Governova

Governova

"AI can build anything. It is us who must tell it exactly what to build, how to build it, what not to build, and who approves every decision."

-green)

On that score. Governova scores itself with the engine it ships, over all five factors at full weight, and currently scores 80 (B) — below its own Certified bar of 85. It was 100 while three of the five factors were unassessed; instrumenting them lowered it. The gap is constitutional coverage: under a fifth of applicable standards are evidenced, and undemonstrated compliance is deliberately uncounted. A governance product that grades itself generously is the one number nobody should trust.

For the exact count, run governova govscore — it is deliberately not quoted here, because it is not reproducible between a full checkout and a CI shallow clone (D-29). The headline is.

That number is checked by a test, not maintained by hand — see testreadmestatesthescoretheenginecomputes`. A stale badge on a governance tool is the same defect as a stale metric inside one.

---

Continuing work on Governova itself? Read START-HERE.md first — current state, the remaining work order, the standing rules, and the traps that cost the last engineer an hour each.

---

Governova is the world's first AI development governance platform — the constitutional layer between AI capability and enterprise trust.

→ Vision · Product · Strategy · Quickstart · Docs index

---

What it is

Governova tells AI tools exactly what to build, how to build it, what not to build, and who approves every decision — for any system, any domain, any stack, any scale. It records every AI action in an immutable audit trail and generates complete documentation for every file ever built. Four-layer architecture

670 live standards and 498 anti-patterns across 15 constitutions (the 594 locked baseline, plus ratified C0 §8 amendments and the four constitutions ratified since — C11 Requirements Engineering, C12 System Modelling, C13 Software Evolution, C14 Data Design).

Every count on this page is compiled/constitution.json, which the engine regenerates and governova compile --check gates on. Regenerate before editing: governova stats, governova validate, governova govscore. Quick start

Then, from the root of any project — one you started yesterday or one that has been running for years:

onboard writes nothing. It reads the repository, proposes a profile, and tells you plainly what it could not detect — a stack it does not recognise, a domain no code can reveal — rather than guessing and quietly narrowing what gets measured.

It runs entirely offline. No account, no server, no telemetry, no licence check — permanently, by ADR-010 §5.1, and asserted by a test that imports every command with the network stack made unimportable. Nothing you scan leaves your machine.

Governing Governova itself, or reading the corpus? Clone the repository:

New here? Start with the documentation…