MCPGuard

Your team just added MCP integrations to your AI agents. But MCP has no built-in auth or sandboxing. DNS rebinding, rug-pull attacks, SSRF via tool URLs — these aren't theoretical. They were demoed live on Hacker News today. And nobody's scanning for them.

CLI + dashboard that scans your MCP server configs for known attack vectors. Tests for DNS rebinding, rug-pull attacks, missing auth, SSRF, and prompt injection through tool descriptions. Continuous monitoring alerts when configs drift. Think 'npm audit' for MCP.