Each runs standalone — its own login, its own data perimeter, its own agreement.
Register once and every demo below unlocks for fourteen days.Register for demo access
The verifiable operating record of a building.
Read-only, one-way feeds from the systems already in use
Property, finance and building feeds mapped to one record
Each fact appended tamper-evidently, externally timestamped
Cited answers and a projection that refuses when unsupported
Scoped, expiring data rooms, embeddable mark and read API
Counterparties re-derive the record offline, without us
Public standards implemented
Implemented against the published specifications. No conformance certification is held or claimed.
Every figure is computed from the verified record and shipped with its citation. The forecast refuses to answer when the history behind it is too thin — a property we consider a feature.
Sectors
Property & built environment · Finance & investment · Environmental technology
Who held what, and when — provable on demand.
Deal sheets and evidence captured into one common format
House-view rules screen, flag and route every deal
Checklist, evidence tiers and committee votes recorded
Each decision written to an append-only, tamper-evident record
Eligibility-gated investor surfaces; every denial logged
An outsider re-derives the record without trusting us
Public standards implemented
Implemented against the published specifications. No conformance certification is held or claimed.
Every score decomposes into the weighted inputs that produced it, and a human can override any component on the record. A language model drafts and extracts; it never decides, and the product runs rules-only without it.
Sectors
Finance · Property & infrastructure · AI & digital
Promotions and rewards a participant can check for themselves.
Existing commerce, loyalty and in-store systems plug in unchanged
Purchases, scans and sign-ups arrive as events
Earn, entry and drop rules turn events into balances
Results published for independent re-derivation and timestamped
Every point, prize and payout reconciles on the ledger
Member wallet, operator console, public verification page
Public standards implemented
Implemented against the published specifications. No conformance certification is held or claimed.
Decision automation with operator-set guardrails — inspectable and auditable today, built behind an interface designed to take a learned model later. We do not claim a trained model.
Sectors
Retail & hospitality · Financial services · Digital platforms & AI · Food & beverage
Custody and evidence, from the factory floor to the invoice.
One record per hand-off, or per evidence item
Match against the order, or grade by provenance tier
Append-only history, hash-chained and re-verified on screen
Scoped party portals, tier-gated rooms, public QR page
Money to the ledger; GS1 EPCIS and Peppol UBL out
Public standards implemented
Implemented against the published specifications. No conformance certification is held or claimed.
Both products score their records — settlement confidence in TraceX, evidence strength in CrestX — with deterministic logic that shows its working, rather than a black-box model.
Sectors
Smart logistics & supply chain · Agritech & food · Advanced manufacturing · Finance
Everyday actions, turned into assurance-ready evidence.
Manual, spreadsheet, document or connector entries come in
Each record carries a provenance score and verification level
Reductions and retirements recorded against measured records
Applied tonnes cancel measured tonnes; residual re-derivable
A self-verifying evidence pack, exported in an open format
An assurer re-derives the chain offline, in any language
Public standards implemented
Implemented against the published specifications. No conformance certification is held or claimed.
The evidence engine is deliberately deterministic — every score and every figure is a rule an assurer can re-run and reproduce exactly, with no model in the loop. Language models appear only in optional back-office assistants that draft copy for a human to approve, never to grade evidence.
Sectors
Environmental technology · Finance · AI & digital
Two decisions we will defend in the room. Everything shown is in the build today; where a capability is dormant until configured, it says so.
A fact is recorded
Appended, never overwritten
Chained to the one before
An edit mid-history shows
Stamped by an outside authority
RFC 3161 — not us
Checked by anyone
Offline, standard tooling, no account
No node to run, no token to hold, no chain governance to trust. Where a platform does not have this, its profile above says so.
Your question
A model reads the question
It may pick an intent, or refuse. It cannot emit a figure.
The record is computed
Deterministic, re-runnable
Answer, with its citation
A lender, an assurer or a regulator has to be able to re-run the arithmetic. A black box cannot be re-run — so it is the wrong tool for the load-bearing part.
The defensible asset is the accumulated record, not the file format.
Export schemas, open document formats and public verification endpoints — free, versioned, ours to maintain.
A counterparty or vendor integrates without buying anything or signing anything.
The verification, settlement and scoring machinery behind the interface is the part we keep.
Every verified day deepens an operating history that cannot be backdated.
The same code path drives a real timestamping authority the moment you point it at one, and labels itself a demonstration stub when you have not. Each historical anchor is re-verified by the authority that issued it — so switching a provider on cannot retroactively invalidate what came before.
Structured output constrains it at the API layer to a whitelisted intent or a refusal, its answer is re-validated server-side, and any failure falls back to deterministic code. The worst it can do is pick the wrong permitted question — it cannot fabricate a figure.
Bands come from replaying the identical method over the record’s own history and scoring it against what actually happened, with a hard refusal gate in code when the history is too thin. The system declines to forecast rather than produce confident-looking nonsense.
A cross-customer benchmark cannot emit a figure until the cohort behind it is large enough. That is a condition the aggregation enforces before it computes, not a promise in a document.
Products on irreconcilable renderers — native mobile and server-rendered web — still compose at runtime through scoped keys and signed embed tokens rather than a shared build. Each stays independently deployable, and separately licensable.
On several platforms the append-only guarantee is enforced by the database itself, which rejects an update or delete against the event chain — including one issued by an administrator, including one issued by us.
Categories, not construction. The deep-dive happens under a mutual NDA.