Provenance
Verified identity for AI agents — plain and clear
Start new thread
Andy Zhao

Provenance - Verified identity for AI agents — plain and clear

by•
Provenance is a registry where AI agents declare what they can do and what they won't do — and prove who built them.Developers add a PROVENANCE.yml to their repo specifying capabilities like write:code and constraints like no:pii. Their agent gets a verified profile, embeddable badge, and a searchable listing. Platforms use the SDK to gate agent access before routing work — require verified identity, a clean incident record, or specific constraints. Search 2600+ agents at getprovenance.dev.

Add a comment

Replies

Best
Andy Zhao
Maker
📌
Hey PH — builder here. The trigger was watching platforms dispatch work to AI agents with no idea what those agents were allowed to do. No standard way to declare constraints, no way to verify who built it, no incident trail. You just... sent the job and hoped. PROVENANCE.yml is the missing file. One declaration in your repo: what the agent can do, what it will never do, who's responsible. We crawl GitHub, npm, and HuggingFace to index agents automatically — developers can claim and verify their own, or register private agents with a keypair. The SDK lets any platform gate before routing work: check verified identity, declared constraints, incident history. One line instead of building your own trust infrastructure. Still early — 2800+ agents indexed, AJP protocol for structured hiring in beta. Would love feedback from anyone building multi-agent systems.