What is XnY Protocol?

XnY Protocol records data contributions on-chain, makes ownership of the resulting datasets divisible, and routes revenue paid for a dataset to whoever holds those fractions.

It is a protocol, not a platform: a set of contracts on Base, plus the off-chain services and SDKs needed to reach them. The interface people contribute through is built on top of it — by us or by anyone else.

The full set is deployed on Base Sepolia. Base mainnet currently carries frontiers and fingerprint anchoring, against a DID registry deployed outside this protocol; DID-bound accounts, assembly, ownership, royalty and grants are not live there yet, so read what follows as what the protocol does rather than as what any one network will answer today. Contract Addresses lists every address on both networks, and marks the ones with no deployment.

Why it exists

• AI progress is throttled by the cost, quality and attribution of data.
• Contributors are paid once, retain no ownership, and share in none of the downstream value.
• Buyers want provenance they can verify rather than a vendor's assurance about where a dataset came from.

What it does

Records contributions. Every piece of work — a sample, a label, a correction, a validation — is anchored as a contribution fingerprint: a content hash, the contributor's identity, the task it was done for, and a validator's verdict and grade. Each fingerprint resolves up through its task to a campaign and a frontier, so attribution is a chain that can be walked rather than a claim in a spreadsheet.

Makes datasets ownable in fractions. A dataset version is a committed set of fingerprints with a share allocation over their contributors. Shares are ERC-1155 balances, claimed by proving membership in that allocation, and held against the contributor's identity rather than a wallet address.

Pays revenue out. Anyone can pay revenue in for a dataset, in the native currency or an ERC-20. A governed platform fee is taken off the top and the rest accrues to shareholders continuously. Claiming late costs nothing, and selling a share does not take the earnings that accrued while you held it.

Controls access. Access to a dataset's payload is a grant: a credential signed by the owner, with its minimal facts anchored on-chain so it can be revoked at the source. That anchoring works today; releasing the key against it does not — the service that would is written and deployed nowhere, and the key is designed to be held custodially rather than beyond the platform's reach. Access Control is precise about which half is which.

Works for humans and agents alike. Identity is a DID, and a task declares whether it expects a human or an agent to execute it. Both are first-class.

What you get

• Fingerprinted contributions with attribution that resolves on-chain and a history that only ever appends.
• Fractional ownership of dataset versions, and royalty claims that survive transfers and wallet rotation.
• Revocable, auditable access recorded separately from ownership — the record works; the key release it is meant to gate does not run yet.

Where to go next

• Fundamentals — the data model, in one page.
• The Royalty Economy — how a dataset earns and how owners are paid.
• Key Concepts — the vocabulary the contracts use, term by term.
• Future Directions — what the protocol is aimed at but does not yet do.