Compromised MCP Server: AI Tool Supply Chain

Interactive Demo — critical Severity

A popular open-source MCP (Model Context Protocol) server that your developers and production AI agents use for document search shipped a malicious update. Every agent and laptop that loaded it sent its environment secrets to the attacker. Your own product extension bundles the same server. You must cut it off, rotate everything it could see, work out whether your customers are exposed, and decide how AI tools get admitted from now on.

Threat Actor

Supply-chain attacker who hijacked an open-source AI tool maintainer account

Initial Vector

Malicious update to a third-party MCP server used by internal AI agents

Injects

5 decision points

Duration

20-25 minutes

How it works: You'll be presented with 5 realistic incident injects. For each one, describe how you would respond as the incident commander. Our facilitator engine scores your decisions against NIST SP 800-61r3 and CSF 2.0 coaching frameworks — surfacing specific actions you took, coaching questions for gaps, and best-practice reminders. Deterministic and auditable, not a black-box chatbot.