How it works

The work happens here.

Kagaz is a static website. There is no server that receives your documents, because there is no server doing any work at all.

What happens when you add a file

Choosing a file hands it to the page through the browser's own file picker. The page reads it into memory in the tab you are looking at, edits it there with two open source libraries — pdf-lib for writing PDFs and pdf.jsfor reading and drawing them — and hands the result back as a download. The file is never copied anywhere else. Close the tab and it is gone from memory.

Even the fonts and the pdf.js worker are served from this site rather than a CDN, because a request to someone else's server is still a request to someone else's server.

How to check this yourself

You do not have to take our word for it, and you should not. Three ways to confirm it:

1. Read the counter. The strip pinned to the bottom of every page reports two numbers live. Uploaded comes from wrapping fetch, XHR and sendBeacon and adding up the bytes of every request body. Offsite uses the browser's own performance timeline to count requests to any origin that is not this site. A plain request count is deliberately not shown: loading our own fonts is a request too, and a number that alarms on every normal page load would only teach you to ignore it.
2. Know what the counter cannot see. A Web Worker keeps its own performance timeline, so requests made inside one are not included in the offsite figure. Kagaz has exactly one of those: the OCR tool's worker fetching its engine and language model, from this same site. Every other tool makes no requests at all. We would rather tell you where the measurement ends than let you over-trust it.
3. Watch the network tab. Press F12, open Network, and use any tool. You will see the page's own scripts and fonts load, and nothing after that.
4. Pull the plug. Load a tool, turn off your Wi‑Fi, then use it. It still works, because nothing it needs is anywhere else.

What it costs

Nothing, and there is no plan where that changes. No account, no daily limit, no watermark, no pass to buy on the fourth file. The processing power is your own device for a second or two, so there is no bill to recover and nothing to meter.

Where it is weaker

Being straight about the trade-offs, because a tool that hides them wastes your time:

• Big files depend on your device. Memory is your laptop's, not a server's. A few hundred pages is comfortable; a thousand-page scan on an old phone may not be.
• Compress and Grayscale are lossy. They rebuild each page as an image, so the text in the output is no longer selectable. Both tools say so on their own pages before you run them.
• There is no PDF to Word or Excel. Doing that well needs far more than a browser can offer, and a bad conversion is worse than none, so Kagaz does not pretend to.
• Password-protected files have to be unlocked first. Kagaz can read them only once you have removed the password yourself.

What is here

51 tools at the moment, and more being added. See the full list.

Who made it

Kagaz is built by Angad Jangir at IIT Madras, and released under the MIT licence.