Safe Agent-Assisted Booking

Inspiration

AI agents are becoming capable of completing real tasks on the web, but actions such as booking should remain transparent and under the user's control. I wanted to explore how WebMCP can let an agent discover services and check availability while keeping the final booking action safe and explicitly confirmed by the user.

What it does

Safe Agent-Assisted Booking is a WebMCP-enabled booking demo.

It exposes three tools to an AI agent:

• search_services — discovers available services.
• check_availability — checks available time slots.
• create_booking — creates a booking only after explicit user confirmation.

The project uses synthetic demo data and does not depend on external booking services.

How I built it

I built a small browser-based booking application and exposed its capabilities through WebMCP tools.

The implementation separates read-only actions from the state-changing booking action. Search and availability checks can be performed safely, while booking requires a confirmation step before execution.

I also added request idempotency and slot-conflict protection so repeated requests do not accidentally create duplicate bookings.

Challenges

The main challenge was designing a write action that remains useful to an AI agent without removing control from the user.

I focused on:

• explicit confirmation before booking;
• predictable tool contracts;
• idempotent booking requests;
• conflict detection;
• safe handling of cancellation or rejected confirmation;
• keeping the demo simple and reproducible.

What I learned

This project helped me understand how WebMCP can turn a normal website into an interface that both humans and AI agents can use.

The most important lesson was that agentic web actions need more than tool access. They need clear boundaries between reading information and changing state.

What's next

The next step would be connecting the same architecture to a real booking backend, adding persistent storage, authentication, and more advanced availability rules while preserving explicit user control for sensitive actions.What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for Safe Agent-Assisted Booking

Built With

• chrome
• cloudflare
• css
• devtools
• github
• html
• javascript
• webmcp
• workers