CrisisRoom AI

Inspiration

CrisisRoom AI was inspired by the need to make AI-assisted incident response safer and more practical. The project explores how an AI agent can investigate a security incident, analyze evidence, and prepare a containment action while keeping a human in control of the final decision.

What I built

CrisisRoom AI is a fictional incident-response dashboard connected to WebMCP. It exposes structured tools that an AI agent can use to retrieve incident information, inspect affected assets, analyze the situation, create an action draft, and generate an incident report.

Human-in-the-loop

The key design principle is that the AI can investigate and prepare an action, but it cannot execute containment on its own. The proposed action remains pending until a human explicitly approves or rejects it.

What I learned

This project helped me understand how WebMCP can connect AI agents directly to useful web application capabilities instead of relying only on UI interaction. I also learned how important clear tool definitions, application state, and human approval are when building AI-assisted security workflows.

Challenges

The main challenge was designing a realistic workflow while keeping the application fictional and safe. I focused on demonstrating the interaction between the agent, WebMCP tools, the incident state, and the human approval interface.

Built With

• css
• html
• javascript
• vite
• webmcp
• webmcp-polyfill