Pricing

No. The official classifier at tasnif.soliq.uz is the only source, and we use its official export too. The portal is convenient for finding one code. Tasnivo checks the whole catalog at once and shows which rows need attention.

For uncertain rows, Google Gemini refines the match: it receives the product name and the unit of measure. For a bank line, Anthropic Claude may suggest the operation type: it receives the payment purpose and the counterparty name, with long numbers hidden in both. Either works only when it is configured. Price, amount, barcode, account number and TIN are never sent to any service. If you prefer, we check your catalog without any external service.

Within one working day of receiving the file, we send the list of risky items to the phone or Telegram you gave. No sign-up needed.

Yes. Russian and Uzbek headers are recognized automatically. Product names may be in Russian, Uzbek Latin or Uzbek Cyrillic: the classifier is loaded in three languages, and names are transliterated from one script to another before matching.

No. There is no document that fixes the one right MXIK code for a product. The system suggests a code and shows its match level and its place in the classifier. You make the decision, and it is written to the decision history.

From approved records you get a file: kl_to_1c.txt for a bank statement, a goods receipt for invoices, a nomenclature Excel file for products. You load them through 1C’s own import window. The files have not yet been tried on a real 1C database: check the first file in a test database. The direct connection to 1C (OData) is in trial mode too. In the app: the “1C import and history” section.

Excel downloaded from the bank (.xlsx or .xls), CSV or TXT, a ZIP of these, or the bank’s own kl_to_1c.txt. If the layout is not recognized, you pick the columns once and the next file with the same headers is read automatically. This path is in trial mode: it has not been confirmed on real bank files yet, so review the result. In the app: the “Bank statements” section.

You upload an archive (.zip), JSON or ESF Excel file downloaded from Didox, or fetch new incoming invoices with your Didox token. Both paths are in trial mode: the structure was checked against Didox’s official documentation and a real export archive, but no real document has been fetched through the Didox API yet, so you review every invoice. Your Didox token is used only to fetch incoming invoices: it is stored encrypted, never shown again and never written to logs. In the app: the “Didox invoices” section.

In the catalog audit: The file is kept on our server, so the corrected Excel can be returned in its original layout. The analysis reads four columns: product name, barcode, MXIK code and unit of measure. To refine a suggestion, only the product name and the unit of measure may be sent to an external AI service. Price, barcode, quantity and every other column stay on the server. In documents: Bank statements, Didox invoices and product lists are also kept on our server. If external AI is on, a bank line sends only the payment purpose, the counterparty name (long numbers hidden in both) and the direction of the money: never the account number, TIN or amount. From product and invoice lines, as in the catalog audit, only the name and the unit of measure may be sent. To learn the package code and the code’s status, only the MXIK code is sent to the official tasnif.soliq.uz. A file sent for the free audit is deleted after 30 days, or at once if you ask. A catalog you upload to the app you delete yourself at any time, together with its findings and decision records. More: privacy policy (in Uzbek).