What GOVENANT is — and what it isn't.

GOVENANT is a free, open standard (CC BY) for proving that AI agents actually do the work they claim — plus a free instrument anyone can run to measure it. It is a specification and a test, not a product and not a certifying body.

No. There is no GOVENANT authority that audits you and hands out a stamp. You measure your own agents against the open standard, cite the probe log that backs your result, and publish it — and anyone is free to challenge it. It works like WCAG A/AA/AAA or a TLS version: self-assessed, publicly checkable, and governed by no one. The gift is the standard and the test; nobody needs our signature to use them.

SOC 2 and ISO require an accredited third-party auditor, carry professional liability, and attest that a company followed a set of controls. GOVENANT is a specification plus a test you run yourself, in your own environment, for free — and it answers a different question: does an agent’s "done" correspond to a verified outcome, or is it self-reported? GOVENANT complements SOC 2; it does not replace or compete with it, and it makes no claim to accredited-auditor status.

They are complementary layers of the same trust stack. Controls-based AI certifications such as AIUC-1 audit your risk posture — policies, adversarial testing, data controls — and answer "will this agent hurt us?" GOVENANT measures your operational record — verified delivery, per-duty coverage, graded predictions, continuously, from the system’s own database — and answers "is this agent actually doing its job, and can anyone prove it?" No controls framework detects performed autonomy: an agent that passes every safety check while shipping nothing leaves no trace a controls audit can see. And GOVENANT does not replace adversarial-robustness testing. Serious deployments carry both — they certify the brakes; we measure the odometer.

They are self-assessed conformance levels — Logged, Gated, Delivered + Covered, Earned — exactly the way WCAG defines A/AA/AAA. You claim a level, cite the probe log that supports it, and publish it openly. Two rules keep a claim honest: the verdict is the minimum of the load-bearing pillars (no averaging away a dead gate), and no product wears a level badge until its record earns it — including our own reference implementation, which wears none yet.

The full specification (Parts 0–16), the audit instrument, and the anti-pattern taxonomy — all under CC BY. You may implement the standard in any stack, run the instrument on your own agents, and publish your results, including as a competing implementer. That is the point: the specification is the gift.

It is the failure mode GOVENANT names and cures: AI agents that look busy — fluent plans, cheerful "done!" messages, dashboards full of motion — while nothing real ships. Once you have a word for it, you see it everywhere. GOVENANT’s three laws (Prevention, Assertion, Coverage) exist to make it impossible to hide.

The standard is free, forever. Buying nothing is a first-class path: read the spec, run the test, publish your level. For teams that want hands-on help, the steward (iii.partners) offers optional implementation support and a deeper review — but that is a service you may want, never a gate you must pass.

Yes — that is the fastest way to understand GOVENANT. Point the free self-check at your agent system and it tells you what your agents deliver versus what they merely perform. Request the self-check kit below and run it today, no strings attached.