This document describes the architecture, data flow, attestation mechanics, and verification procedures of the four systems in the KeepOne ecosystem. Intended for technical reviewers at licensed institutions, family-office due-diligence teams, and developers.
The credibility of a trading track record has long rested on two kinds of evidence: platform screenshots and third-party endorsement. Both have structural weaknesses.
KeepOne converts "this record existed at this moment and has not been altered" from a question of trust into a cryptographic question. At the end of each settlement period, the digest of that period's data is written to a smart contract on Base.
Key design: only digests go on-chain, never raw data. No account details, strategy parameters, or position-level information ever leaves the institution. The chain proves that a dataset existed at a point in time and is unaltered — it does not publish the dataset.
The engine connects to four categories of sources through read-only interfaces. Credentials are constrained at creation time and carry no order, transfer, or withdrawal rights.
| Source | Access method | Retrieved |
|---|---|---|
| MetaTrader 5 | Investor password | Historical fills, positions, equity curve, net value |
| Interactive Brokers | Read-only API (TWS / Gateway) | Executions, net liquidation value, unrealized PnL |
| Crypto exchanges | Read-only API key (trading and withdrawal disabled) | Trade history, balance snapshots |
| Third-party evaluation systems | Platform-provided read-only export | Full trade ledger for the evaluation period |
One digest is produced per account per settlement period (UTC day):
PROOF_HASH_SALT, never published) is appendedThe salt defends against dictionary attacks on low-entropy data: even with the digest in hand, an attacker cannot brute-force plausible trade combinations to reconstruct the original records.
A day is sealed the moment its digest is anchored. Afterwards:
409 DAY_CLOSED with the sealed proof identifier and both equity values00001.hk is an analytics terminal. It executes no trades and custodies no funds. What distinguishes it is that both its inputs and outputs are verifiable.
Cross-market pricing and historical behaviour data come from records already anchored by base.keepone.dev rather than being collected independently. The inputs to any conclusion can therefore be reproduced by a third party.
Every AI-generated risk alert and analytical report is hashed with SHA-256 and anchored on-chain at creation time. Users can then establish whether a given report existed at a given moment and whether it was altered afterwards.
What this solves: if an AI model's predictions are only cited after the fact, there is no way to rule out selective memory of the correct calls. Anchoring outputs turns hit rate into a computable fact rather than a curated narrative.
Trader performance during simulated evaluation is continuously verified by the attestation engine, forming a tamper-proof on-chain track record.
A trader wants to demonstrate competence to an asset manager without handing over strategy parameters. Traditional due diligence forces a choice between the two: disclose everything, or prove nothing.
Zero-knowledge proofs let a trader prove a proposition — "my maximum drawdown stayed below 10% over twelve consecutive months" — without revealing a single position. The institution gets the risk evidence it needs; the trader keeps the strategy boundary intact.
| Layer | Content | Verifiability |
|---|---|---|
| Raw layer | Full trade record from the evaluation period | Visible only to the holder and authorised parties |
| Digest layer | SHA-256 fingerprint of the record | Public on-chain; proves existence |
| Claim layer | ZKP-proven risk metrics | Public on-chain; proves truth of the claim |
The problem with financial content is not whether it is wrong. It is whether, after it turns out to be wrong, you can still see what it originally said.
When a creator publishes market research or a data visualisation, the platform computes a content fingerprint and anchors it to Base. Readers can verify whether that chart existed at a point in time and whether it changed afterwards.
The mechanism does not judge correctness. It guarantees integrity and ordering. For financial information that is often more important than correctness, because post-hoc edits, deleted unfavourable conclusions, and swapped chart data are the most common forms of contamination in the information ecosystem.
This section describes how a third party independently verifies a record without trusting KeepOne. No call to our API is required at any point.
Note: the salt is not published on-chain and must be obtained from the data holder. Verification therefore requires their cooperation — but they cannot unilaterally deny a record, because the digest is already on-chain and they cannot produce a different dataset with the same digest.
| Role | Permissions | Constraints |
|---|---|---|
| Platform (KeepOne) | Read trade records within the authorised scope | No ordering, transfer, or withdrawal rights |
| Data source (broker / exchange) | Provide read-only interface | Boundary fixed at credential level, not escalatable |
| On-chain contract | Accept digests, serve queries | Receives no raw data and no funds |
| Third-party verifier | Read on-chain digests | Read-only; cannot write |
If the salt were on-chain, an attacker could brute-force common trade combinations and reverse the original records. Keeping the salt with the service and the data holder leaves only the digest public; both are needed to complete verification.
This section states what the system does not do, to prevent misinterpretation.
Technical note: the systems described here are under continuous development. Interface details, parameters, and deployment addresses are governed by the runtime state of base.keepone.dev.