The Anklet Protocol

Every payment in India can travel between any two apps because UPI is a protocol, not a product. Commerce found the same shape in ONDC. Dispute resolution found it in PULSE. But the moment two people need to forma legal agreement — a couple before marriage, families across borders, a founder and a partner — they fall back into walled gardens: one platform’s database, one vendor’s process, one operator’s word that everything was done fairly.

We think agreement formation deserves open infrastructure. So we are writing down the protocol we run on — and opening it up.

What it is

Ankletis an open, IETF-style specification for forming legal agreements online, built as the Agreement Formation track of PULSE — the Beckn protocol’s legal-services adaptation. PULSE covers online dispute resolution; Anklet covers what happens before a dispute ever exists: intake, disclosure, negotiation, drafting, professional review, execution, and sealed archival — between parties who may not, and need not, trust each other or any central operator.

The specification standardizes the things that determine whether an agreement survives scrutiny years later:

• A lifecycle with teeth. Agreement state is derived from recorded facts and guarded by server-side gates — disclosure complete, conflicts reconciled, risks dispositioned, identities verified — each failure reported as a typed, actionable reason. Once signing begins, the matter is sealed: nothing changes without leaving evidence.
• Selective disclosure between adverse parties. Each party captures privately; sharing is an explicit, recorded act of consent. Safety information — the protocol’s most protected class — can never be shared, never inferred, never surfaced to the other party.
• Confidentiality at the protocol layer.Public, standard, privileged, sealed, and restricted-safety are wire-level classes with fail-closed handling — attorney–client privilege is an envelope property, not an application promise.
• Pluggable trust. Identity proofing, statutory e-signature, and privacy-preserving claim attestation (prove an income band, not a bank statement) are abstract rails any qualified provider can serve — and every provider callback is a hint that must be re-verified, never a fact.
• An audit trail built for court. Every material act lands in a tamper-evident process record that exports as a voluntariness annexure to the instrument itself, with hash-anchored document versions and an evidence seal that re-blocks execution if anything shifts.
• Rules for AI. AI may draft and guide — under disclosed use, audience-scoped context isolation, output egress control, grounded citations to the record, quality gates, and bounded, expiring authority. The enforceable skeleton of the instrument is rendered deterministically; the machine never freelances the parts a judge will read first.

Why “Anklet”

In the Silappadikaram, the great Tamil epic, Kannagi proves a fatal miscarriage of justice with a single object: her anklet, broken open before the king, containing evidence no testimony could refute. That is the standard this protocol holds itself to. An agreement formed over Anklet carries its own anklet — a tamper-evident record of who disclosed what, who consented when, who advised whom, and what exactly was signed — so that its fairness can be proven by the instrument itself, not by anyone’s word.

Why marriage agreements first

Marriage agreements are the hardest ordinary case: two people who intend to cooperate for life but are legally adverse at signing, often across borders, with asymmetric information and — sometimes — asymmetric safety. A protocol that holds there holds anywhere. Our first applicability profile specifies the marriage-agreement process end to end, with jurisdictional annexes (India first) carrying the local formalities. Just Agree is the first implementation.

How it’s governed

• The specification is Beckn-compatible and additive to PULSE, written as RFC-style drafts with normative requirement language and mandatory security and privacy analysis.
• Every change goes through Anklet Improvement Proposals — a numbered, EIP-style process. Implementation experience in, normative text out; the spec never changes without a merged AIP.
• CourtXAI is one implementation, not the protocol. We built it, we run it, and we wrote down what we learned — so that anyone can build another, and so that the people signing agreements on any conforming platform get the same guarantees we give ours.

No single entity should hold everyone’s agreements. Every act should have an audit trail. Trust should be verifiable, not assumed. That’s the protocol.

The working drafts

Published here as -00 working group drafts, in BCP 14 requirement language. Nothing is a stable standard yet.

• draft-anklet-core-00 — the core protocol
• draft-anklet-marriage-00 — applicability profile: marriage agreements
• annex-india-00 — jurisdictional annex: India
• generic workflow — non-normative end-to-end walkthrough

If you want to shape them, join the founding network.