Six verbs. One Bitcoin address. The protocol surface of the sovereign web.

Identity, confidentiality, legitimacy, provenance, authority, commitment — six composable protocols on Bitcoin. No custody. No token. No issuer. Every artifact verifies offline, forever.

A small Bitcoin balance, left unspent for a while, is the only credible economic signal on the open internet. OrangeCheck is the family of open protocols that turns that signal into portable, cryptographic, platform-neutral proofs — for sybil resistance, end-to-end encryption, sovereign declarations, stake-weighted polls, scoped agent delegation, and bonded commitments.

No custody. No KYC. No aggregated reputation score. No captured intermediary. Honest users pay nothing but time; attackers pay real Bitcoin opportunity cost — ruinous at scale.

Hosted reference apps · docs · sign-in: ochk.io · docs.ochk.io

Every protocol shares the same shape: a BIP-322 signature from a Bitcoin address over a canonical, human-readable message. Learn that one shape and the rest of the family pattern-matches.

The family is closed at six. OC Attest is the base layer; the other five are peers that may optionally reference an Attest proof as a stake-at-signing signal. Every protocol's specification, threat model, and reference impls live in their own repo. Everything composes — gate predicates, vote-resolved disputes, stamp-published pledges, agent-signed envelopes — but no protocol depends on another at the wire level.

Every protocol ships under the same canonical layout: README.md / PROTOCOL.md / SPEC.md / WHY.md / SECURITY.md / CHANGELOG.md / LICENSE / test-vectors/ /examples/. Specs are CC-BY-4.0; example code is MIT.

A protocol-level conformance vector in any spec repo is a (input, expected_output) JSON tuple that every reference implementation MUST pass byte-identically. A new implementation joins the family by passing every vector for the protocols it implements.

All code is MIT. Specs are CC-BY-4.0. Fork, re-implement, self-host — that's the point.

Core (one per protocol):

Integrations + middleware:

CLIs:

pip install orangecheck — sync + async client against the OrangeCheck hosted API, plus offline canonical primitives. Shares conformance vectors with @orangecheck/sdk; a byte-identity drift in either implementation breaks the other's CI.

Sybil-gate a route in five lines:

import { ocGate } from '@orangecheck/gate';

app.post(
    '/post',
    ocGate({ minSats: 100_000, minDays: 30, address: { from: 'header' } }),
    postHandler,
);

Sign in with Bitcoin end-to-end (cross-subdomain auth host):

yarn add @orangecheck/auth-client

import { OcSessionProvider, OcSignInButton, useOcSession } from '@orangecheck/auth-client';

function App() {
    return (
        <OcSessionProvider>
            <Page />
        </OcSessionProvider>
    );
}

function Page() {
    const { status, account } = useOcSession();
    return status === 'authenticated' ? <Dashboard /> : <OcSignInButton />;
}

Per-protocol quickstarts: docs.ochk.io. One-page narrative for any new reader: docs.ochk.io/getting-started/which-protocol.

Read this list twice. These are visible refusals — design constraints, not omissions.

• Not custodial. No protocol in the family ever holds sats. Funds never move. Bonds and stake are economic signals, not balances under contract control.
• Not slashing-based. No actor — protocol, counterparty, voter, oracle — can seize bonded sats. Ever. Teeth come from public records, not lockup.
• Not aggregated reputation. Every protocol ships raw metrics (sats_bonded, days_unspent, kept_count, broken_count) and refuses to ship a canonical score. Platforms compose policy from raw queries.
• Not proof of personhood. One address can be one human, ten, or zero. For one-human-one-vote, look at World ID or BrightID.
• Not legal. Produces evidence admissible in jurisdictions that recognise digital signatures. Does not produce judgements. Does not replace courts.
• Not a smart-contract platform. No VM, no gas, no on-chain state writes. Everything is off-chain canonical envelopes plus reads of public Bitcoin and Nostr state.

If your design only works on Ed25519 (no economic layer, no non-malleable time, no enumerable identity namespace), it doesn't belong in this family.

• Bug or spec question? Open an issue on the relevant oc-*-protocol repo (specification), the relevant oc-*-web repo (reference app), or oc-packages (publishable code).
• Writing a new implementation (Rust, Swift, Go, Elixir, anything)? Start from the conformance vectors in the relevant oc-*-protocol repo. If your implementation passes the vectors byte-identically, it agrees with the reference TypeScript and Python implementations exactly.
• Building something on top? Get in touch — happy to wire up gate predicates, pledge resolution mechanisms, or custom integrations.
• Licensing. Specs CC-BY-4.0 · Code MIT · No CLA.