TradeFinity
People have stuff they do not need anymore, and they often want things they cannot easily afford, find, or justify buying.
Core trick
Gemini extracts offered and wanted items from photos and chat, then a search over listings assembles multi-step trade chains.
Description
People have stuff they do not need anymore, and they often want things they cannot easily afford, find, or justify buying. Barter should be the perfect solution, but in practice it almost never works because finding a direct one-to-one trade is really hard. You might have a bike and want a guitar, but the person with the guitar probably does not want your bike.
The Oracle solves this by finding trade paths instead of simple matches. A user can upload a photo or describe something they want to trade, then say what they are looking for. The Oracle looks through real local barter listings and finds a chain of trades where everyone gives something they no longer need and gets something they actually value, with no money involved.
For the MVP, we are focusing on real Craigslist barter listings from the Bay Area. A user can chat or upload a photo, and Gemini helps understand what they are offering, predicts what people might realistically want in return, searches the listing pool, and finds multi-step trade rings.
For the hackathon demo, the goal is simple: prove the core loop works. The product should understand an item, figure out trade intent, search messy real-world listings, and show a specific trade path that feels possible in the real world.
README excerpt
# The Oracle The Oracle is a managed-agent barter platform built for the Google I/O Hackathon 2026. It helps people turn something they already own into something they actually want by coordinating Gemini-powered agents around real secondhand-market data and Bay Area Craigslist barter listings. ## Problem Statement People often have valuable items sitting unused, while the things they want are locked behind cash prices, fragmented marketplaces, and one-to-one barter matches that almost never line up. Barter should be powerful, but in practice it is too hard to price items, find compatible counterparties, avoid scams, and discover multi-step trade paths where everyone receives something useful. In an ideal world, a person could upload a photo or describe an item, state what they want, and let an intelligent agent network search, price, reason, and protect the transaction without taking away human control. Our vision for The Oracle is to become the trusted agent layer for secondhand exchange: a system that can understand messy listings, coordinate safe trade paths, and make no-cash commerce feel as direct as buying from a normal marketplace. The MVP proves the core loop with real Bay Area Craigslist barter data. A user can upload a photo or chat with the Concierge Agent, Gemini extracts item and intent, managed agents price and research the item, the app searches messy listing pools, and The Oracle reveals realistic multi-step trade paths where everyone gets something usef