Here is a test you can run on any industry in one sentence. If the question "what did we know at that exact moment, who did it, under what authority — and can I prove it to a third party?" is business-critical, then you need a system of record that can answer it. If it is not, a normal database is fine.
Kivi is an event-ledger database built for the first case. Only events are durable; every table, graph or view is compiled from them on demand; and no answer comes back without the record numbers that produced it. An untraced answer literally cannot be constructed.
So which industries live in the first case? Here is where we see Kivi fitting, strongest first.
Insurance
A claim payment ends up in court. The question is never "what does the file say now" — it is "what did the file say the moment you authorized the payment." With as-of, Kivi returns any past state provably: the file at FNOL, the rationale when a reserve changed, a point-in-time snapshot for Solvency II. Every answer carries its receipt. And on the fraud (SIU) side, the traced graph shows who is connected to a claimant within N hops — with a dated proof for every link.
Banking & capital markets
Order books, reconciliation, KYC/AML approvals — all need a record that cannot be deleted, cannot be tampered with, and is timestamped. In Kivi every event is sealed into a hash chain; change one byte after the fact and the chain breaks, and Kivi names the flaw by address rather than hiding it. "Who approved this, on what basis" lives inside the record (actor, purpose). SOX evidence, MiFID II order trails, AML history — one dated place.
Healthcare
A patient asks who saw their data, when, and under what consent. Each access is an event; the purpose field carries the legal basis for processing. When consent is withdrawn, a per-record key is destroyed — crypto-erase gives the right to be forgotten without breaking the chain. as-of proves which result was visible at the moment of a clinical decision; 21 CFR Part 11 electronic records and signatures sit on the same ground.
Public sector, forensic & regulatory
Evidence has one question in court: has it changed since it was seized? Registries, chain of custody and inspection records all need a tamper-evident, timestamped record a third party can verify. Kivi traces custody as a graph, anchors head hashes to a qualified timestamp authority (RFC-3161 / eIDAS), and ships an independent, single-file verifier — so an auditor can validate the record offline, with no server and no access to the closed core.
Supply chain & logistics
A product is recalled and you have hours, not days: where did this batch come from and where did it go. Every movement is an immutable event; the graph returns every actor within a few hops of a batch, each step dated. Across organizations, federation and mutual witnessing let supplier and buyer verify the same record without trusting a single middle.
Energy & ESG
Emissions and grid figures are becoming audited numbers. "We reported it" is giving way to "we did not alter it, and we can prove it." Each measurement is a sealed event; a later correction happens only through a traced compensation event, never a silent overwrite. as-of proves what a figure was at a quarter's close.
Manufacturing, aviation & automotive
Traceability here is a safety matter. Part history, maintenance logs and quality events must be immutable and dated. as-of proves what was known about a part at the moment maintenance was signed off, and broken or partial data is reported by address instead of silently skipped.
Legal & compliance technology
In a dispute: which version of a contract was in force on which date, and who changed it. Every change is a traced event with actor and purpose; as-of reconstructs the text on any date; and the independent verifier lets even the opposing party validate the record with no server.
And one that is arriving: AI data lineage
With the EU AI Act, data lineage is shifting from good practice to obligation. Which data, whose consent, which model version. In Kivi data movements and training steps are traced events; as-of proves what a dataset was when a model version was trained; and if consent is withdrawn for a record, crypto-erase applies.
Where Kivi is not the right tool
We would rather tell you this up front. High-volume, low-value CRUD (a shopping cart, a social feed) is better on a classic OLTP database. Multi-statement ACID transactions, heavy JOINs and referential constraints belong in a relational database — put one next to Kivi and keep Kivi as the traced system of record. Data that must be freely mutable or deletable fights an append-only design. And heavy graph analytics want a real graph engine; Kivi offers bounded, traced traversal, not PageRank.
The common thread
Every sector above shares one question: who did what, when, under what authority; what was visible at the time; and can I prove it to a third party. When that is critical, Kivi fits. When it is not, it does not — and that is a fine answer too.
Honest status: Kivi's integrity claims are backed by in-repo proofs — a six-language conformance exam, fault-injecting test servers, seeded tamper traps, an offline whole-dataset audit — i.e. our evidence, not a third party's. It is not yet independently security-audited and has no production mileage. We would rather you know that before you build on it.
- Docker Hub — try it in one command: hub.docker.com/r/iwasoftcom/kivi
- GitHub — docs in 8 languages: github.com/iwasoftcom/kivi
- Product page: iwasoft.com/products/kivi