About the standard
Governance
How OARS is maintained, versioned, and evolved — and who is accountable for it. A standard is only as trustworthy as the body that keeps it.
Who maintains OARS
The Open Agent Readiness Standard is published and maintained by Knov.ai. We author the specification, run the verification process, and operate the directory that exposes verified entities to AI agents. The standard is open: any entity may implement it without license or fee, and the machine-readable schema and full specification are published for anyone to read and build against.
Open and free to implement
Implementing OARS never requires payment to Knov.ai. Assessing your site, generating an
oars.json, and listing in the directory as self-declared are free. Verification adds
a human review by a Knov.ai reviewer who confirms what an automated check cannot; the
standard itself remains open regardless of whether you pursue verification.
How the standard changes
Changes to the specification are proposed publicly, reviewed with community input, and published
with version history. Every oars.json records the schemaVersion it was
written against, and each manifest’s @context points at the permanent schema URL
so existing files never break when the standard advances.
The current version is OARS v1.1. The full change history lives at the bottom of the specification.
Certification and monitoring
Verification reflects our assessment at a point in time against the standard. To keep verified status meaningful as sites change, the specification defines:
- Cumulative compliance — each level requires full compliance with every level below it.
- Recertification — verified entities recertify on a defined cadence, with automated monitoring between cycles for higher levels.
- Regression handling — if monitoring detects a previously passing requirement now failing, the entity enters a cure period before its level is adjusted.
- Appeals — an entity that believes it was incorrectly assessed may submit an appeal, which a second reviewer assesses independently.
The exact cadences and timelines are defined in the certification policies section of the specification.
What we do and don’t do
Knov.ai is infrastructure, not a marketing service. We define and verify the standard; we do not optimize sites for AI-search rankings, write content, or build websites. That separation keeps verification credible — we are not grading work we were paid to produce.
Participate
Feedback on the standard, proposed changes, and corrections are welcome. To weigh in, report an issue with an assessment, or ask about implementation, write to [email protected].