Your documents never
leave this tab
Sign, compress, convert, redact and scan — 59 tools that run on your device. No account, no server, nothing uploaded. Further down, one of them runs on a real file and counts every byte that leaves your machine.
Crops to the size a consulate asks for, at the resolution they print at.
Every engine below runs inside your browser — none of them is a service we send your file to
Nothing is sent anywhere
No cloud, no API, no analytics, no error reporter — not to us and not to anyone else. Turn your wifi off once the page has loaded and every tool still works.
There is nothing to sign up for
No account, no email, no cookie banner, because there is nothing to put in one. Files you choose to keep live in this browser and are yours to clear.
Even the AI runs here
Summaries, translation and background removal come from models that download once and then run on your machine. Your document is never the thing being uploaded.
Do more than one thing at a time
A single tool is a single tool. What people actually do is a sequence — read the scan, compress it, take the metadata off, file it — and doing that by hand four times a week is the part worth removing.
Pipelines
Chain the tools together
A template is a list of steps — scan, read the text, compress, strip the metadata — that runs top to bottom on whatever you hand it. Save one, and the whole sequence becomes one click.
Browse the templatesAutomations
Or let them run themselves
Point a rule at a trigger instead of a button. Every PDF you save gets compressed; anything with 'contract' in the name loses its metadata; a folder tidies itself once a day.
See the scheduling- Compress to 500 KB
- Strip the metadata
- File it in Contracts
The Library
Locked with a passphrase you keep
Everything you save lives in this browser. Set a passphrase and the rows and the file bytes are encrypted at rest with a key that exists only in memory — which also means we cannot help you recover it.
Open the LibrarySharing
Send a workflow, not a document
A template exports as a short code with no file in it — just the steps. Paste one and it is checked against the tools that actually exist before anything runs.
See how sharing workspw1_c2NhbitvY3IrY29tcHJlc3M- 3 steps
- 0 documents
- Checked against the real tools
Set it up once. Then it is one press.
Here is one running. Not a recording of one — press the button and a real PDF is built in this tab, the three steps run on it, and the card that comes back reports what actually happened, including how many bytes left your machine.
The template
Draft for review
Watermark it, number the pages, strip the metadata.
- 1WatermarkStamps text across every page
- 2Add page numbersNumbers each page in a corner
- 3Strip metadataRemoves author, producer and the XMP block
Press run, and this fills in.
A real PDF is built in this tab, the three steps run on it, and everything you see afterwards — including the bytes uploaded — is measured.
Rules that run the tools for you
Compress every PDF you save. Strip the metadata off anything with 'contract' in the name. Tidy a folder once a day. The steps are the same tools you would have clicked, running in the same worker, with nobody clicking.
See how the scheduling worksWhat it cannot do
A browser cannot run cron — a closed tab does not execute code. So rules catch up instead: one that came due while you were away runs the next time you open the app, once, not once per missed day.
Keep what you want, lock it if you like
Save results into folders, reopen them in another tool, and keep a history of what was run. It is SQLite in a worker with the file bytes in your browser's own filesystem — the same storage a photo app would use.
Open the LibraryWhat it cannot do
Set a passphrase and the rows and bytes are encrypted at rest with a key held only in memory. Forget it and there is no recovery, because there is no one to recover it from. The page says so before you set one.
Describe the job, watch it run
Ask for “split this into pages and compress each one” and it assembles the steps from the same registry the sidebar is built from, then runs them here. Every action it takes is one you could have taken by hand.
Open QuillWhat it cannot do
The model runs on your device or in your browser's own built-in AI. Your document is never the thing being sent — the only download is the model's weights, once, and only if you open a tool that needs them.
Primary benefits
What you get for giving up the server
Losing the backend costs a few things and buys rather more. Here is the trade, in full.
Nothing waits on a network
The file is already where the work happens, so there is no upload and no queue.
- A 40 MB merge starts instantly
- No size cap beyond your own memory
- No tier, because there is no meter
Private by construction
Not a policy we follow — a transfer that does not exist, enforced by your own browser.
