Open Standard / v0.2 / MIT License

Disclose is the confidence layer for AI agents.

An open standard for publishing verified operational signals. Agents evaluate businesses, services, and digital resources before making decisions.

View on GitHub Read the Spec

01 / Discovery

One file. Any agent.

Merchants publish a /.well-known/disclose.json on their domain — the same convention as robots.txt, but for AI agents. No registry. No platform dependency. Any agent that knows the convention can find it.

02 / Attestation

Named verifiers. Staked reputations.

Every signal is either verifier-attested or merchant-reported. Agents know which is which.

03 / Architecture

Three scopes. One document.

Signals are published at Merchant, Offer, or Item scope. An agent comparing the same product across two sellers can read merchant-level confidence, SKU-level performance, and item-level attributes — all from a single endpoint.

Premise

When an agent helps someone buy something, it reads the marketing layer. Price. Reviews. Availability. Not the operational layer.

Disclose is the schema for the operational layer. No scores. No badges.

Starting with commerce. Built to go further.

Governance

Disclose is MIT-licensed. The spec is designed to be owned by no single company — not by a platform, not by a verifier, and not by its founder.

Independent governance is in formation. The working group is the first step.

Working Group

The spec exists. The working group is in its first cohort.

We're looking for three types of participants: merchants with clean operational data and existing verifier relationships, verifiers who already hold the signals the schema defines, and infrastructure builders who will find the gaps before they become problems at scale.

The ask is specific — read the spec, stress-test the schema against your data, and tell us where it breaks. That's it.

Read the Spec Join on GitHub →