Calendarly

Inspiration

Imagine scheduling a one-hour meeting for 20 people this week. How much time will that cost? Imagine the hassle - setting meeting timeslots, and sending the timeslots through an instant messaging app for voting, and someone will say that this time does not work for them, and now we need to add extra timeslots for voting. Moreover, some people will have time preferences (e.g. Not prefer mornings), some people can only attend from this time to this time, and because you know no one's schedule, you found yourself just spending hours just to schedule it. I have experienced myself first-hand, so I want to build a platform that helps with work groups scheduling their meetings.

What it does

Using Calendarly, the organizer can simply describe the meeting in plain language, stating their rough time requirement, who is required, and everyone's preferences. The owner can also connect to their calendar application, and AI will check the free/busy times of their calendar, and suggest the best timeslots for the meeting. The owner can then share the link for everyone to vote. For the guests, they can also connect their calendar to automatically vote for the timeslots, and the AI will suggest alternative timeslots for the meeting. They can also describe their own availability in their own words, or suggest a time. After everyone finishes voting, the organizer will receive an email and can visit the dashboard, and AI will give a brief note on which meeting timeslot is the best. Finally, the organizer can decide the meeting time, and an email is automatically sent to notify everyone about the meeting time.

How we built it

In technical terms, I have used Next.js on Vercel for the web framework, Supabase for database and Google authentication, Google Calendar API, GLM through z.ai, Brevo for email, and Tailwind for CSS and designing the website.

The app is based on an engine later, where the slots, time zones, constraints and ranking format are set and processed, and it is fed by the AI layer, which parses any plain text statements into a schema and explains the reasoning behind the final timeslots chosen. The calendar provider layer also supports the calendar import from Google, and only the free/busy times are read (i.e. not the event information) to create and match the meeting timeslots.

Planning documents such as the scope, PRD and spec of the product are planned thoroughly, and the idea is refined by interviews with AI. Then, a fast build is done by Claude Code, and for the verification, the platform itself has been tested by Claude Code by unit tests and demo tests on a live browser using the Google Chrome Claude extension. The vercel and Supabase connectors are also used to speed up the development process.

Challenges we ran into

Challenges with defining the problem: My intended audience was broad at first, which caused my idea to be vague and hard to present, and it was difficult for me to choose a right direction for my product. Time used for testing the app: Even though Claude Code could test and make sure the app is functioning, it still cannot replace humans in testing the whole process and giving feedback for aligning with the idea in my mind.

Accomplishments that we're proud of

The application is actually working on Vercel, and most features defined by my scope has been implemented, tested and shipped. The build of the app only cost a short time of around two days, and only occasional commitments after Claude Code has automatically built the product and run the test cases itself. Feedback sessions from our users also showed great feedback, and some even mentioned that this would have saved over 50% of their time in scheduling the meetings.

What we learned

Think big and start small: I have learnt that it is impossible to start from a large and vague idea, and it is better to present one specific pain point first, and first build a minimum viable product for my idea. Automated testing: This is my first project where I have equipped my coding assistants with connectors such as Vercel and most importantly, Google Chrome extension, and I understand that not every part of the development process can be replaced even though AI has the vision and tests the product, and human feedback is still important in the software development process.

What's next for Calendarly

Connecting to other calendars: Our app will be able to support Outlook and Apple calendars in the future Google verification and a custom domain Further improvements in UI/UX, and invite individuals for testing and registration in our platform Collaborating with SMEs to provide this service to the business world

Built With

• brevo
• html
• next.js
• oauth
• supabase
• typescript
• vercel