also does not clarify whether to use the three analysis methods together or separately, or which combination is advantageous. The momentum index on a 10-point scale produced by the pipeline came out low at 1.3, and no other indicators were produced. For teams that operate or integrate MCP servers, it looks like a tool that can be put to use right away for organizing inspection targets. That judgment is better left until after the 4.8.7 changes and detection accuracy have been confirmed.