- No account, so nothing to breach
- A CSP naming every host we may contact
- Measured on the front page, not asserted
It still works with the wifi off
A service worker keeps the whole app in your browser's cache after the first visit.
- Every tool runs offline
- Reloads come from your disk
- Add it to a home screen if you want an icon
Real engines, not approximations
The same libraries a desktop tool would use, compiled to WebAssembly and run here.
- Ghostscript for compression and repair
- Tesseract for reading scans
- pdf.js and pdf-lib for everything else
A Library instead of an account
Folders, history and results in this browser's own storage, under a passphrase if you want one.
- SQLite in a worker, files in the origin's filesystem
- AES-GCM at rest with a key held only in memory
- Nothing syncs, so nothing leaks
It can do the work without you
Templates chain the tools; rules run them on a trigger or a schedule.
- Eighteen ready-made workflows
- Share one as a code with no document in it
- Catch-up scheduling, because a browser has no cron
59 tools, none of them a round trip
Documents and photographs, from a two-second merge to a passport photo at 600 dpi. Every one of them runs in a worker on this device.
Organize
Edit
Rotate or remove pages, add text and shapes, and save a copy.
PDFDraw a crop box to trim margins on one page or all pages.
PDFStamp a text or image watermark on every page with a live preview.
PDFDraw your signature, drop it where it belongs and download the signed PDF.
PDFFill the fields of an interactive PDF form and optionally flatten it.
PDFStamp page numbers in any corner, starting from any number.
PDFOptimize
Secure
Convert
Turn a PDF into PNG or JPG images, a Word document or an Excel sheet without uploading it.
PDFCombine PNG and JPEG images into one PDF, one image per page.
Picture · severalUse your camera to capture pages straight into a PDF.
Picture · severalPhotograph a page, mark its corners, and get a straightened, cleaned-up scan or a multi-page PDF.
PicturePhotos & images
Crop a photo to Nepali PP, passport or visa size at print resolution, and lay out a sheet to print and cut.
PictureLay out a staff or student card at exact ID-1 (85.6×54 mm) size, front and back, and export it print-ready.
PictureDesign a business card at any standard size and export a print-ready PDF with bleed and crop marks.
PictureMake a QR for a link, wifi network or contact card, or read one out of a picture.
PictureTrim a photo to any ratio with a draggable crop box.
PictureTurn a photo in 90° steps or mirror it, then save it back out.
PictureSet exact pixel dimensions or scale by a percentage, with proportions locked.
PictureShrink a photo to fit an upload limit, lowering quality only as far as it has to.
PictureChange an image's format on this device, with a white ground when dropping transparency.
PictureCut the subject out of a photo with an on-device model.
PictureBlur, pixelate or black out any part of a picture, with an on-device face finder to propose the areas.
PictureEnlarge a picture with a super-resolution model that downloads once and runs on this device.
PictureAdd text, shapes and stickers to a photo and tune its colour, all on this device.
PictureCaption a picture top and bottom in the classic style.
PictureStamp text across a photo, once or tiled over the whole frame, on this device.
PicturePut five markers on a face and measure its proportions against φ and against the values that scored best when they were tested.
PictureMeasures the darkest and brightest tones in your photo and stretches them back to the ends of the range.
PictureTune brightness, contrast and colour, or apply black & white, sepia and document looks.
PictureBrush over a blemish, a stray hair or a speck of dust and the surrounding skin grows back over it.
PictureText & AI
Read the number, dates, totals and line items off an invoice, bill, statement or payslip, check them, and export to Excel, CSV or JSON.
PDFPull text out of scanned PDFs with on-device optical character recognition.
PDFGet a summary of any PDF from an AI model that runs.
PDFTranslate a PDF's text into another language with a model that runs.
PDFMake a document
Fill in your client and line items and get a print-ready invoice.
PDFEnter earnings and deductions and get a payslip with the net pay worked out.
PDFFill in your experience and download a one-page CV in two designs.
PDFQuote a price with line items, tax and a validity date.
PDFOrder goods or services with quantities and prices, ready to send.
PDFList what was delivered, with quantities and no prices.
