PdfLuxe › Trust & Transparency
Trust & transparency
PdfLuxe asks you to open your documents in it, so here is exactly how it treats them: what is free, where your files go (almost always: nowhere), which open-source components do the work, and what little visit data is counted. Nothing on this page goes beyond what the site actually does — and if you find a claim here that does not hold, tell us and it will be corrected.
Free, always — no hidden charges
- Every tool is free. All 151 tools are free in full — there is no paid plan, no premium tier and nothing held back behind a result.
- No sign-up, no account. No tool asks for an email address, a name or a password before it works.
- No watermark. Nothing is stamped on what you download; the file you get is the finished file.
- No trial, no credits, no task caps. There is no trial period that expires, no credit balance that runs out, and no file-count or hourly limit of the kind other free PDF sites apply. Where a tool has a practical limit of its own, the tool states it plainly — OCR reads up to 10 pages per run, for example, and Edit PDF's server finish has the 25 MB gate described below.
- How a free site pays its bills: PdfLuxe is supported by advertising (Google AdSense and Monetag), which helps keep the tools free. The ads do not come from the tools, have nothing to do with your files, and never change a tool's result.
Your files stay on your device
By default, every PDF, scan and image tool on this site runs as code in your own browser. When you open a file in a tool, it is read, changed and saved again on the device in your hands. There is no upload step, no server copy, and nothing of your file is sent to us or stored anywhere. Close the tab and your file is gone from the page. The exceptions are few and optional — each one is stated on this page in full, and each tool states its own case on its page before anything is sent.
The exceptions, stated plainly — Edit PDF's True Replace and thirteen server converters
Editing itself never uploads anything: opening a PDF in Edit PDF, tapping lines and placing text, highlights and shapes all happen in your browser. Exactly one step can involve a server, and it is a finish you trigger yourself. When you tap Done with lines edited in place and the file is 25 MB or smaller, the PDF is sent over HTTPS to PdfLuxe's own edit server (api.pdfluxe.com), which replaces the edited lines inside the file itself — the old text is removed, not just covered. The file is processed in memory only: it is not stored, not logged and not shared, and the finished PDF is returned to you.
If you moved a text block to a new position, the finish happens on your device instead — a moved block has no original position for the server to replace — and the same on-device finish applies to any file over the server's 25 MB limit. Both upload nothing, and Done tells you plainly which finish applies rather than failing silently. Find & Replace's separate "True replace all" button uses the same server in the same way, and its label says so.
Thirteen converters run on the same server. Three of them convert formats that cannot be read inside a web browser at all: XPS to PDF, MOBI to PDF and CBR to PDF send the file you choose (up to 25 MB) over HTTPS to api.pdfluxe.com, where it is converted in memory only — not stored, logged or shared — and the finished PDF is returned to you. The fourth, Webpage to PDF, sends no file: the web address you type goes to the same server, which loads the page in a real browser engine — a page cannot be fetched and rendered from inside your own browser, because the web's cross-origin rules block it — and returns the rendered page as a PDF (up to 25 MB). The fifth, PDF to PDF/A, sends your PDF to the same server, because PDF/A conversion needs server-only validation libraries — every file it returns has been validated with veraPDF. The sixth, DJVU to PDF, sends your .djvu file to the same server, because DjVu needs a server-only decoder. The remaining seven — Word to PDF, Excel to PDF, PowerPoint to PDF, ODT to PDF, ODS to PDF, ODP to PDF and Pages to PDF — send your office document to the same server, because rendering those layouts faithfully takes a full office engine, which a browser page cannot provide; inside Excel to PDF, a .csv file is the one exception — it is converted in your browser and never sent. Each of the thirteen states the server step on its own page and inside the tool, before anything is sent.
Outside services you choose to use
A small number of tools can work with an outside service — always one you choose, and each tool page says so before anything is sent. The AI writing tools can use an AI service you connect with your own key; they send the text you type to that service, never your files. Translate PDF's cloud engines receive the text being translated (its on-device engine sends nothing). OCR PDF offers optional cloud engines — OCR.space, Google Cloud Vision and Azure — that work only with your own API key; only if you pick one are page images sent to that provider, under your key. The default OCR engine, Tesseract.js, runs entirely on your device, and the on-device AI options, WebLLM and Transformers.js (listed below), likewise run in your browser and send nothing anywhere.
Open-source components
PdfLuxe is built on established open-source libraries rather than a closed, unauditable engine. These are the main components, with the versions the site loads today (17 November 2026), and each name links to its official project page. QPDF runs as WebAssembly in your browser; PyMuPDF is the only component on this list that runs on a server — the edit server described above.
| Component | What it does here | Licence |
|---|---|---|
| pdf.js 3.11.174 | Mozilla's PDF engine — displays and reads PDF pages in your browser. | Apache-2.0 |
| pdf-lib 1.17.1 | Creates and edits PDF files — merge, split, forms, page numbers, watermarks and more. | MIT |
| QPDF WebAssembly | Password protection with AES-256 encryption (Protect PDF), structure repair (Repair PDF) and unlocking — all on your device. | Apache-2.0 |
| Tesseract.js 5.0.5 | OCR — reads the text out of scanned pages in OCR PDF. | Apache-2.0 |
| Fabric.js 5.3.1 | The drawing canvas behind Design Studio and Magic Canvas. | MIT |
| JSZip 3.10.1 | Builds ZIP downloads and the ZIP-based Office files (DOCX and PPTX) that some exports produce. | MIT |
| WebLLM 0.2.79 | Runs AI models on your device for the local-AI tools. | Apache-2.0 |
| Transformers.js 2.17.2 | On-device AI models for the browser AI features. | Apache-2.0 |
| PyMuPDF | The edit server's PDF engine — it performs True Replace, in memory only. | See project page |
Privacy & statistics
There is no account on PdfLuxe, so there is no name, email address or password of yours here to leak — and we do not sell data.
What is counted is small, and it is counted openly. When a page on PdfLuxe opens, the page path and the time of the visit are counted, so the site owner can see how many visits each page receives. On its own, that counting stores no IP address and no visitor ID — and if you reject the privacy banner, never answer it, or your browser sends a Do Not Track signal, counting is all that ever happens, and every tool keeps working exactly the same.
Only with your permission — given in the small privacy banner at the bottom of the page, and changeable at any time from the "Privacy choices" link in the footer — a visit also records a random Visitor ID (it identifies this browser only; it is not a fingerprint), an IP address and an estimated location. Visitor records are kept for 90 days and the plain day counts for 120 days, then deleted. The statistics are never sold, never shared for advertising, and never connected to your files.
The full Privacy Policy opens from the "Privacy Policy" link in the homepage footer.
Questions about any of this
Email support@pdfluxe.com with any question about privacy, processing or a claim on this page. The inbox is read by the people who build the tools, and a claim that turns out to be wrong will be fixed on this page, not argued about.
Updated 17 November 2026.