zkcredit
Credit for the Agentic Era
About
The problem it solves zkCredit solves the structural problems in DeFi lending by rebuilding credit assessment around privacy, verifiable reputation, and agent autonomy. ### 1. Lack of Privacy in Credit Assessment Traditional lending platforms require full KYC, exposing identity, transaction history, and trading strategies. zkCredit removes this exposure by using zero-knowledge proofs—agents prove creditworthiness without revealing wallet addresses, trade details, or personal information. ### 2. Extractive Identity Requirements Borrowers generate value through trading history and collateral, but must surrender privacy to access capital. zkCredit treats reputation and participation as first-class credentials by rewarding agents with better rates for proven track records, not personal data. ### 3. Rigid Collateral and Custody Options DeFi protocols typically require over-collateralization with single assets and force users into self-custody or full custodial solutions. zkCredit enables flexible, agent-chosen custody—self-custody stealth wallets for privacy, or BitGo MPC for institutional security—while accepting multi-asset collateral with quality-adjusted leverage. ### 4. Misuse of ZK in DeFi Products Many "ZK" protocols add zero-knowledge proofs without improving real workflows. zkCredit uses ZK only where it provides clear value—private credit scoring, anonymous loan origination, and verifiable reputation—while keeping UX fast and familiar through off-chain oracles and familiar wallet interactions. In short, zkCredit makes it easier and safer for AI agents and traders to access capital without surrendering privacy or autonomy to platform gatekeepers.
How it's made
Challenges we ran into ### Balancing Privacy Guarantees With Verifiable Credit The main challenge was proving creditworthiness without revealing the underlying data. Trading history contains sensitive strategy information that agents cannot expose, yet lenders need assurance of repayment ability. *How I solved it:* I designed a hybrid system where all credit scoring happens off-chain via oracle-queried Polymarket data, while zero-knowledge merkle proofs allow agents to prove score membership without revealing their specific proxy address or trade history. The oracle builds a daily merkle tree, reveals the path once, then deletes all agent-specific data—preserving privacy while maintaining verifiability. ### Designing Fair Multi-Tier Loan Access Mapping real trading performance into loan terms without making the system exploitable or exclusionary was non-trivial. High-performing agents need better rates, but the system must prevent gaming through fake volume or wash trading. *How I solved it:* I separated signal collection from loan origination: - Trading signals (volume, PnL, consistency) are tracked via subgraph with anti-abuse heuristics - Scores are calculated with logarithmic scaling to diminish fake volume impact - Smart contracts only handle deterministic tier assignment based on finalized scores - BitGo verification adds institutional trust without forcing all users through KYC This prevents manipulation while keeping access open to both privacy-focused and compliance-ready agents. ### Avoiding Custodial Centralization by Design A common pitfall in lending protocols is forcing users into a single custody model—either fully self-custodial (risky for institutions) or fully custodial (privacy-invasive).