AegisPay

What it is: AegisPay is a decentralized settlement layer built for the machine economy, DePIN, and dynamic services.

The Problem it Solves: Web3 and stablecoin payments are incredibly dumb right now compared to the intelligence of the Visa or Mastercard networks. If a user wants to use dynamic pricing—like an Uber ride or an EV charger—they are forced into terrible UX choices. They either have to sign a dangerous infinite ERC20 approval (giving the merchant the keys to their wallet), or they have to over-collateralize by locking up massive amounts of upfront capital just in case the price changes.

How it Works: We are fixing this by bringing Visa-style "Auth & Capture" to smart contracts, guarded by an offchain AI risk engine. Users deposit into a liquid userFreeBalances ledger. When interacting with a merchant, they sign a strict off-chain authorization which goes through an offchain fraud prevention check as well. If the price dynamically changes, the merchant requests a secureIncrement. This request routes through our Chainlink CRE AI Risk Engine, which evaluates the transaction against historical data and variance limits. If approved, the CRE safely executes the increment on-chain. The merchant captures the final amount, and the excess is instantly released.

AegisPay is architected across three core pillars:

• The AI Risk Engine (Chainlink CRE): The off-chain intelligence layer. It uses HTTP triggers to process merchant increments, queries Firestore for historical context, and uses a Gemini LLM to dynamically evaluate fraud risk based on MerchantCategory variance limits. EVM Log Triggers automatically sync on-chain Captured and FundsReleased events back to Firestore to maintain a perfect audit trail.
• The Smart Contract (Solidity/Foundry): A Singleton Ledger utilizing a 3-tier accounting system (userFreeBalances, authorizedHolds, and merchantSettledBalances). Critical functions like authorizations and increments are strictly modifier-gated to the Chainlink CRE Forwarder to prevent bypassing the AI risk engine.
• The Frontend (Next.js 16): Built with Tailwind CSS and shadcn/ui. We designed a split-screen "Simulation Console" that allows judges to see the seamless Web2 mobile UX on the left, while watching the real-time Chainlink CRE terminal execution and LLM reasoning on the right.