The Village Governance API — governance as a service

Early access · on application

AI agents now propose actions all day: send this, pay that, publish, delete, commit. The hard question was never whether the agent can act — it’s who decided it should, and what record exists of that decision afterwards.

The Village Governance API answers both. Your system submits a proposed action; you get back a verdict — allow, gate (a human must decide), or deny — and a signed receipt you can verify yourself, offline, without asking us. Behind each verdict sits a tamper-evident, customer-owned record of the whole deliberation, exportable as a self-verifying dossier.

It is an interface, deliberately narrow. What you send and what you get back is the entire surface. How a verdict is reached never crosses the wire.

The record is the product

Most audit trails record that something was approved. Boards, regulators and counsel want to see how — what AI was used, what it produced, how a person actually engaged with it, and what they decided.

The API captures that as a structured kōrero — the deliberation itself:

AI output and human reasoning are kept structurally separate, and the API never fills in the reasons for you. A record that says “a director read the AI’s summary and challenged two figures, and here is what they wrote at the time” is a different class of evidence from a tick in a box. That separation is the point.

Verify it without us

Every receipt is Ed25519-signed with your tenant’s own key, published in your tenant’s DID document. Verification takes a receipt, that public document, and about forty lines of code — no call to us, no account with us, no trust in our word. The client SDK ships the verifier; the algorithm is documented so you can re-implement it in any language. The receipt checker is a single file that needs only Node 20 or later.

Export the full record and the dossier carries the receipts, the proof chain and the deliberation together. A third party can check all of it years later, whether or not we still exist.

Integrity on your terms

By default the record is tamper-evident: a signed, append-only proof chain in which a signed entry can be checked against its signer’s key. Since 24 September 2026 each new entry also carries the hash of the one before it, so an entry removed from the middle of a chain, or swapped for another, is caught; earlier entries are unlinked, and removing the last entry is not caught by links alone. That gap is narrowed by a signed length anchor taken hourly. Chain heads are anchored to an outside RFC 3161 authority by a daily sweep, and the estate root is stamped by Certum (Poland) and by Portugal’s Cartão de Cidadão authority, with one witness signature from our own New Zealand server, which is not independent of us. Nothing here has been audited by a person; AI agents checked it against the running code. We hold that a record whose tampering can be demonstrated is stronger evidence than one that simply claims to be unchangeable — the case is made in full in Tamper-Evident Governance on this site. Tenants whose obligations require it can instead switch on write-once mode, which refuses all amendment after the fact.

Daily anchoring on Midnight: live on the main network. Mainnet anchoring was switched on on 12 November 2026: the hash of the estate root, which we sign, is written to Midnight’s main network once a day. Only the hash goes on the ledger, never content. As at 16 November 2026 the public estate-root index lists a main-network anchor for every daily root from 5 to 15 November; the roots for 5 to 11 November were also read back from the chain with the checker. The claims register says how the earlier roots came to be anchored on the evening of switch-on. The 17 anchors made from 20 October to 5 November were on Midnight’s Preview test network; that ended on 6 November. A route, GET /api/v1/governance/records/{id}/anchor-proof, is built and deployed (the claims register has its status and limits); it returns the path that joins a covered receipt to that fingerprint, and the downloadable receipt checker reads it offline (--anchor-proof and --anchor-check) and prints anchored or anchor_not_checked after its result. Only hashes go on the ledger. A receipt’s verdict never depends on the anchor, and the anchor does not change a sandbox receipt. What an anchor shows is that a value was recorded under a date key. It does not show that any record is true or complete; the entry is made with our service key, so it is our attestation, and Midnight is federated, not decentralised. It covers the estate on one of our two servers and not a tenant on a dedicated database, and a receipt sealed after a day’s root is built joins the next day’s range. Until a receipt’s day has been anchored, the route answers not_yet_anchored. The claims register keeps the detail.

Signing is configurable the same way: a single tenant key, or per-director signing, where each participant’s engagement carries an individually attributable signature.

Custody and control by design

What we do not claim

The record is tamper-evident and signed. We do not market it as tamper-proof or court-defensible — whether it satisfies a legal duty is a matter for your counsel and, where it comes to it, a court. What we give you is the strongest contemporaneous, self-verifiable record of how a decision was made. Where a customer enables the write-once toggle, the record can no longer be amended at all; that hardens the record, but not the claim — even then it is evidence of how a decision was made, not proof that any legal duty is met.

Availability

The API is built, tested and security-reviewed internally. We are opening it in stages, and we would rather stage it carefully than expose it early:

  1. Now — the full interface documentation and client SDK are public (below). You can design and build your integration today; the contract is versioned and additive-only within v1.
  2. On application — sandbox access: approved applicants receive a tenant-scoped key against a live sandbox. Applying is light — who you are, what you intend to build, and agreement to the terms; enough to keep bad actors out, no more.
  3. Production keys follow counsel-reviewed terms and a data processing agreement — no production key is issued before both exist.
Request early access

Documentation

OpenAPI specification (v1) User guide Integrator changelog Conduct Record walkthrough (shell) Conduct Record walkthrough (Node) Versioning & deprecation policy Client SDK + verifier

Pricing

The Governance API is an add-on to a Village subscription — every tenant is a Village, and the records live there, so there is no standalone API plan. Indicative pricing — a Village subscription plus a metered governance add-on, in your currency — is published at pricing page. All figures are indicative until general availability; assurance-level engagements are priced by conversation.


The Village Governance API is a product of My Digital Sovereignty Ltd (NZ), built on the governance architecture documented on this site. Questions: governance-api@mysovereignty.digital.