05

Governance

Built to be examined.

Every payment is recorded. Every decision can be replayed. The ledger is the evidence a supervisor, an auditor, and a correspondent bank can sit with.

Ledger

The book is the evidence.

The design writes every posting to the Cush Blockchain, an append only ledger hash linked with BLAKE3. Multi-entity books across companies, countries, and currencies. Agents follow rules the institution writes.

Permanent payment record
Every posting is hashed into an immutable BLAKE3 ledger. The book is the evidence.
Intelligence inside mandate
Agents operate only inside policy you authorise. Decisions are explainable and replayable for audit and supervisors.
Books across companies
Multi-entity isolation across the UK, Singapore, the United States, and holding structures. One control plane.
Supervisory readiness
Architecture prepared for PRA, MAS, and peer sandbox programmes, and for multi-jurisdiction licensing. Prepared is not the same as granted.

Supervisory access

Give the examiner a view.

Cush Core is designed so a supervisor can read the same book the institution posts to. Data on the Cush Blockchain is encrypted, and a read only examiner viewer is planned to show the supervisor the records the institution grants. The viewer can read. It cannot post.

QCan our supervisor read the ledger directly?+

That is the design. A read only examiner viewer is planned, so the institution can give its supervisor access to the records it grants. Scope, location and terms are agreed between the institution and its supervisor. The viewer never carries the right to post.

One chain

Money, risk and incidents. One record.

Every posting is hashed into the BLAKE3 ledger, with the actor, the time and the outcome. The design carries the same tamper evident chain to risk decisions and operational incidents, so the payment, the score behind it and anything that went wrong are read as one history. Internal audit and the supervisor examine one record, not three systems.

QAre risk decisions and incidents on the same chain as postings?+

That is the direction of the design. The design records postings and contract executions on the Cush Blockchain, hash linked with BLAKE3. Carrying the same chain to risk decisions and incidents is part of the build, and the briefing sets out what is complete.

Object centric ledger

Every customer, account and contract keeps its own chain.

The Cush Blockchain is object centric: every customer, account and contract has its own chain of events, hash linked with BLAKE3. Records are append only and never modified. Chains are spread across shards backed by PostgreSQL, a spare node takes over a failed node's shards in one to two seconds, and real time projections give current state without replaying the full chain.

QWhat happens to the ledger if a node fails?+

Chains are spread across shards on separate nodes, each backed by PostgreSQL. When a node stops responding, the coordinators remove it from the registry and a spare node takes over its shards in one to two seconds. A failed node never resumes ownership. It can only return as a spare.

The Cush Blockchain

A blockchain built for banks, not for public consensus.

The Cush Blockchain is the ledger inside Cush Core, licensable with the platform. It is designed for institutions that run their own infrastructure and answer to their own supervisors.

  1. 01

    Not full consensus.

    There is no single global chain and no general purpose consensus. Every customer, account and contract keeps its own chain, so a posting never waits on the rest of the network.

  2. 02

    Permissioned by design.

    Nodes are known and registered, not open to anyone. Each chain is tamper evident, hash linked with BLAKE3.

  3. 03

    Built on PostgreSQL.

    Shards are backed by PostgreSQL. When a node stops responding, a spare takes over its shards automatically in one to two seconds.

  4. 04

    Every write is attributable.

    Every commit is backed by a transaction signature, and nodes check each message so that a removed or out of date node cannot write.

  5. 05

    Encrypted and examinable.

    Data on the chain is encrypted. A read only viewer for examiners is planned, so an examiner can see the records the institution grants.

  6. 06

    Models do not post.

    Models explain and suggest. In the Cush Blockchain design, postings are made by deterministic smart contracts, inside the permissions the institution sets.

Lineage

Controls the officer already knows.

Designed for KYC, AML, PSD2, and DORA-aligned controls. Architecture prepared for PRA, MAS, and other sandbox programmes, and for licences in more than one country. Preparation is a design claim. It is not a licence already granted.

Direct CHAPS lineage: the founder led the first non-bank PSP onboarded by the Bank of England as a Direct CHAPS member.

Multi-jurisdiction books: UK Ltd, Singapore Pte, and US Inc. Isolation on one control plane, not three vendor estates.

