Filozy is a free set of 39 PDF tools that run in your browser. It is built and run independently by the Filozy team, from Ontario, Canada. Pages on Filozy are drafted with AI assistance, then checked against the tool code and by running the tool. That check has missed things; the corrections below list the mistakes that changed what a page told you, and when we fixed them. If you find something wrong, write to admin.filozy@gmail.com.
Who runs Filozy
Filozy is built and run by the Filozy team, operating from Ontario, Canada. That is the operator named in the privacy policy and the terms of use. It is independent: it is not owned by or affiliated with any of the companies whose products it compares itself with.
The team is small, and there is no support department. That cuts both ways. The people who answer email are the people who build and test the tools, and decisions get made quickly. There is also no help desk, no guaranteed response time and no service-level agreement.
Filozy is a static website. Its 39 tools organise, convert, edit, secure and sign PDFs. It shows you the pages before you make changes you cannot undo, and it explains when a result came out weak or unchanged. You can pass a file from one tool to the next without choosing it again.
Why it exists
Most PDF websites upload your file, process it on a server, and delete it some hours later. That is a reasonable engineering choice, and the honest services document it clearly. The objection is narrower than “servers are bad”. People put contracts, identity papers, medical letters and bank statements through PDF sites, and “we delete it after an hour” still means a copy existed.
So Filozy does the work in your browser instead. The constraint shapes everything else: which features are possible, which are not, and why a phone can run out of memory on a file a server would finish.
You do not have to take that on trust. The verification page gives three checks you can run yourself, including one with the internet switched off.
How the tools are built
The document work runs in a Web Worker, a background thread in your browser, using open-source engines: pdf-lib to build and change PDFs, pdf.js to read and draw them, Tesseract for text recognition, HarfBuzz to shrink embedded fonts and fflate for ZIP files. The engines and their data files come from Filozy’s own domain. Opening a document therefore does not tell a third-party server which engine you needed or which file you opened. The software notices list each engine’s license.
How we test
Every release is tested automatically before it is published, and a release that fails a check does not go out. The test documents are generated by the test code itself. No visitor’s file, and no real customer document, is ever used as test data.
The PDF operations themselves
One test suite runs the document engines directly on generated PDFs. It checks what comes out, not just that a file came back:
- merging keeps every page;
- splitting makes the groups asked for;
- rotating changes only the selected pages;
- a protected file opens again with its password;
- form answers survive filling and flattening;
- metadata changes are confirmed by reading the file back;
- compression keeps form fields, metadata and signatures;
- a batch ZIP keeps every file intact.
Every tool, in a real browser
Each live tool is then opened in a real Chromium browser, given a generated file and run from start to finish. The test fails if the tool shows an error or produces no download. Where it applies, it also fails if the output has the wrong number of pages or a changed page size. Four of the workflow recipes are also run step by step, passing the file from tool to tool. Every tool page is also loaded at desktop, phone and narrow-phone widths. It must load without errors, show its file picker and fit the screen without sideways scrolling.
The tools where mistakes would cost most get their own tests:
- Compress is run at every level. It must never return a bigger file or lose a page, and pictures must stay at Keep quality unless you choose otherwise.
- Edit and Sign are tested with touch gestures. Pinching to zoom must never stamp a signature on the page.
- Split is checked for contrast, touch-target size and keyboard use.
- Batch results must save, share and ZIP every file intact.
Privacy, checked by the build
- Every place in Filozy’s own code that can send a network request has to be listed, with where it goes, what triggers it and what it sends. An unlisted one stops the build.
- Every use of browser storage is on an approved list too, and any code that sets a cookie fails the build. The same check makes sure the OCR engine loads its language data from Filozy’s own domain and does not save a copy in browser storage.
- The site is served with a Content-Security-Policy generated from the finished build. Browsers will run only Filozy’s own scripts and Cloudflare’s analytics script. Pages can send data only to Filozy and to that analytics service.
- No page may load advertising or consent scripts, or contain ad slots. The build checks every page.
Once a release is live, the home page, a tool page and the privacy and terms pages are fetched again. This confirms the tested release is the one being served. It also confirms that those pages load no advertising and no outside script except the analytics script the privacy policy discloses.
What the tests do not prove
The tests use generated documents and an automated browser. They show that the tools work on those files. They do not prove that every real document or every phone will behave the same way. Where a tool has a known limit, its page says so. You can also check any job yourself. Under Processing details, most tools show a receipt. Its “0 document bytes sent” is printed by the page, and it adds the browser’s log of requests the page itself made while your file was being processed, which leaves out requests made inside the background worker. The verification page explains what it does and does not prove.
How we write and review
Pages on Filozy are drafted with AI assistance, then checked against the tool code and by running the tool. That covers tool pages, guides, answers, comparisons and this page. The check has missed things; the corrections below list the mistakes that changed what a page told you, and when we fixed them.
- Claims about Filozy’s tools are checked against the code that does the work. If a page says a tool keeps form fields or leaves pictures alone by default, the check is made in that tool’s code and by running the tool. The check is not perfect. When it, or a reader, finds a page promising more than its tool does, we correct the page. The corrections below are examples.
