PDF Arranger

Last reviewed Sep 13, 2026 · 859 words

Last month I scanned a 40-page lease as four separate PDFs, two of them upside down, and had the single correct file 90 seconds after opening PDF Arranger. That is the whole job description. It is a small Python and GTK desktop app that shows every page of every PDF you drop into it as a thumbnail; you drag pages into order, rotate, delete, split and save. Because it uses pikepdf (and therefore qpdf) underneath, the output pages are the original page objects, byte for byte, not re-rendered images. Nothing gets blurrier and file size does not balloon.

A desktop tool in a directory of servers

PDF Arranger has no Docker image, no web interface and no multi-user anything, and that is fine: it is the tool you reach for on the machine where the PDF already is. In a self-hosted document stack it occupies the slot before the server. Scans and downloads get tidied here, then land in Paperless-ngx for OCR, tagging and search, or in Calibre if they are books. Paperless can rotate and split documents itself now, but its editing is deliberately basic; for anything beyond a single rotation, doing it locally first is faster and does not create a consumption task that then has to be re-processed.

The server-side sibling is Stirling-PDF, a much bigger project that does everything PDF Arranger does plus OCR, compression, conversion, signing and about 40 other operations through a browser. If you want a family member to fix a PDF from their phone, Stirling is the answer. If you personally want to reorder 12 pages with the least friction, a native window with drag and drop still beats uploading to a web app and downloading the result, and PDF Arranger opens in under a second.

What it does and what it refuses to

Merge any number of files, split by page ranges or into single pages, reorder by drag, rotate 90 degrees at a time, duplicate pages, delete pages, crop margins, and edit the title and author metadata. Recent versions add page-level zoom and a "generate booklet" imposition, which folds a document into printer spreads. There are keyboard shortcuts for all of it and multi-select works the way a file manager's does.

It does not OCR, does not compress or re-encode images, does not edit text or annotations, does not fill forms, and does not sign. Those are features, not omissions; each of them would require rendering the page, and not rendering is the reason the output is trustworthy. Send those jobs to Stirling-PDF or a command-line tool.

Cropping hides content, it does not remove it

The one behaviour that can bite. PDF Arranger's crop sets the page's crop box, which tells viewers what to display. The full original page, including whatever you cropped away, is still inside the file and any PDF library can read it. Do not use crop to redact a signature, an address or an account number and then email the result. For true redaction you need a tool that rasterizes or rewrites the content stream, and that is a different job.

The same logic runs the other way: because pages are moved rather than rebuilt, form fields, links and bookmarks from the source files survive merging more often than in tools that print to a new PDF. Bookmarks in particular come through intact from the first file in a merge.

Installation on the three platforms

Linux: sudo apt install pdfarranger on Debian and Ubuntu, dnf install pdfarranger on Fedora, or the Flatpak com.github.jeromerobert.pdfarranger from Flathub, which is what I use because it tracks upstream releases within days. Windows: an installer and a portable zip are attached to each release on GitHub. macOS: MacPorts and Homebrew both package it, though the GTK look is visibly foreign there. Memory is negligible; the 256 MB figure is a ceiling for a document with several hundred thumbnailed pages, and the Flatpak's disk footprint is larger than the app because of the GNOME runtime.

Where it fits for ebook people

In the ebooks category it is the odd one out: not a library manager, not a reader, not a converter. Its role is preparation. Scanned books that arrived as one PDF per chapter get merged; a scan with a blank page between every leaf gets cleaned; a sideways-scanned spread gets rotated. Then Calibre or Calibre-Web takes the result. For document archives rather than books, the document management category is where the server-side pieces live.

What I'd do

Install the Flatpak on any Linux desktop that touches scanned paper and keep it in the dock next to the scanner app. Use it for everything that is reorder, rotate, merge or split; send OCR, compression and redaction to Stirling-PDF running on the server. Never crop to hide information. And if you find yourself wishing PDF Arranger had a web UI, that wish is called Stirling-PDF and it is one container away.