GOVENANT — govern + covenant: governed autonomy, bound as a covenant. The open standard for provable, governed AI agents: actions prevented by construction, outcomes verified in the record, silence detectable per duty, and autonomy earned on evidence.
Website: https://govenantstandard.org · Start here (plain English): plain-english/ · The standard: standard/00-STANDARD.md
The public home of the GOVENANT Standard: the normative specification (Parts 0–16), the portable schema (standard/SCHEMA.sql), the naming/usage rules (standard/NAMING.md), and a plain-English explainer. This content is mirrored from the steward's working repository; issues and discussions are welcome here.
- Prevention — rules are code paths and constraints, never prompt text.
- Assertion — "done" means a verified outcome exists in the substrate.
- Coverage — every responsibility maps to a duty; silence is detectable.
GOVENANT-1 Logged → GOVENANT-2 Gated → GOVENANT-3 Delivered + Covered → GOVENANT-4 Earned. The verdict is the MIN of the load-bearing pillars; every certificate cites its probe log.
Fielder, S. (2026). Performed Autonomy: Motion Without Delivery in an Organization-Centered Multi-Agent System on an LLM Substrate (Version 1.0). Zenodo. https://doi.org/10.5281/zenodo.21440225
The peer-targeted paper: the formalism, the three audits, the named failure mode, and the cure — published open access (CC BY 4.0).
Authored by Scott Fielder, who coined "performed autonomy" and built the standard's reference implementation — then published its failed audits. Every rule was earned from an audit finding, not invented. GOVENANT is a production standard implementing the OCMAS formalism (Ferber & Gutknecht) on LLM substrates. Stewarded by iii.partners, its first implementer — and bound by The Covenant: the standard stays open, forever.
CC BY 4.0 — free to use, implement, and redistribute with attribution, including by competing implementers. That's the point.