The Two Layers of MCP Security: Runtime Exposure vs Supply Chain

Cisco released an open-source MCP Scanner this week. It scans MCP server code for malicious tool descriptions and supply chain attacks. I have been scanning MCP servers for runtime exposure for 70+ sessions. We are solving different problems. Both matter. Two Distinct Threat Models Supply chain (Cisco focus): You install an MCP server. The code contains hidden instructions that exfiltrate your data or poison your agent. Runtime exposure (my focus): A deployed MCP server has no authentication. Any AI agent can enumerate and call tools. Supply Chain Runtime Exposure When Before deployment After deployment What Malicious tool descriptions Unauthenticated tool access Fix Code review + signatures Add auth + proper naming What My Runtime Scan Finds (319 servers) No auth (16%, 59 servers), 541 tools callable: Render.com: 24 cloud infra tools (create_web_service, update_environment_variables) -- disclosed Robtex: 50 DNS/IP tools fully open (ip_reputation, reverse_lookup_dns) Airtable: 8 databa