Every free PDF tool makes you upload your file. That's not a technical requirement.

Merge two PDFs on ilovepdf, smallpdf, or almost any free PDF site and the first thing that happens is your file goes to their server. Then you wait, then you get it back - often watermarked, and usually with a page cap if you didn't pay.

I spent the last stretch building the version that doesn't do that: 20 tools, all of them running in the browser tab. So I now have an unusually specific answer to "does this actually need a server?"

Mostly, no.

Page operations — merge, split, rotate, reorder, delete pages, page numbers, metadata — are pure object-graph edits. pdf-lib does them in milliseconds on a file you already have in memory. There is no computation here that a laptop from 2015 can't do instantly. Rendering and compression run through pdf.js and canvas. Password protection is AES; your browser has had that in SubtleCrypto for years. OCR is tesseract.js, which compiles Tesseract to WASM and runs offline once the language data is cached.

But I want to be honest about where it's harder, because this is the part nobody selling either model says out loud:

- PDF to Word is a real tradeoff. I extract text with pdf.js and rebuild a .docx. You get the content; you do not get a pixel-perfect reconstruction of a complex layout. A server running LibreOffice does that better. If that's what you need, use the server tool — I'd rather say so than oversell.

- Memory is a genuine ceiling. A 500 MB scanned PDF will fight you in a browser tab, and mobile Safari especially. A server has as much RAM as you're willing to rent.

- OCR is slower. No GPU, and the first run downloads language data.

- One server pipeline is easier to maintain than the same code behaving across four browser engines. That's a real engineering cost, and it's on me, not the user.

So the honest version is: some of it is harder without a server, and a little of it is genuinely worse. Almost none of it is impossible.

Which leaves the question of why the upload is universal anyway. My read: it isn't a technical decision, it's a metering one. If every free file costs you server time, the free tier has to be capped or watermarked or it bleeds money. The upload is what makes the cap enforceable. Take the upload away and there's nothing left to meter — which is also why I can offer no page limit without it being a promotional trick.

Curious whether people who've built on the server side see it differently. Is there something that genuinely forced your hand that I'm glossing over?

(I'm launching PDFLove today - free, in-browser PDF tools. But I'm more interested in the argument than the plug.)