Inspect a CSV with a real RFC 4180 parser: column types and ranges, rows with the wrong field count, duplicates, and a cleaned file you can copy back out.
Waiting for a CSV file
Parsed properly — a quoted comma, a doubled quote and a newline inside a field are all handled. Nothing is uploaded.
The report appears here
Every column with what it looks like and what is in it, every ragged row, and every exact duplicate.
CSV Inspector runs entirely on your device — nothing is uploaded.
You’re on the Free plan. Compare plans
What is actually in this file, before you load it into something that will fail halfway through and tell you a line number and nothing else. Column by column: what the values look like, how many are filled, how many are unique, the length range, and the two things that most often need fixing.
It uses a real parser. Splitting each line on commas is wrong for any file containing a quoted comma, a doubled quote, or a newline inside a field — which is most real exports from most real systems. This runs the RFC 4180 parser the site's CSV converters already use, so what the inspection reports is what a correct reader would see rather than what a naive one would.
Types are reported and never applied. CSV has no types; every cell is text. Saying what a column looks like is useful, converting it is destructive, and the case that matters is a leading zero. A value like a padded postcode is a perfectly good identifier and a ruined number, so columns containing one are flagged with a count rather than quietly widened to an integer.
Ragged rows are measured against the header, not against the widest row in the file. That sounds like a detail and is not: measuring against the widest row means one over-wide row makes every other row look wrong, including the header, and the tool then reports the problem on line one. Here the header defines the expected field count and the rows are checked against it.
The cleaning controls do the small set of things that are nearly always what a file needs: drop exact duplicate rows, drop rows that are entirely empty, trim every cell, transpose, and write the result with a different delimiter. A ragged row is padded rather than dropped, so nothing is lost on the way out. Nothing is uploaded — the file is parsed in the page, which is what makes it usable on an export with real customer data in it.
Yes. It uses the RFC 4180 parser the site's CSV converters run on, so a quoted comma, a doubled quote and a newline inside a field are all handled. Splitting on commas would get all three wrong.
No, deliberately. CSV has no types and converting them is destructive. The clearest case is a value with a leading zero: converting it to a number destroys the zero, and a postcode or an account number does not survive that.
A row whose field count differs from the header's. It is the usual reason an import fails partway through, and it is almost always a missing quote earlier in the file rather than a problem with the row itself.
Because this counts exact repeats of a whole row. A row differing in one field is a different row. Spreadsheet duplicate finders often compare a single column, which finds many more.
Because most readers treat it as part of the first header name, so the first column arrives with an invisible character stuck to its name and every lookup by that name fails. It is harmless in a document and a real problem in a CSV.
No. It is parsed in your browser and never leaves the page, which is what makes it safe for an export containing real customer data.