ProvenanceKit lets any application record, verify, and query the provenance of content — who created it, which AI systems contributed, what source material was used, and how rights flow — all in a cryptographically verifiable, chain-agnostic way.
The problem
When humans and AI systems co-create content, answering “where did this come from?” becomes hard fast:- Which model generated which part of a document?
- Who owns the training data a model was fine-tuned on?
- If a remix earns money, who gets paid and how much?
- Can a publisher prove an article was human-reviewed?
What ProvenanceKit provides
ProvenanceKit is a layered open-source framework built around three primitives:| Primitive | What it represents |
|---|---|
| Entity | A participant — a person, an AI agent, an organization |
| Action | Something that happened — a file was generated, a model was called, a remix was created |
| Attribution | A claim that links an Entity to an Action or resource — “this human reviewed this output” |
How the layers work
Choose your integration depth
Types only
Use
@provenancekit/eaa-types for the schema definitions and write your own storage and indexing.SDK + API
Use
@provenancekit/sdk against provenancekit-api for a full managed provenance backend in minutes.Extensions
Add domain semantics — AI metadata, licensing, payments, privacy — without changing your core setup.
On-chain
Anchor provenance records on any EVM chain. Pluggable chain adapters — bring your own wallet or node.
Next steps
Quickstart
Record your first provenance event in under 5 minutes.