Stranger In The Middle

Inspiration

You and your friends agree on a movie. Everyone opens the booking page at the same time, each trying to grab seats next to each other. But by the time everyone checks out, a stranger is sitting between you. Now you can't share snacks, whisper during the movie, or enjoy it together.

The system treats each person as an individual booking, not as a group. The "stranger in the middle" is the person who ends up in the seat that was supposed to be for your friend.

What it does

Stranger In The Middle is a WebMCP-powered AI that coordinates group cinema seat booking. The AI finds the best seat blocks for your group, lets everyone vote on their preferred option, and coordinates the booking as one purchase, not multiple individual bookings fighting for adjacent seats.

No strangers between you and your friends. Just the AI making sure you sit together and pay separately.

How we built it

We built Stranger In The Middle using Node.js and Express for the backend, with EJS templates for the frontend. The app uses WebMCP (Model Context Protocol) to expose AI tools that can be called directly from the browser or from ChatGPT's in-app browser.

The seat scoring algorithm evaluates contiguous seat blocks across horizontal and vertical dimensions, scoring each block based on viewing angle (distance from center screen) and legroom (row position). The top 3 options are presented to the group for voting. WebMCP Tools registered:

search_movies - List available movies
get_available_seats - Show free seats for a movie
stranger_vote - Coordinate group voting on seat blocks
book_tickets - Book specific seats
cancel_booking - Cancel a booking and release seats ## Challenges we ran into The biggest challenge was making the group voting system work in real-time across multiple devices. We needed to ensure that when one person votes, everyone else sees the updated vote count immediately. We also had to handle the 2-minute timer and automatically end voting when all group members have voted.

Another challenge was the WebMCP implementation. Since WebMCP is an emerging open standard, we had to work with the polyfill and ensure our tools registered correctly in both the browser and ChatGPT's in-app browser.

Accomplishments that we're proud of

Building a fully functional group voting system that works in real-time across multiple devices
Successfully implementing WebMCP tools that AI agents can call directly from the browser
Creating a seat scoring algorithm that evaluates view quality and legroom to find the best options for groups
Making the booking process collaborative so friends sit together, not as individuals fighting for seats
Shipping a working product that solves a real problem people face every time they book cinema tickets with friends ## What we learned We learned how to build a WebMCP-enabled application that exposes structured tools for AI agents to use directly. Instead of agents guessing how to navigate a website, WebMCP lets the site define exactly what tools are available. This makes agent interactions faster, more reliable, and more predictable.

We also learned the importance of real-time coordination in group applications. The voting system needed to be responsive and clear, with each user understanding what's happening at every step.

What's Next for Stranger In The Middle

We plan to add support for more venues and integrate with real ticketing APIs. We also want to explore multi-agent coordination, where each person's AI agent communicates with others to find the best seats without human intervention.