The problem DorkFi solves

DorkFi is a modular, cross-chain lending protocol designed to make DeFi safer and more automated.

Today, users must manually manage borrowing positions, rely on off-chain keepers, and navigate fragmented liquidity across chains — exposing them to liquidation risk and infrastructure failures.

DorkFi enables overcollateralized borrowing and stablecoin minting (WAD) with deterministic risk models and protocol-native automation.

By integrating scheduled on-chain execution and layered architecture, it removes reliance on centralized bots while improving capital efficiency.

DorkFi transforms lending from a manual, fragile process into programmable, self-executing financial infrastructure.

Challenges we ran into

Hedera Schedule Service and Forge script: non-contract at 0x16B

This project is a Hedera-native lending pool that uses Hedera’s Schedule Service (HIP-1215) so a keeper runs on-chain at fixed intervals without an off-chain bot. The pool extends a base lending design with a Hedera-specific contract that calls the Schedule Service precompile at 0x16B to reschedule itself (e.g. for interest accrual or maintenance). Starting the keeper from Foundry failed with “call to non-contract address 0x16B” because Forge still runs the script in a local VM that doesn’t have Hedera’s precompiles; we fixed it by sending the start transaction with cast send so execution happens only on Hedera. The result is a self-scheduling keeper that relies on Hedera’s native scheduling instead of external infrastructure.

Use of AI tools and agents

Use of AI Tools and Agents

AI tools were used during development to accelerate UI creation, debugging, and frontend iteration. However, there are currently no live AI agents embedded in the deployed protocol. All lending logic, settlement, and automation are deterministic and executed directly through smart contracts.

The system leverages x402 scheduling primitives for guaranteed future execution, enabling network-level automation without centralized keepers.

Looking ahead, the architecture supports a User-Controlled Protocol (UCP) model where AI agents could operate as advisory or automation layers within user-defined permissions. In this design, AI would analyze risk or optimization opportunities, prepare transactions, and optionally schedule them via x402 — while smart contracts enforce final rules and custody remains fully on-chain.

This approach preserves decentralization today while enabling intelligent automation in future iterations.