NPI Redactor - clean your files before using AI

100% Local

Everything happens inside your browser, on your computer. Your files are never uploaded anywhere. Don't take our word for it: open your browser's Network tab (F12) and watch while you use this. Nothing leaves your device.

Drop files here, or click to choose

.txt .md .csv .docx .xlsx .pdf and scans/photos (.png .jpg). Add your files, set the options below, then click Run Redaction.

Options Redact all dates (any date in a loan file could be a DOB or closing date) Also redact dollar amounts & credit scores (kept by default, useful for AI and not identifying once names/SSNs are removed) Aggressive mode — also redact every common first/last name, any all-caps name-like text, and any long number. Over-redacts on purpose; use when you'd rather be safe than sorry and clean up the output afterward. Call transcript mode — for recorded calls and other unformatted text: also catch numbers said digit-by-digit or spelled out (SSN, phone, account), spoken emails (“name at gmail dot com”), and conversational name intros (“this is John Smith”). Account holder & other names to redact

Put the account holder's name here (plus any co-borrowers, spouses, or businesses). Enter each name once — the tool automatically generates the variations a document might use, so you don't have to:

first last · last first · last first middle-initial
the comma form too, e.g. an IRS/ACH line like SMITH, JOHN A

Separate different people with commas: John Smith, Jane Doe. To enter a single name that itself contains a comma, wrap it in quotes: "Smith, John A".

Always keep (never redact)

Terms the tool should leave alone — your company or lender name, loan products, anything that gets caught by mistake. Helpful with Aggressive mode. Comma-separated: Acme Title, Main Street Capital.

Always spot-check the cleaned file before sharing it. No automated tool catches 100% — use the Review panel on each result to preview the output and double-check anything that wasn't redacted. PDFs are rebuilt at their original text positions containing only cleaned text, so the layout is preserved and no original data hides in any layer. Word docs come out as cleaned text; scanned PDFs and photos are read with on-device OCR. Nothing is uploaded.

Before you use this tool

Strip client NPI from any file before it touches AI.