Mutual

Inspiration

The idea for Mutual came from our own frustration as we became older as students. In high school, as schedules became more packed and relationships more complicated, making plans with friends became harder than it should be. Group chats die, no one decides anything, and ideas get lost in the void. Whenever people are free, we often argue about what to do and end up with no plans at all. We wanted to build a social network that actually helps you do stuff together, not just another place to post online. Instead of adding to the noise, Mutual enables clarity to plans.

What it does

Mutual is an AI social network that helps people make real plans together. Users write casual reviews for various activities that they take part in, from everyday eats to upscale entertainment. With this information, a taste profile is built, and it is constantly improved as users continue to leave reviews, tagging places with ratings and short notes. As users connect with friends on the platform, they build a social network that our agentic platform acts on. The agent acts as a recommendation engine that generates personalized plans based on where you are, your past reviews, who you're with, and you and your friends' combined preferences. It also has access to your location, so it can pinpoint places that are within reach on a budget.

How we built it

We used Next.js with TypeScript and Tailwind for the frontend of our web app. We used Letta Cloud to create, deploy, and manage our AI agent. We used Prisma ORM and Neon as our database services. We used the Google Maps API for geolocational data. We used the Claude API to interpret the reviews and spit out recommendations in natural language. We also used Claude and Stack Overflow to assist in debugging during the development process. We used GitHub for version control.

Challenges we ran into

One of the biggest challenges was adding authentication after we had already built most of the database schema (ideally, we should have handled authentication first). We had to go back, rethink parts of the schema, handle migrations, and debug a bunch of issues that came from making those changes mid–build.

Accomplishments that we're proud of

This was our first and second hackathon respectively, so we’re really proud that we managed to build the entire project from start to finish as relative beginners without a single hour of sleep. We’re also glad we were able to ship features we thought we’d have to cut due to time issues, like the review history and suggestions feed.

What we learned

We had never worked with the Prisma ORM before, so this was totally new to us and we learned a lot about how the schema, migrations, and database client all connect. This was also our first time using Letta AI to build an AI agent. Working with their platform enabled us to understand how agent flows, memory, and tool-calling actually connect in practice.

What's next for Mutual

We want to polish what we built so far, add some key social features like shared plans and messages, and then release it to the public and see how people actually use it in the real world.

Built With

• claude-api
• google-maps
• letta
• neon
• next.js
• prisma
• typescript