The file tools that never see your files
FileAnvil is 22 free tools for images and PDFs that run entirely inside your browser tab. Not "we delete your files after an hour" — there is no upload endpoint on this site at all. Your file is read, processed and saved without ever crossing the network.
Why it exists
Nearly every online file tool works the same way: you hand over your file, a server somewhere does the work, and you download the result. You are then left trusting a privacy policy about the copy they still have — how long they keep it, who can read it, what happens if they are breached.
For a holiday photo, fine. For a signed contract, a payslip, a passport scan, a medical record or a client's document, that is a genuine problem — and for some professions it is a compliance problem, because the upload itself is the disclosure. The awkward part is that the upload was never necessary. It is an artefact of how web apps were built in 2010, not a technical requirement in 2026.
Browsers can now do this work themselves. So the honest version of a file tool is one that simply never asks for the file.
How it actually works
Every tool here is WebAssembly running on your own device — the same battle-tested libraries a server would use, compiled to run in the browser instead. The site itself is static: no functions, no edge runtime, no database, no API. There is nothing on the other end that could accept a file even if it wanted to.
| When you… | …this runs, on your device |
|---|---|
| Compress or convert an image | jSquash — the same WebAssembly codecs Squoosh uses: MozJPEG, libwebp, OxiPNG, plus AVIF and JPEG XL |
| Open an iPhone photo | libheif, for decoding HEIC |
| Merge, split, rotate or watermark a PDF | pdf-lib |
| Turn PDF pages into images | pdf.js, Mozilla's PDF renderer |
| Password-protect, unlock or repair a PDF | qpdf, compiled to WebAssembly (AES-256) |
| Make a scan searchable | Tesseract OCR |
| Download a batch at once | fflate, zipping in the browser |
It also installs as an app and keeps working with the network off, because the codecs are cached on first visit. That is not a gimmick — it is the cleanest proof available that nothing is being sent anywhere.
Don't trust this page — check it
The entire point is that you should not have to take my word for it. Two ways to confirm it yourself, both taking about ten seconds:
- Turn off your Wi-Fi and use any tool. It still works. A tool that uploaded your file could not.
- Open your browser's Network tab and process a file. No request appears, and nothing is sent — there is a recording of exactly that if you want to see it first.
Why it's free
Because your device does the work. The expensive part of a file service is the server farm doing conversions; there isn't one here, so there is no bill to pass on. Serving a static site to a lot of people costs very little.
There are no accounts, no file-size tiers, no watermarks and nothing to upgrade. The only thing that ever leaves your browser is anonymous usage counting: a pageview, and an event that says "a file was processed" with the tool's name. No cookies, no filenames, no file contents, ever — there is a test in the build pipeline that fails the release if a test file's name so much as appears in an outgoing request. The privacy policy spells out the rest.
What it doesn't do
A page like this is usually where the overclaiming happens, so here are the real constraints:
- Very large files depend on your device. The ceiling is your available memory, not a paid plan. A 500 MB PDF on an old phone will struggle.
- Redaction rasterises the pages it touches. True content-stream redaction in the browser only exists in AGPL or commercial engines, so the honest approach is to flatten the page instead of pretending.
- Signing is a simple electronic signature — an image you place on the page. It is not a certified or qualified signature, and I won't market it as one.
- Squoosh is better for a single image. It has a side-by-side quality preview and it is also fully client-side. This is for batches, PDFs and HEIC — and the comparison page says so plainly.
- No accounts means no history. Nothing syncs between your devices, because nothing about you is stored anywhere.
Who builds it
FileAnvil is built and maintained by Thomas van der Bruggen, an independent developer. It is not a startup, not venture-funded, and not a data company — there is no business model that depends on knowing anything about you, which is precisely why the architecture could be built this way.
Tools get added and improved regularly; the changelog is the honest record of what has shipped.