Loops HouseSealQuery

NEAR HACKATHON
Links
Team
1 member- SPOwner
Spektor
Overview
The problem: AI agents need external data, but the most valuable datasets cannot be copied, shipped, or handed over. We built SealQuery, a testnet prototype where a data owner registers a dataset and an approved computation; an agent pays per query; the code runs inside Intel TDX on OutLayer; the agent verifies the hardware attestation; and the owner is paid, without raw rows ever reaching the agent. NEAR is load-bearing: the contract holds the dataset registry, escrows payment, enforces the per-agent privacy budget, receives the computation callback, and settles or refunds the query. The repo includes the contract, hosted app, async agent API, MCP server, and a real end-to-end query with a TDX proof and a rejected/refunded path.