Verifiable Decision Infrastructure

Rules-as-code that can be proven.

What is this

QuantZK lets your software prove it followed a declared, machine-checkable rule set — without exposing private inputs. Issued evidence remains independently verifiable offline using its pinned verification material. Fresh revocation or governance status may require current trust data.

Install

Offline verifier, SDK, and QZR-1 conformance ship in the public extract ayitsomar/quantzk-verify (not yet published as standalone public npm packages). Full issuer stack remains in the monorepo.

Clone (public) 
git clone https://github.com/ayitsomar/quantzk-verify.git
cd quantzk-verify
npm install --legacy-peer-deps
JavaScript 
# workspace packages
@quantzk/sdk
@quantzk/vdi-verifier
@quantzk/verify
Python 
# local package (monorepo)
cd packages/vdi-verifier-python
pip install -e .
Go 
# module under monorepo
packages/vdi-verifier-go

Hello World

import { attest, verify } from '@quantzk/attest';

const proof = await attest('fair-lending-v1', {
  inputs: { credit_score: 720, income: 85000 },
  outcome: 'approved'
});

const { valid } = await verify(proof);  // true

Three lines. The manifest defines the rules. The proof is automatic. Verification is independent.

What it proves

What the verifier learns What stays private
Encoded constraints in the declared manifest evaluated successfully Input data (credit score, income, etc.)
Included cryptographic proof and issuer signature verify under pinned material Model weights and algorithm
Authority / revocation checks only when the chosen profile requires them How a human reviewer interpreted the regulation
Proof is cryptographically valid for the embedded circuit / vkey (e.g. Groth16) Any internal state

Available Manifests

Manifest Regulation What it proves
fair-lending-v1 Fair Lending Act (template) Encoded controls mapped to the referenced fair-lending profile were satisfied — not a legal determination of non-discrimination
eu-ai-act-v1 EU AI Act (template) Encoded controls mapped to selected EU AI Act requirements evaluated successfully — not a certification that the system meets the Act
hipaa-v1 HIPAA (template) Encoded handling constraints mapped to the referenced HIPAA control profile were satisfied — not a finding of HIPAA compliance
soc2-access-v1 SOC 2 CC6 (template) Encoded access-control checks mapped to the referenced CC6 profile evaluated successfully — not a SOC 2 opinion
nist-ai-rmf-v1 NIST AI 100-1 (template) Encoded controls mapped to the referenced NIST AI RMF profile evaluated successfully
model-safety-v1 Internal / NIST 600-1 (template) Encoded output constraints in the declared safety profile evaluated successfully

Documentation