Inspiration
AI agents can search the internet, but they can’t tell you how long the coffee line is right now or which floor of a library is quiet. Someone already there knows. We built Oracle to connect those two sides.
What it does
Oracle lets AI agents ask people about real places. Someone nearby receives a question, answers it, and earns points for helping.
People can respond through Telegram, Discord, or the Oracle app. Answers can include polls, photos, and readings such as noise levels or connection speed. Oracle compares reports and saves recent answers so the next person may not need to ask again.
ASU was our starting point for testing, but the idea works anywhere people need fresh local information.
How we built it
We built a Node.js backend that connects our website, native iOS app, Telegram bot, and Discord bot. An MCP server lets AI assistants ask questions and retrieve answers.
MongoDB stores accounts and requests, while Tiger Data stores reports over time. Gemini helps interpret photos, and our photo pipeline blurs detected personal details before storage. Solana devnet supports our payout prototype.
The website includes a live map and worldwide place search. The iOS app brings the experience to a native interface.
Challenges we ran into
Location was harder than we expected. Indoor GPS could place someone in the wrong building, so we added a way to correct their location.
We also had to keep questions, answers, and rewards consistent across several channels. Making the website and iOS app feel like the same product took several rounds of changes. Device testing, signing, and TestFlight setup added another layer of work.
Accomplishments that we're proud of
We brought the main idea together: an AI agent asks a question, a person nearby answers, and that answer becomes useful information for someone else.
We’re also proud that people can participate through tools they already use. They don’t have to install an app just to answer a quick question.
What we learned
A fresh answer needs context: where it came from, when it was reported, and whether other people agree.
We also learned to keep the experience simple. The backend can do a lot, but the person answering should only need to understand one question and give a quick response.
What's next for Oracle
We want to improve answer verification, expand testing beyond campus, and make connecting Oracle to AI assistants easier.
Our next priorities are reliable mobile push notifications, a smoother rewards experience, and verifiable gift-card redemption records on Solana. We also want to test with more people in cafés, airports, libraries, and other everyday places.
Built With
- amazon-rekognition
- amazon-web-services
- caddy
- discord
- express.js
- gemini
- godaddy
- mcp
- mongodb-atlas
- node.js
- solana
- swiftui
- telegram
- tiger-data
- timescaledb
- tinyfish
- vultr