MindBridge

• 

  Home Page

• 

  Select an option

• 

  Waiting page -- select AI or human

• 

  Create an account

• 

  Choose a listener

• 

  Connecting to a listener

Inspiration

Mental health resources are scarce, expensive, and often inaccessible, especially for college students. We wanted to build something that makes real human connection available to anyone, at any hour. We built MindBridge around that idea: instant, anonymous, human support with zero barriers to entry.

What it does

MindBridge connects people seeking emotional support with verified human listeners in real time. Users select their mood and what kind of support they need, then our matching engine pairs them with an available listener. While they wait, a calming ElevenLabs AI voice companion (personalized to their mood) keeps them company until the human picks up. Every listener is verified through World ID, guaranteeing no bots, no fake accounts, and one identity per person. Sessions happen over live video and text chat.

How we built it

We started with a full Figma prototype to map out every screen. The frontend is built in Next.js with Tailwind, and the backend runs on a Node/Express server deployed via ngrok so participants on different devices can join the same session. We integrated World ID's IDKit for zero-knowledge proof-of-human verification, ElevenLabs' Conversational AI SDK for the personalized waiting room voice, and WebRTC real-time video calls. User accounts, roles, and session data are stored in MongoDB hosted on AWS.

Challenges we ran into

Getting World ID verification working took significant debugging to get right. Building the ElevenLabs Conversational AI agent was hard to make the voice feel genuinely calming rather than robotic. The hardest problem overall was synchronizing the live video call — getting two users on different machines to join the same call at the same moment, with chat running alongside it without the session state getting out of sync.

Accomplishments that we're proud of

Our biggest win was getting real-time connection working between a user and a listener, complete with live chat. At one point we almost gave up on the matching system entirely, so watching two verified humans connect in real time on a product we built from scratch in 36 hours made it all worth it.

What we learned

We learned how to architect a multi-agent system, using World ID's zero-knowledge verification and ElevenLabs' Conversational AI audio. We also learned how to deploy and manage a live server, which was a first for our team.

What's next for MindBridge

We want to build an intelligent matching algorithm that improves over time and pairs users with counselors based on session history, feedback ratings, and specialty alignment. Longer term, we see MindBridge expanding into scheduled sessions with both licensed therapists and peer volunteers and crisis escalation pathways that automatically route users to certified professionals.

Built With

• elevenlabs
• figma
• mongodb
• next.js
• ngrok
• node.js
• typescript
• webrtc
• worldid