- Claims about other services are taken from those services’ own published pages. They are quoted where the wording matters, linked, and dated, as in “checked September 2026”. We include a number only if the other service’s page states it. If a comparison cannot be backed up that way, we do not publish it. Every comparison also says when the other service is the better choice.
- The PDF upload test was measured, not estimated. On 22 September 2026, from Toronto, Canada, each service’s Compress page was opened in a fresh automated browser. Every request and cookie in the first fifteen seconds was recorded, then the page was given a test PDF. An upload counted only when the file itself, or a piece of it, left the page. Every figure on that page is calculated from the raw data, which is published with it. Services that could not be tested are listed as not tested rather than guessed. Filozy was measured the same way.
- Dates. Guides, answers, comparisons, recipes and the upload test carry a line reading “By the Filozy team, with AI assistance · Reviewed” and a date, then a link back to this section. The date is when that page’s content last changed. Changes go through the same checks.
- What we do not publish: testimonials, ratings, user counts or awards.
Corrections
When a page is wrong, we correct it and its reviewed date changes. If a mistake changed what a page told you, such as a feature a tool does not have or another service’s terms stated wrongly, we list it here with its date. Product fixes also appear on What’s new. To report an error, see contact below.
- 19 October 2026, Excel to PDF and OCR descriptions. The Excel to PDF description and its result note said dates appear exactly as the sheet shows them. The tool writes every date as year-month-day, whatever format the sheet uses. The OCR description said small jobs stay on your machine, which suggested large ones might not. OCR always runs in your browser, whatever the size of the file. Both now say what the tools do.
- 30 September 2026, Adobe Acrobat online comparison. An earlier version said every Adobe online tool needs a sign-in before you can download the result. Adobe’s own pages show that Compress, Merge, Protect PDF and Word to PDF do not. The page now quotes Adobe directly.
- 22 September 2026, tool pages. Three claims were removed because the tools did not do these things: that Word to PDF keeps tables, that the Organize PDF pages tool adds blank pages and imports other PDFs, and that Crop accepts typed edge values.
- 22 September 2026, Page Numbers. The Prefix box was being ignored, so “Page” never reached the page. The tool now prints it as “Page 1”, “Page 2” and so on.
- 16 September 2026, the processing receipt. The receipt said “0 bytes sent”, although loading the page itself takes bytes, and it said “1 pages” on single-page jobs. It now reads “1 page · 0 document bytes sent”, naming which bytes it counts, and the verification page quotes it exactly.
How Filozy is paid for
It is free to use, with no account, no daily limit and no watermark. There is no paid tier, and there is no advertising on the site today. The plan is to fund it with advertising later. If that happens, ads will sit only on informational pages. They will be kept apart from choosing a file, previewing, editing, processing and downloading, and any permission the law requires will be asked for first. The privacy policy will be updated before any advertising starts. These commitments come with that plan:
- Advertising never receives documents. Not their contents, filenames, passwords, recognised text or signatures. The way the site is built makes this structural rather than a promise: the file is never sent to us, so there is nothing to pass on.
- Tools will not move behind a paywall. What is free today stays free.
The site is hosted by Cloudflare, which adds its cookieless Web Analytics to count visits. That is the one third-party script on the site. The privacy policy says what it records and how to block it.
Independence
Filozy is not affiliated with, endorsed by or sponsored by Adobe, iLovePDF, Smallpdf, PDF24, Sejda, Soda PDF, PDF Candy, TinyWow, CloudConvert, pdfFiller, Foxit or any other company named on this site. Their product names appear only to identify what is being compared. Nothing on Filozy is sponsored, and no link to another product earns us a commission.
Why the code is not public
Filozy is closed source. That is a fair thing to hold against a privacy claim. The honest reason is that it is commercial software, not a principled position on openness. It is also why the verification page relies on what you can observe rather than on reading code. A network trace and an offline run show what the software does, whether or not you can read it.
What it is not
Filozy is not a mature commercial suite. It has no native apps, shared workspace, multilingual interface, support desk, uptime agreement, certificate-backed signing service or cloud document storage. Its text recognition reads English only. Its PDF to Word conversion struggles with multi-column layouts. A very large scan can use up a phone’s memory where a server-based tool would finish.
When those things matter more than keeping the file on your device, larger established products are a better choice. The comparison pages say which ones, and for which tasks they win.
Contact and reporting an error
Email: admin.filozy@gmail.com
Videos: Filozy on YouTube, short videos about keeping PDFs private and doing everyday PDF jobs with the tools.
Write about a wrong statement on any page, a tool that gave a bad result, a privacy or security concern, or anything else. The most useful reports include:
- the address of the page;
- what you expected and what happened;
- your browser and device.
Please do not attach private documents or include passwords. A description of what went wrong is usually enough. If a sample would help, remove personal and confidential information first. Email is separate from the tools and is handled through Gmail. The privacy policy explains what happens to what you send.
More
See what’s new since launch, browse every tool, try a connected workflow recipe, or read the guides.