The problem we set out to solve
HTS misclassification is one of the most common and costly compliance failures in U.S. trade. A wrong classification means the wrong duty rate, the wrong trade remedy exposure, and a CBP audit finding if the importer cannot document their reasoning. It costs money. It delays shipments. In serious cases, it results in penalties.
The tools available to address this are, almost universally, inadequate. Legacy ERP systems carry forward classifications that haven't been reviewed since the tariff schedule changed. Spreadsheets diverge the moment two people edit them. General-purpose AI tools produce different answers to the same question depending on how it's phrased, then provide no documentation of their reasoning.
"An HTS code without reasoning is not a classification. It is a guess."
The professionals who do this work know the problem cold. They've been living with it for years. They don't need to be taught what an HTS code is. They need a tool that treats the problem with the same seriousness they do — and produces a result they can actually stand behind.
That is what Kanon is for.
The name
The name is deliberate. It carries two meanings at once — and the tension between them is the entire point.
Kanon
/ ˈkeɪ.nən / — the record
In literature, religion, and law, the canon is the definitive, accepted body of work — the official version that everyone agrees on. In Kanon, it means exactly that: one golden record, one source of truth, one version of the product classification that the entire organization trusts. It carries authority without arrogance. It is simply the truth.
Cannon
/ ˈkæn.ən / — the weapon
A cannon is a precision instrument of overwhelming force. When it fires, everyone knows. Kanon borrows this energy — not as aggression, but as celebration. Every time a product is correctly classified and the reasoning is locked, the cannon fires. It is the product's signature moment of triumph. Your data is right. Now let it speak.
Two words. One sound. One brand. The double entendre is not a gimmick — it is the product's entire philosophy made audible: establish the truth, then let the cannon speak.
The pirate thread
Kanon carries a third layer beneath the surface — the world of the Caribbean, the Age of Sail, and the pirate as a figure of radical competence. This is not a costume. It is a positioning — and it is earned.
Pirates were not ignorant of maritime law. They were intimately familiar with every trade route, every manifest, every tariff, every port authority, and every rule governing the movement of goods across open water. They understood the system more completely than anyone operating inside it. They simply refused to accept that the system was the best version of itself. They charted their own course.
"It's better to be a pirate than to join the navy."
— Steve Jobs, on building the original Mac team
Kanon's first customers are the pirates of trade compliance. The ones with war stories about legacy ERPs, disconnected data sources, and spreadsheets that haven't been trusted since 2019. The ones who know exactly what the right answer looks like and are tired of systems that can't deliver it. They don't want to join the navy. They want a cannon.
The Tropic Dusk color palette is not arbitrary. Deep teal, turquoise, warm gold, shell white — these are the colors of the Caribbean at golden hour. The place where trade routes converged, where fortunes were made and lost, where the rules were written and rewritten on open water. The palette connects the product to the world it inhabits without ever having to say so directly.
Our mission
One authoritative answer. Every time.
Every U.S. importer deserves to know exactly what code their product belongs under — and to have the legal reasoning that proves it.
-
Determinism is not optional
The same product attributes, evaluated against the same corpus, must always produce the same HTS code. This is not a feature — it is the baseline requirement for any tool that claims to support audit defense.
-
Documentation is the product
An HTS code without a reasoning trace is not a classification. The Classification Support Package — the immutable record of every GRI decision — is what separates a defensible classification from a guess.
-
The corpus is the moat
Kanon's classification engine is only as good as the corpus it runs against. Every HTSUS chapter note and section note, parsed into structured condition trees. The corpus is why Kanon's answers can be trusted.
How we work
Four principles that govern every decision we make about the product.
The LLM never writes alone
Every LLM-generated classification follows a propose-then-confirm pattern. The system proposes. The user confirms. No LLM output is written to a classification record without human review. This is not a limitation — it is the audit trail.
One active record per product
At any given time, each product has exactly one active classification record. Reclassifications create new records — they do not overwrite old ones. The history is permanent. The audit trail is complete.
Corpus versioning is non-negotiable
Every classification is pinned to the exact HTSUS corpus version that was active at classification time. When the USITC publishes a new edition, Kanon versions it. Your old records remain valid against the corpus they were made against.
Transparency in confidence
Kanon reports its confidence level on every classification: Confident, Reasoned, Split, or Insufficient. A Reasoned answer is still a good answer. An Insufficient answer means you need to provide more product information before Kanon will commit.
Ready to chart your own course?
Classify any product under the HTSUS for free. The full corpus. Every decision documented. A Classification Support Package that holds up on audit day.