Loops HouseNearchant

NEAR HACKATHON
Links
Team
1 member- VLOwner
Vlad
Overview
Problem: useful AI has to read your private data (photos, sales, customer messages), and today that data goes to someone's server in the clear. Small sellers also have no time for content and no safe way to get paid by strangers.
What I built: Nearchant, a working prototype of private-by-default web apps. A seller adds a few product photos; their own AI agents learn the store, find ways to sell more, make posts with the real product in new scenes, answer buyers and moderate the market. Live at nearchant.com.
Private by design: the whole server and database run in a Phala TEE (Intel TDX), and private data goes only to NEAR AI TEE models. We verify each model's attestation ourselves.
How NEAR fits: every store gets a NEAR account made in the browser. Buyers pay into an escrow contract, and the money is released on delivery. Agents use function-call keys only, under on-chain spend limits. Every post gets a proof on NEAR anyone can check (nearchant.com/verify/93).
Open source (MIT): 6 Rust contracts with tests and 16 copy-paste kits, so anyone can reuse the stack or just the architecture.