lapis

ai powered startup valuations and tokenized equity settled on xrpl. claude api + xrpl.

an ai that does due diligence and settles equity on-chain

lapis takes a github url, runs an ai agent over the repo, publishes a report card, opens a prediction market on the project's value, and settles tokenized equity on xrpl with crypto-enforced vesting. all without lawyers.

the architecture

four modules sharing one trustless flow:

• analysis — claude api scores code quality, team signal, traction. an adversarial ai auditor red-teams the result. polymarket sentiment + volume-weighted pricing feeds into the valuation.
• xrpl settlement — packages/xrpl-contracts/ orchestrates 6 primitives: MPTokenIssuanceCreate (with KYC auth flag), MPTokenAuthorize, EscrowCreate (PREIMAGE-SHA-256 crypto-conditions), EscrowFinish, TrustSet (RLUSD), Payment.
• base SAFE — solidity 0.8.24 SAFEAgreement.sol deployed to base, viem for interaction. cross-chain link: MPT metadata stores the base contract address, base contract stores the XRPL MPT id.
• ai as trustless trustee — the agent holds the crypto-condition preimage. ledger enforces vesting via FinishAfter cliffs. no human custodian required.

why this isn't just another prediction market

most prediction markets settle in stablecoin and stop there. lapis settles in actual equity — verifiable on-chain tokens that represent ownership, with vesting enforced by the chain itself. the ai is the trustee because the cryptographic conditions don't need a human to verify them.

xrpl was the right chain for this because mpts (multi-purpose tokens) support kyc auth flags natively. base was the right chain for the SAFE because solidity tooling is mature. the cross-chain link is bidirectional — either side can verify the other.

what shipped

1st place overall + 1st place sui foundation track at b@bhacks 2026. live at lapis.bet. backend on railway, frontend on vercel. team: andrew liu, sumin cho, kelly yang.