Loops HouseSpendGuard AI

NEAR HACKATHON
Links
Team
1 member- FAOwner
Fatemeh
Overview
I built SpendGuard AI, a security and policy layer for AI agents using NEAR Intents. Users can create agents, set per-swap and daily spending limits, restrict allowed assets, and review exact swap details before approval.
Building on my existing NEAR AI risk-analysis integration, I added Meteor wallet ownership verification, agent and policy management, time-limited authorizations, encrypted credentials, swap previews, and transaction tracking. The backend enforces mandatory security checks independently of AI decisions, binds approvals to the reviewed operation, and guards against duplicate execution.
Automated security and browser tests cover the workflow using simulated wallet and provider responses. Live mainnet signing and settlement validation remain pending. SpendGuard’s goal is to make onchain agents useful while keeping users in control of their funds.