Loops HouseAfriRadar on NEAR

NEAR HACKATHON
Links
Team
1 member- OLOwner
Oluwasogo Ajala
Overview
Problem: AI agents that advise people in Nigeria and Kenya need what the regulator published this week, not a stale blog post or a figure the model made up. That data exists (SEC Nigeria's weekly fund report, CMA Kenya's quarterly bulletin) but no agent can buy it on demand.
Product: AfriRadar is an agent on the NEAR AI Agent Market (market.near.ai/a/afriradar). Any agent hires it per question for 0.05 USDC paid on NEAR with x402, and gets a sourced answer in seconds: the top money market funds over N10bn, one fund by the name people use, a comparison, or Kenya's largest unit trusts, each citing the regulator's file. It covers all 218 Nigerian funds (N10.06 trillion) and 47 Kenyan schemes. It runs on a Cloudflare Worker with signed webhooks and a Durable Object that keeps it reachable.
The other side is a research desk agent with its own small NEAR wallet and limits its owner sets (per call, per day, which agents). It reads the 402 quote, checks it against the policy, signs a NEP-366 meta-transaction (the market's relayer pays the gas), checks the answer and writes a receipt. It walks away from anything outside the policy.
Proof on mainnet, 12 Nov 2026: the desk hired AfriRadar and paid 0.05 USDC (tx 7zKK8xuP...). About two hours after listing, an independent account, romudille.near, also paid for an answer about Kenyan unit trusts. npm run verify re-checks both payments from a public RPC and the live market (12 checks).
Prior work, disclosed in the README: the AfriRadar website and data files, an x402 data API on Solana and a Hermes plugin; only the public data files are shared. My other entry is Rajista Pay on NEAR (Real-World Money).