Do my files get uploaded anywhere?

No. Everything runs locally, and the app is sandboxed with no network access to upload through.

Does it work without an internet connection?

Yes, entirely. Nothing in the app needs a connection, including the SQL engine and every export.

How large a file can it open?

Larger than a spreadsheet will. Files stream rather than loading into memory, so the practical limit is disk space, not RAM, and Excel's 1,048,576-row ceiling does not apply.

Which computers does it run on?

On the Mac, macOS 11 Big Sur and later, universal for both Apple silicon and Intel; the App Store download is about 3.9 MB. On Windows, version 1809 and later, x64 and ARM64. Same engine and same file formats on both.

Is it a subscription?

No. It is a one-time purchase - $19.99 on the Mac App Store - with no in-app purchases and nothing to renew.

Is there a free version?

Yes. unfckr.app is the same engine running in your browser, free and with no sign-up. The desktop app adds the folder grant, the command-line runner, and native file handling.

Is there a Windows or Linux version?

Windows is done and its Microsoft Store listing is approved. Linux is planned after it. The engine, the pipeline and every file format are already platform-independent; what differs is the shell around them.

Wranglify - the workbench for files that break everything else

why this exists

Three ordinary files that nothing else will open

Every analyst has the same three problems, and the usual answers each fail on a different one.

The spreadsheet gives up

A spreadsheet stops at 1,048,576 rows and slows down long before that. The export you were sent has more.

Wranglify opened a 497 MB, 5,096,070-line file and scrolled every record at the same speed at the end as at the start.

The website is not an option

The free online converter wants an upload. The file has customer records in it, so that is the end of that conversation.

Sandboxed, and it makes no network requests at all. Not a policy. A property you can check.

One-click tools do one thing

The encoding is wrong AND a column holds three values AND the statuses collide by case. Four tools, four exports, four chances to lose a row.

One pipeline, in order, reversible, and saveable as a recipe you can run again next quarter.

power tools

Built for people who do this all day

Not a converter with a nice icon. The parts below are the reason people keep it open. Each is shown in a different one of the eleven themes.

A real SQL engine, on your own machine

Load several files as tables and join, filter, group and sort them with SQLite running locally. Every step shows the SQL it generates, so the chain is auditable rather than magic.

Joins across files, a read-only console, full-text search
Export the result as CSV, JSON, XLSX or a real .sqlite database

shown in blackout

Schema and INSERTs, typed from the whole file

Turn a file into CREATE TABLE, INSERT, SELECT and UPDATE for PostgreSQL, MySQL, SQL Server or SQLite. The types come from a scan of every row, not the first ten.

VARCHAR sized to what is actually there, INT or BIGINT by magnitude
DECIMAL by observed scale, NOT NULL only where there are no nulls
A primary key only where the column is provably unique

shown in antique

Pivot by dragging, computed in SQL

Drop fields into rows, columns and values. Totals and subtotals are their own queries, so they are measured rather than added up from what happens to be on screen.

Number formats per measure, heat-mapped cells
Out as Markdown, HTML, TSV or a styled workbook

shown in reef

Charts that draw what you built

Build a pivot, press chart, and the picture is of that, aggregated in the engine rather than sampled from a preview. Drag across it to filter the data behind it.

Export SVG, PNG, or the Vega-Lite spec that drew it

shown in cyberunc

Do it once, then do it to the folder

The pipeline saves as a recipe: a plain JSON file listing your steps in order. Authorize a folder once and apply that recipe to every file in it, from the app or from a terminal.

The terminal command is the same sandboxed app, not a second program
A dry run prints the plan before anything is written

shown in planet

the part that matters

Do not trust us. Make the machine say it.

Anyone can write "we respect your privacy" on a page. The app ships a test that tries to smuggle data out through every channel there is and shows you each one refused, naming the rule that did it. Nothing is actually sent: the target is under .invalid, the suffix reserved so it can never exist.

There is no account, no telemetry, no crash reporting and no server. The privacy page explains every entitlement the app holds, including the one that looks like a contradiction until you read why it is there.

Read the privacy model

✓fetch()refused by connect-src
✓XMLHttpRequestrefused by connect-src
✓WebSocketrefused by connect-src
✓tracking pixelrefused by img-src
✓<script src>refused by script-src-elem
✓form POSTrefused by form-action

6 of 6 channels refused by the browser itself. Every route out is closed.

where it runs

Mac and Windows. Then Linux.

One codebase, three targets. The engine, the pipeline and every file format are already platform-independent; what differs is the shell around them. Rather than promise dates, here is the honest state of each.

Available

macOS

Universal for Apple silicon and Intel, macOS 11 and later. Sandboxed for the Mac App Store, with the folder grant and the terminal runner working inside the sandbox.

