Business file to data
A price list, a customer list or an order file — as data you can use
Hand in the file as you received it. You see what was recognised, you check a proposed mapping to the fields you need, and you download the result together with a list of every row that needs a look. Article number 00412 stays 00412, 1.249,50 stays twelve hundred forty-nine fifty, and a date that reads two ways is left alone until you say which way. Free, no account.
Free · no account · nothing is stored
- Limits: 3.0 MB per file, 50,000 rows, 200 columns, 10,000 characters per cell. A larger file is refused with the limit named — never half processed.
- Nothing is stored. Your file is sent to this website only, processed in memory and gone with the answer. Nothing about its content is logged.
- Formulas and macros are never executed. A formula cell gives the value last saved in the file, and is reported.
1. Hand in a file
CSV or other delimited text, Excel (.xlsx) or JSON. Or try one of these — they are invented, and each shows something that usually goes wrong.
This is the target schema: the list of fields your data has to end up in, each with a type and whether it is required.
What this is, and what it is not
It does
- Read CSV and other delimited text (comma, semicolon, tab, pipe), Excel .xlsx, JSON and NDJSON, in any language and the common encodings.
- Keep codes as they are: leading zeros, long numbers, spaces inside an id.
- Work out per column whether 1.234 means a thousand or one-point-two — and say so when the file does not settle it.
- Turn dates into the international form (2026-01-31) only when the day/month order is certain or you set it.
- Give the same result for the same file and the same mapping, every time, with fingerprints to prove it.
It does not
- Correct anything silently. A doubtful value stays as written and is listed with its row.
- Fill in a missing value. Missing stays empty, and a required field that is empty is reported.
- Run formulas or macros. Ever.
- Use AI to guess your mapping in this version. The proposal is plain matching on column names and contents, and it is only a proposal.
- Take large files yet: 3 MB and 50,000 rows per file. Within those limits the answer is immediate; a larger file is refused with the limit named — there is no queued long job in this version.
- Keep your file. There is nothing to fetch later: run it again and you get the identical result.
What it costs
Nothing, within the limits above, for everyone — the work is done in memory and we pay no third party for it, so nothing is charged. If larger files become possible, their price will be worked out from what processing them is measured to cost before it applies. No price has been set.
For your AI agent and for developers
An agent never has to click through this page. The same two steps are the free tools file_to_data_inspect and file_to_data_transform on https://jithox.com/api/mcp — no key — and two plain HTTP routes, POST /api/file-to-data/v1/inspect and /transform. A lost connection is repaired by repeating the call: the same file and mapping give the same run reference and the same result fingerprint. Connect your AI · everything you can get done.