About ArthaFlow Global
We don't publish a number we can't trace.
Export decisions are made on figures — a duty rate, a rebate, a landed cost. A wrong one is expensive and invisible, because a wrong duty rate looks exactly like a right one. Most trade tools answer every question. We answer the ones we can source, and say plainly where the evidence runs out.
Never invented
An HS code. The model picks from the real nomenclature or it fails.
Never averaged
Two sources that disagree stay disagreeing, and we say which is which.
Never implied
A rate is dated. If we last checked it in June, the page says June.
Where the answers come from
Eight datasets, each built from its primary source.
Not scraped from a competitor, not bought from an aggregator, not asked of a language model. Each is parsed from the notification or schedule that publishes it, by a script in the repository, and re-checked against published totals before it ships.
Indian tariff lines, with export policy
12,087 Free · 130 Restricted · 93 Prohibited
DGFT ITC(HS) 2022 Schedule 2
12,310
India's export values, per line, per year
Seven financial years. Totals reconcile to India's published book.
DGCIS, Department of Commerce
12,402
International nomenclature
Vendored, so a resolved code is guaranteed to exist.
UN Comtrade HS 2022 reference
6,939
RoDTEP rebate rates
Amendment chain replayed, reproducing the official counts.
DGFT Appendix 4R
10,610
Duty Drawback rates
Across 1,014 headings.
CBIC Notification 77/2023-Cus (N.T.)
2,123
Destinations priced
Every reporter code verified against a live query.
UNCTAD TRAINS via World Bank WITS
82
Import VAT and GST rates
Single-sourced entries say so in their own note.
VATupdate and PwC, cross-checked
67
Trade agreements
Status tracked separately — only in-force can be claimed.
DGFT and treaty texts
16
The arithmetic built on top of these carries 134+ assertions, each pinned to a value computed by hand. They are checked by deliberately breaking the calculation: if charging US duty on the wrong basis, or compounding VAT on the wrong value, does not fail the suite, the suite is not doing its job.
That standard was set by getting it wrong once. An early importer read DGCIS's comma-formatted values inside a bare exception handler and silently turned each one into a zero; the national total came out 13% light and looked entirely plausible. The test meant to catch it accepted any figure between $250bn and $600bn, so it passed. Both the parser and the test were rewritten.
Who builds it
Two founders, and one standard applied to both halves.
The conviction the company is built on is narrow and unfashionable: a trade tool earns trust by what it declines to answer. Every refusal on this page cost a feature a competitor ships without hesitation, and each is recorded in the codebase with the reasoning beside it — because a decision that is not written down gets quietly reversed by whoever opens the file next.
Sarthak Wage
Founder & CEO · Data layer, tariff engine, product
Builds the parts of ArthaFlow that have to be right rather than merely convincing: the tariff datasets and the scripts that construct them, the classification engine, and the arithmetic a manufacturer prices a shipment on.
- Previously
- CTO of a 3D design studio — raised multiple funding rounds, revenue to seven figures
- Studied
- K. J. Somaiya College of Engineering, Vidyavihar
- Builds
- The datasets, the tariff engine, and the landed-cost arithmetic
Sameer Morya
Co-Founder & CTO · Export Saathi, AI architecture
Owns the layer a manufacturer actually talks to. Export Saathi turns a plain question about someone's own goods into a sequence of queries against real trade data — and, harder, into an answer that says where the evidence stops.
- Previously
- SeekMyCOURSE
- Route in
- Strategy and operations, into AI architecture
- Builds
- Export Saathi and the agent layer above the data
What we refused to ship
Three questions the product will not answer.
Each of these has an answer a competitor is comfortable giving. We looked for the source, could not find one that holds, and shipped the refusal instead. This is the part of the product that is hardest to copy, because copying it means giving something up.
What Duty Drawback rate applies to my leather bags?
What we could say
2.4%
What we say
The rate is somewhere between 1.5% and 3.5%. Your broker decides which.
The Drawback Schedule carries its own numbering, following the Customs Tariff only to four digits and then subdividing on its own terms. Not one of its 340 eight-digit codes is a valid ITC-HS line. Heading 6302 spans five items at different rates, none of which map onto a tariff line, so returning one would be inventing a join that does not exist. Where a heading carries a single rate — 781 of them do — we answer precisely.
What will my buyer pay under the India–UAE trade agreement?
What we could say
0%
What we say
CEPA is in force and your buyer likely pays less. We don't hold the rate.
No reachable source publishes preferential rates. WITS TRAINS serves none through its SDMX endpoint — verified against USA–Mexico under USMCA, which would certainly carry data if the endpoint served preferential lines at all. India's own CEPA schedules sit in annexes behind dead links. A guessed preferential rate is the worst number this product could produce, because it is precisely the one a manufacturer would price a shipment on.
What RoDTEP rate will I actually receive?
What we could say
1.7%
What we say
The notified rate is 3.4%. DGFT currently limits benefits to 50% of it.
Halving the notified rate would produce one clean number, and it would be wrong the moment the limitation changes — it has already been extended once. So the notified rate and the limitation are stored and shown as two separate facts, and the product never multiplies them together. A single figure here would look more authoritative and rot silently.
Check it yourself
Don't take the page's word for it.
Everything above is inspectable. The tools run without an account, and the repository carries the builders, the tests and the reasoning.