Available Approved, arriving

Windows

The port is done and the Microsoft Store listing is approved. Windows 10 version 1809 and later, x64 and ARM64. It ships as a packaged desktop app, so the folder grant, batch runs and the terminal command all work exactly as they do on the Mac.

Planned

Linux

Next. Same shared core again, so the remaining work is packaging and the desktop integration rather than the product itself.

Wranglify: an offline CSV, JSON and SQLite workbench for Mac and Windows

Wranglify is a native desktop app for Mac and Windows, for the files that defeat everything else: the 500 MB CSV that freezes the spreadsheet, the export with mojibake through every accented name, the nested JSON that is not a table yet, the extract you are not allowed to paste into a website. It profiles, cleans, joins, queries and converts them on your own machine. Nothing is uploaded, because the app has no network code to upload with.

What it opens

CSV and friends - comma, semicolon, tab or pipe delimited, detected rather than asked about, plus anything pasted out of Excel, Google Sheets, a SQL client or a web page.
JSON and NDJSON - including responses that wrap their records in an envelope, where the array inside becomes the table.
Excel workbooks - .xlsx, every sheet.
SQLite databases - open a .sqlite file directly and query it.

Files stream through the pipeline rather than loading whole, so the ceiling is your disk rather than your RAM. That is what makes it a viable large CSV viewer for Mac and a large CSV editor for Windows alike: files well past Excel's 1,048,576-row limit open and stay responsive.

What it does

Profiles every column from the whole file, not a sample: distinct counts, top values, min, max, mean, date ranges, null share, proven uniqueness.
Cleans and transforms: flatten nested JSON, filter, dedupe, sort, rename and drop columns, split one column into several at a delimiter, extract new columns with a regular expression, fix individual cells. Every step is visible, reorderable, switch-off-able and undoable.
Scrubs PII: finds emails, phone numbers, SSNs and card numbers, then masks, hashes or drops them.
Repairs broken encodings: JosÃ© back to José. Non-UTF-8 files are detected on open, and each repair is proven by a byte roundtrip first, so words that only look wrong are left alone.
Runs real SQL on a full SQLite engine: joins across several files, GROUP BY, a read-only console, full-text search over millions of rows, and a relationship map that measures each candidate join before you build it, so you know whether it drops rows, multiplies them, or matches cleanly.
Pivots and charts: build a pivot by dragging, computed in SQL rather than in memory, and chart any table as bar, line, area or point.
Reads logs as events: layouts detected, stack traces kept whole, timestamps normalized.
Writes what the next tool needs: CSV, JSON, NDJSON, .xlsx, a real .sqlite database, or a .sql script - CREATE TABLE, INSERT, SELECT and UPDATE for PostgreSQL, MySQL, SQL Server and SQLite, with column types suggested from your actual data. Also TypeScript interfaces and JSON Schema.
Does it again to the folder: grant Wranglify a folder once and a saved recipe runs over everything in it, including from Terminal with the wranglify command, so the cleanup you worked out by hand becomes the one you repeat every month.

Why offline is the feature

Most data tools that look like this are websites, which means the file you were told not to email leaves the building anyway. Wranglify reads only the files and folders you hand it - sandboxed on macOS, a packaged desktop app on Windows with no network code in either. There is no account, no sign-in, no telemetry, no crash reporter and no analytics, so there is no session log of which files you opened - the pipeline you built is the only record, and it stays on your disk. The privacy page sets out exactly what that means.

Frequently asked questions

Do my files get uploaded anywhere?: No. Everything runs locally, and the app is sandboxed with no network access to upload through.
Does it work without an internet connection?: Yes, entirely. Nothing in the app needs a connection, including the SQL engine and every export.
How large a file can it open?: Larger than a spreadsheet will. Files stream rather than loading into memory, so the practical limit is disk space, not RAM, and Excel's 1,048,576-row ceiling does not apply.
Which computers does it run on?: On the Mac, macOS 11 Big Sur and later, universal for both Apple silicon and Intel; the App Store download is about 3.9 MB. On Windows, version 1809 and later, x64 and ARM64. Same engine and same file formats on both.
Is it a subscription?: No. It is a one-time purchase - $19.99 on the Mac App Store - with no in-app purchases and nothing to renew.
Is there a free version?: Yes. unfckr.app is the same engine running in your browser, free and with no sign-up. The desktop app adds the folder grant, the command-line runner, and native file handling.
Is there a Windows or Linux version?: Windows is done and its Microsoft Store listing is approved. Linux is planned after it. The engine, the pipeline and every file format are already platform-independent; what differs is the shell around them.

Your dataset is broken.
It is also confidential.