Replay for the officer: risk scores, rail choice, and postings are a single trace. Internal audit does not reconstruct the payment from three systems.

Compliance by design

Compliance is a property of the platform.

Strong customer authentication, field level data minimisation, PEP and sanctions screening, risk grading, suspicious activity flagging, management information exports and a full change history are built into Cush Core. Together they are built to support evidence for supervisory review.

Authentication

Strong proof for everyone who signs in.

Customers, operators and partners all authenticate through one high assurance framework: FIDO2 passkeys, authenticator app codes (TOTP), device biometrics, one time codes and passwords hashed with Argon2id. Step up authentication asks for stronger proof automatically on high value operations.

QDo operators sign in the same way as customers?+

Yes. One framework serves customers, operators and partners, with passkeys, authenticator apps, biometrics and step up checks for high value operations.

Defence in depth

Security at every layer.

TLS on every connection, including calls between internal services. Short lived sessions bound to the device, with automatic revocation on suspicious activity. Authorisation enforced on the server, sensitive fields filtered before they reach the wire, and immutable, tamper evident logs of every access and change.

Access control

Who can see what, down to the field.

Every request passes three layers: who you are, what you may do, and which fields you may see. Sensitive fields are filtered on the server before the response leaves. People act as themselves, as an agent, or with delegated authority such as a power of attorney, and every authorisation decision is written to the audit log. No silent grants.

QCan a power of attorney act for a customer?+

Yes. The representation model lets a person act as themselves, as an agent, or with delegated authority such as a power of attorney. Each decision is logged.

Observability

Nothing runs unobserved.

Every request and event is logged in structured form, searchable by customer, transaction or component, with retention set to the institution's policy. Distributed traces follow a request across every service, so a customer complaint can be tied to the exact call chain. Dashboards and alerts sit beside management information, and every runtime configuration change records who changed what and when.

QCan we trace a complaint to what the system actually did?+

Yes. Distributed tracing links a customer complaint to the specific chain of calls behind it.

Examination

What a CIO will ask.

01Must we replace the incumbent core?+

No. Sidecar coexistence is the default path. Run Cush Core beside the existing book for a product line, a geography, or correspondent flows. Replatform only when the evidence justifies it.

02Who holds the system of record?+

The institution. The design writes postings to an append only ledger the institution holds, under its licence. Cush Core is not a multi-tenant SaaS interposed between you and the books.

03How do agents remain inside mandate?+

You authorise the policy. Agents score, route, reconcile, and escalate inside those rules. Nothing material occurs that cannot be replayed for internal audit or a supervisor.

04Which rails are in scope?+

Inbound: Faster Payments, SEPA, Fedwire / ACH, card acquiring, SWIFT. Outbound: CHAPS, TARGET2, SWIFT, and local instant schemes. Routing is governed by cost, success rate, and mandate.

05Can a supervisor replay a payment end to end?+

Yes. Posting, risk score, rail selection, and scheme acknowledgement form a single trace. That is the purpose of an examinable core.

06How is the platform commercialised?+

Licence economics on a per-customer basis, approximately one dollar per month, instead of a stack of vendor licences for ledger, hub, cards, and overlay. That figure is a commercial model, not reported revenue. Exact terms are set in the briefing.

Principals

Names on the memorandum.

Request a briefing

Matthew Ekow Folson

Founder and Group CEO

Sole founder. Twenty-five years in banking technology and payments. Former HSBC (EFOSS Board Chair), Metro Bank (PSD2), and Orwell Group. Led the first non-bank PSP onboarded by the Bank of England as a Direct CHAPS member.

Jose Luis Caldeira

Chief Technology Officer

More than twenty years building banking platforms in London. Former CTO of QOrbital and head of architecture at Metro Bank and TSB. At Orwell Group he delivered the first Faster Payments and CHAPS certification in the UK for a non-bank, and at Lloyds Banking Group he built the UK's first enterprise Kafka streaming platform. Author of the house thesis on the vertically integrated bank.

Paul Singh

MLRO and Chief Compliance Officer

Registered MLRO. Former Bank of England / PRA supervisory background. Previously MLRO for Revolut Payments and Crypto. Clears material external claims and licensing conversations.