Privacy
This site is a measuring instrument, and it treats visitors the way it treats data: it keeps only what it can justify.
What we do not collect
No accounts, no API keys, no cookies set by us, no advertising or tracking pixels, no fingerprinting, no sale or sharing of any visitor data. The MCP endpoint and every dataset are free and anonymous by design.
If you upload a customs entry
What you send is a filed customs document, so it is treated differently from a page view.
We keep the file itself, and the whole of it. The PDF or spreadsheet you upload is stored on our own server, and everything printed on that form is in it — including parts we never read into the database, such as the bill of lading. From it we read and store the importer of record and its number, the entry number and type, the ultimate consignee, both address blocks, the manufacturer ID, the port, the dates, the origin, the totals and fees, and every line’s tariff code, value, duty, weight and quantity. The filename is kept too, and a filename often names the importer by itself.
The report at your private link carries less than the record we hold. It has your importer name, importer-of-record (IOR) number, entry number, dates, origin, totals and tariff lines. It deliberately does not carry the port code, your own goods descriptions, the consignee, the addresses or the manufacturer ID.
It is read here and goes nowhere else. A 7501 is parsed on this server in about a twentieth of a second, by our own reader. Until 2026-08-29 that extraction ran through Google’s Gemini; that path is gone and nothing is sent to a third party. Your upload is never web-served, never sold, never shared, and never used to build anything that identifies an importer publicly.
An importer-of-record number can lawfully be a Social Security Number for a sole proprietor (19 CFR 24.5). We do not want it and we do not keep it: any IOR in that format is masked automatically before the report is written to disk. Company names and EINs are kept, because they are what make the record useful.
How long. Your report link expires after 30 days unless you sign in to keep it. A copy of the upload and the report is archived on our server indefinitely, so the checker can be improved against real filings and a re-filed entry can be compared with what it replaced. If you would rather we held nothing, say so at mrigank@intoglo.com and it is deleted, no questions.
Signed in, you can delete your account yourself at any time. Since 2026-08-30 that removes the stored files as well as the records — both the live report and the archived copy. Before that date it removed the records only, which is a correction we are stating rather than quietly fixing.
What we log, and why
IP addresses are never stored in analytics. For each tool call the server records the tool name, argument shapes, latency and result size, keyed to a session hash derived with a key that rotates weekly — so the hashes cannot be linked across weeks.
The purpose is finding failure modes, so agents get better answers. Detail rows are deleted after 90 days. A narrow allowlist of low-risk values is kept (product search phrases, country and chapter names) to fix vocabulary gaps.
Infrastructure
Traffic passes through Cloudflare, which processes IP addresses transiently for security and caching under its own policies. The origin server is in the United States (Ashburn, Virginia).
Contact
Questions or removal requests: [email protected]. Publisher: Mrigank Singh. Data licence: CC BY 4.0.