Free browser tool
Clean a CSV online
Clean a CSV with explicit, reversible choices. Identifiers such as 00123 stay as text.
Your file stays on this device. No upload, account or tracking.
Your CSV
Up to 600,000 characters, 5,000 data rows and 100 columns.
Result
Result
Technical output
Long text is shortened on screen. The download and the copy contain the complete file.
Clean a supplier CSV and reconcile the rows
Select Try sample to load three synthetic product rows. The first two contain the same SKU 00123, title and price; the third has SKU 00456. The sample turns on Trim spaces and Remove duplicate rows. Select Clean CSV: expect three input rows, two output rows and one removed row.
The output retains 00123 and 00456 as text. The title Widget A becomes Widget A. Deduplication compares every cell after the selected trimming step; it does not merge different records that happen to share a SKU, sum quantities or choose which conflicting product record is correct.
Keep your original export. If repeated rows represent separate orders or events, turn off Remove duplicate rows. If edge spaces carry meaning, turn off Trim spaces. Compare the input, output and removed counts before downloading cleaned.csv.
Choose the export mode for the destination. Spreadsheet protection can prefix formula-like text with an apostrophe; machine export preserves that text and needs a destination that handles it safely. Neither mode makes another application's import rules known, and spreadsheet programs can reinterpret leading zeros when opening a file.
Try the cleanup with the synthetic sample above. To inspect unfamiliar headers first, use CSV profile; to check the output against explicit destination rules, use CSV validator. Files remain on this device.
About this tool
What this tool does
Trim cell whitespace and remove exact duplicate rows without uploading your file.
Limits and safe use
Up to 600,000 characters, 5,000 data rows and 100 columns. Keep the original file and review the output before using it elsewhere. This tool does not certify an import, infer missing meaning or send data to a model.