Attendly

Inspiration

This is built on another project of mine, which was originally written for my childrens PTA who were looking for a simple to use platform to help them keep track of ticket sales. The typical users are mums and time-strapped parents, and wanted something that reduces friction for them and the target audience.

I wanted to explore expanding on this using the new WebMCP to see if it was possible for a parent to quickly get tickets for upcoming events, without having to navigate through the site - by using natural language inside ChatGPT.

Taking this further, I'd expand it further by adding the ability to create events.

What it does

It uses natural conversation to discover relevant events for the user, allowing a parent, for example, to sign-up for upcoming events at a school without having to step through the site hierarchy. It provides everything they need to make an informed decision, and a confirmation record once registered.

For event organisers, it can provide quick feedback, i.e. 'Who is yet to arrive?' or following an evacuation 'Produce an evacuation report'. In what is becoming a familiar interface (ChatGPT).

How we built it

I used GPT 5.6 Sol to brainstorm a lightweight prototype of my existing application, and create Epics and user stories for the work which was kept in Linear. We then tackled each story in sequence, building out and pivoting as the demo became more fleshed out.

I used Fable 5.1 to pickup some edge cases where Sol was too quick to finish (i.e. it implemented a partial persistent storage), and independent agent analysis of each story and unit tests.

Challenges we ran into

GPT 5.6 Sol was quick to to a partial implementation, loosing context of the full scope of the build, and twice delivered partial features that I passed over to Fable 5.1 to complete. It also had a tendency to lean too heavily on the 'prototype' guidelines, and stamped this into the site.

I had some challenges understanding how to trigger the WebMCP standard, which didn't work in ChatGPT Chat mode - it needed to be in Work mode to load the Internal Browser tool and discover the exposed WebMCP on the site. Despite several prompts, the Chat could not even explain why it was unable to find it.

Accomplishments that we're proud of

I am glad I took the time to play with WebMCP - this is my first ever challenge I've entered. I have spent the last several years working with LLMs and AI, building out a few projects - I am someone who learns by doing and I feel like I got some real value out of this which I can carry forward into my other projects.

What we learned

I learned how to implement WebMCP! I also learnt a few more tricks in how to prompt GPT 5.6 to keep it on track.

What's next for Attendly

This will eventually be implemented into the full Attendly product. The stories are in the backlog, but it's a big backlog.

Built With

• cloudflare
• react
• typescript
• vite
• wrangler