AiGovOps Foundation
Open Source v4 · AI Governance Operations

Get to YES.
Stay at YES.
Recover to YES.

An end-to-end, open-source platform for governing AI in production — built on three products and orchestrated by a manager-agent. Every control is a gate that answers PASS or FAIL. No maybes allowed.

Try the Yes-Gate ↓ View the source ↗
The platform

Three products, one gate logic

Each product enforces the same contract — a control either passes or fails, and every failure ships a path back to YES. They share one source of gate logic in shared/gate.mjs.

Audit & proof

Beacon

Listens, identifies models and patterns, proposes heuristics, and signs verifiable ed25519 evidence receipts you can check offline.

PASS / FAIL — never MAYBE
Monitoring & drift

Lantern

Semantic diff, drift detection, change monitoring, and escalation — so a system that reached YES is caught the moment it slips.

PASS / FAIL — never MAYBE
Policy & gates

Umbrella

Policy compiler, code gates, automated enforcement, and framework profiles that turn written rules into deterministic checks.

PASS / FAIL — never MAYBE

Jeeves — the manager-agent — orchestrates all three gates and reports a single verdict. Agents do the bureaucracy; humans hold the keys.

The discipline

Three Laws of Gates

Get to YES. Every control has a deterministic pass condition. Ambiguity is a bug.
Stay at YES. Continuous monitoring re-runs the gates on every change — drift is caught, not discovered later.
Recover to YES. Every FAIL ships a mitigation checklist and an automated path back. A gate never just says no.
Live

Try the Yes-Gate

This is the real gate logic from shared/gate.mjs, running in your browser. Toggle the checks — the verdict is binary, and every failure carries its way back to YES.

    The real pipeline in Node, zero dependencies — policy, then drift, then a signed proof:

    import { orchestrate } from "@aigovops/jeeves";

orchestrate({
  umbrella: { profile: "baseline",
              payload: { model: "claude-opus-4-8", humanApproved: true } },
  lantern:  { baseline: { cost: 100 }, current: { cost: 102 }, tolerance: 0.05 },
});
// → { status: "PASS", receipt: { …ed25519… } }   signed, verifiable offline, no maybes
    Standing on proven open source

    A tiny core, production-grade backends

    The core stays small and auditable on purpose. For scale, each component maps to a mature Apache-2.0 project (10k+ stars, verified live). Define the contract once; enforce it with the strongest backend each environment allows.

    Umbrella · policy

    OPA · Casbin · Trivy

    Rego policy engine, embeddable authz, and ready-made code/dependency gates.

    Lantern · drift

    Prometheus · Great Expectations

    Time-series monitoring and declarative data expectations for drift + escalation.

    Beacon · proof

    MLflow · Sigstore

    Model system-of-record to sign over; the production form of signed receipts.

    Jeeves · orchestration

    Airflow · Argo · Keycloak

    DAG/workflow engines for the pipeline; identity to scope every view.

    Full mapping with live-verified star counts: INTEROP.md ↗ · the complete product blueprint: BLUEPRINT.md ↗ · the path to production: ROADMAP.md ↗ · the 1-click product plan: PRODUCTIZATION.md ↗ · hosted & open-core: SAAS.md ↗