How browser-based tools work

Two tools can look identical, with the same drop zone and the same "Download" button, and still treat your file in completely different ways. One sends it to a server; the other never lets it leave your device. Here's the difference, and how to tell which one you're using.

Step through it

Pick a kind of tool, then click through the four steps of compressing a PDF. Watch where the file goes and what your browser's Network tab would show.

Why both kinds exist

Server tools made sense when browsers were slow and could only display pages. Heavy work like video encoding or OCR happened on the company's machines. They still make sense for jobs that need a large AI model or a big database.

Modern browsers are far more capable. A handful of built-in features let a page do real work locally:

• File API: reads a file you picked or dropped, straight from your disk into the tab's memory.
• Canvas and image decoding: resizes, crops and re-encodes images to JPG, PNG or WebP.
• WebAssembly: runs compiled code (PDF engines, video codecs, compression libraries) at close to native speed.
• Web Crypto: generates truly random passwords and keys on your device.
• localStorage and IndexedDB: keep your settings and drafts on your device instead of in an account.

Put together, that's enough for a PDF merger, an image compressor or a password generator to work with no server at all. Mozilla's MDN Web Docs describe each of these APIs in detail.

The one-minute check

1. Open the tool's page on a laptop or desktop.
2. Open developer tools (F12, or Cmd+Option+I on a Mac) and click the Network tab.
3. Clear the list, then use the tool on a harmless test file.
4. Look for a request whose size is roughly the size of your file. A big POST or PUT means it was uploaded. A quiet list, or only tiny requests for analytics, means the work stayed local.

A second, rougher test: once the page has loaded, switch off Wi-Fi and try the tool. A purely client-side tool keeps working offline.

Stepping through a system one move at a time is often the quickest way to really get it. For that kind of learning elsewhere, ahaboo walks you through topics from tax brackets to plant biology using diagrams you control.

What "privacy-first" should mean

When a tool claims it doesn't upload your files, the Network tab should back that up. Tools that do need a server (anything using a hosted AI model, for example) should say so plainly and tell you how long they keep your data. The privacy-first shelf collects tools that describe themselves as working locally, and the AI tools shelf collects the ones that send text to a model by design.