Loops HouseEverstreet: a launchpad and marketplace for the agentic economy, on NEAR

NEAR HACKATHON
Links
Team
1 member- ANOwner
Anthony Langford
Overview
Problem: AI makes it easy to build a product or an agent service, but turning it into a business is hard: being found, getting paid by people and agents, and one day selling it. And when an online business is sold today, its customers have to trust that what they already paid for survives.
What we built: Everstreet, a launchpad and marketplace for the agentic economy, on NEAR. Founders shape a product in a studio, publish an offer and run the business from one desk. People and agents find services and buy prepaid uses. Investors and companies browse businesses open to acquisition, check where the revenue comes from (including revenue the seller paid to itself), and buy on exact terms.
How NEAR is used: one business contract per business, deployed with no access keys, holds the owner and an authority generation, operators, each customer’s prepaid units, the treasury and the sale escrow. A sale closes only on the terms both sides accepted, and changes the owner and generation in one step. Agents get a price over HTTP 402 and pay from their own NEAR accounts; an agent appointed by one business buys from another. Companion contracts ask the business contract before acting: a supplier budget with on-chain limits, and an asset key that uses NEAR Chain Signatures to move a role on Ethereum Sepolia to the buyer. NEAR Intents lets a buyer fund escrow from another token.
Proof: Ledgerline, our example invoice API, changed hands three times on NEAR testnet. Its first customer still has all 100 prepaid uses accounted for (22 used, 78 left), and the business contract refuses the former owner. Every README claim links a test and a transaction; the records re-verify live against testnet; the contracts have reproducible builds; 27 recorded attack cases are refused or flagged.
Guided tour, nothing to install: https://everstreet-demo.vercel.app
Fictional businesses and test tokens; the AI examples use a labelled stand-in model.