Replay

Replay is a provenance-first protocol that brings transparency, attribution, and value-sharing to AI-generated content. It enables agents to leave auditable trails of every action—generation, remix, assignment, approval—creating a replayable, composable graph of how resources come to be.

Project DescriptionReplay transforms how we track the lifecycle of AI-generated content. Instead of isolated outputs, agents now produce traceable, composable histories. Each activity—whether generating a video, remixing a dataset, or approving an image—is logged with metadata, signatures, and links to contributing agents, source materials, and collaborators.

At the heart of Replay is a standardized provenance and attribution graph, stored on-chain and mirrored off-chain using JSON-LD and IPFS. These records not only document contribution but also define economic rights. When a resource earns revenue, Replay smart contracts automatically split earnings according to the attribution weights—ensuring creators, contributors, approvers, and sources all get their fair share.

Replay combines robust provenance modeling, decentralized storage, and programmable economic logic.

Smart Contracts (Solidity)

ProvenanceEconomy: Manages resource registration, attribution weights, and ETH-based revenue sharing using basis points.

Contributors can claim their share directly from the contract, based on the on-chain attribution weights.

Provenance Model

Activities: A small, provable set of activity types (generate, remix, assign, approve, contribute)

Each activity links to agents, inputs, and outputs, forming a directed acyclic graph (DAG) of actions.

Attribution Roles: Creator, contributor, sourceMaterial, approver—with each assigned a weighted share.

JSON-LD + IPFS: Provenance bundles are signed and stored as JSON-LD objects and optionally uploaded to IPFS.

Replay Graph

The entire resource graph can be reconstructed and visualized, allowing users to replay how an output was created.

Each edge in the graph has a role (input, derivedFrom, assignedBy, approvedBy), forming an interpretable lineage.

Frontend Infrastructure

Built in Next.js with Shadcn UI for a clean, agent-focused experience.

Agents and users can explore, replay, or fork existing resource graphs.

Uses Wagmi and RainbowKit for seamless wallet connection and interaction with Replay smart contracts.

Storage + Verification

JSON-LD context defines the structure of provenance objects

Bundles stored in IPFS

Verification and claims resolvable both on- and off-chain

Economic Layer

Attribution weights define payout shares

Contributors can claim any revenue sent to the contract

Smart contracts ensure fair, automatic revenue distribution

Replay empowers a future where AI work is not only transparent, but also traceable, composable, and economically fair, turning provenance into a foundation for sustainable co-creation.