Autonomous AI agent navigating cross-chain yields via ENS identity and zero-fee state channels.
The project integrates three core pillars:
Sovereign Identity: Using ENS, users store their risk profiles and "Safe Haven" settings directly on-chain. This ensures the AI agent follows the user's rules without relying on a centralized database.
Macro-Mobility: Powered by LI.FI, the agent scans the entire ecosystem for the highest USDC yields. When a superior opportunity is found on another chain (e.g., moving from 2% on Ethereum to 8% on Base), ORION orchestrates the cross-chain migration.
Micro-Efficiency: Once on a chain, ORION utilizes Yellow Network’s state channels to perform high-frequency, zero-gas rebalancing. This allows the agent to chase even marginal 0.1% yield improvements that would otherwise be eaten by gas fees.
By combining stable asset focus (USDC) with cross-chain intelligence and gasless execution, ORION makes institutional-grade yield farming accessible to the everyday user.
ORION is built using a "Digital Twin" hybrid architecture designed to bridge the gap between production-ready logic and fragmented testnet infrastructure.
The Tech Stack: Agent Intelligence: Built with Node.js/TypeScript, the agent's brain pulls real-time yield data from the DeFiLlama API and cross-chain routing data from the LI.FI Mainnet API.
Execution Layer: We integrated the Yellow Network Nitrolite SDK to open state-channel sessions. This enables the agent to sign off-chain state updates for instant, gasless swaps.
Identity Layer: We used Ethers.js to resolve ENS text records on Sepolia. This acts as our decentralized configuration store, allowing the agent to read risk_profile and safe_haven_address parameters in real-time.
Smart Contracts: We developed a Uniswap v4 Hook (deployed on Sepolia) that acts as a "Risk Guard," programmatically reverting trades if they violate the parameters stored in the user's ENS profile.
The "Hacky" Brilliance: Since LI.FI and Yellow Network have limited cross-chain testnet liquidity for native USDC, we implemented a Digital Twin Simulation.
Real-Data Intelligence: The agent uses LI.FI Mainnet Quotes to determine the actual best route and fees.
Mocked Execution: Instead of a broken testnet bridge, we built a "Logical Bridge" tool. When the agent triggers a bridge, it locks Sepolia USDC into an escrow contract and mints a "Simulated Balance" in our agent's ledger.
Hybrid Flow: This allows us to demonstrate a full "Scan → Bridge → Trade" loop in under 2 minutes, proving the system's logic and UX are production-ready even while external testnets are catching up. We effectively "mocked the plumbing but kept the brain real."