PDFAcknowledge rent received, with the period, property and amount.
PDFList what is in the box with quantities, ready to pack.
PDFWrite the letter that goes with your CV, on a clean page.
PDFOffer a job with the role, salary and start date set out.
PDFCertify that someone worked for you, and for how long.
PDFState an employee's salary for a bank, landlord or embassy.
PDFConfirm someone has been relieved of their duties and settled up.
PDFRecommend someone for a job, a course or a tenancy.
PDFSet income against expenses for a month and see the balance.
PDFInvite people to an event with the date, time and place.
PDFDesign a certificate on a canvas and print it landscape, ready to frame.
PDFDesign an A3 poster or shop placard on a canvas, built to read from across a room.
PDFThe evidence
Don't believe us. Here is the browser's own record.
Every privacy promise on the internet is worth what you can verify of it, and no more. What follows is not a screenshot and not an illustration — it is performance.getEntriesByType("resource"), the log your browser keeps of every request this page made, plus a count of the bytes leaving through the only four APIs that can carry any.
Measured live in this tab
0 Bytes
uploaded from this page, since it loaded
- 0
- Requests seen
- 0
- Left this site
- 0
- Carried a body
Try it against a real operation
Builds a PDF and stamps it, then reports what crossed the network.
Reading the browser's record…
If this stays empty, your browser has not exposed Resource Timing to the page. Your own devtools still will.
This counts what this page does. Code in a worker has its own network stack and would not appear here — that gap is closed by the Content-Security-Policy below, which your browser applies to workers too, and by your own devtools, which see every context. This is the quick check, not the last word, and the last word was never going to be something we wrote.
Those /_next/ requests are not your file
Why measuring the body is the whole trick
fetch, XMLHttpRequest, sendBeacon and a WebSocket. All four are wrapped before the app starts and every byte handed to them is counted. An upload we could not measure would be reported as unknown, not as zero — the flattering rounding is the one thing a proof like this must never do.The two exceptions, stated plainly
And it is enforced, not promised
connect-src 'self' blob: https://huggingface.co https://*.huggingface.co https://hf.co https://*.hf.coFive checks, about a minute each
This is the part of the page where a product normally shows you strangers saying nice things about it. On this particular claim a stranger's opinion is not evidence. These are, and none of them requires trusting us.
Watch the network panel
F12, the Network tab, then sign the document above. No request appears, because none is made.
20 secondsPull the wifi out
Load the page, go offline, then merge, compress, sign, redact — anything. It all still works, because it always ran here.
1 minuteRead the response headers
The Content-Security-Policy names every host this site may contact. Your browser enforces it whatever we intended.
30 secondsLook in Application → Storage
Anything you saved to the Library is there, on your disk, under your control, and Clear site data really does clear it.
30 secondsFilter the whole session by POST
Tick Preserve log, use the app for as long as you like, then filter by method:POST. The list stays empty.
as long as you likeThe difference is architectural, not editorial
Nothing below is an accusation about anyone. If a site does the work on a server then your file has to get to that server, and everything in the right-hand column follows from that one fact.
| On this site | A tool that uploads | |
|---|---|---|
| Where your file is processed | This tab | Their servers |
| Copies of it on someone else's disk | None | At least one, for a while |
| Needs an account | No | Usually |
| Works with no internet | Yes, after the first load | No |
| Size and page limits | Whatever your machine can hold | Set by your plan |
| What happens on a breach | There is nothing of yours to leak | Depends on their retention |
Questions
The questions people actually ask
Including the ones with an answer we would rather not have to give.
If nothing is uploaded, what are all those requests in my network tab?
The program arriving. A page that does something has to load its own JavaScript, styles and — here — the PDF and image engines that do the work. Every one is a GET or a HEAD to this same site: the two verbs that cannot carry a body, so there is nowhere in one for a document to be. The panel above lists them as the browser recorded them, with the number of bytes each one sent.
Can you get rid of those too?
Not entirely — the code has to reach you once. After that a service worker keeps the app in your browser's cache and serves it from there, which is why the site keeps working with the network off. That is a better test than any sentence on this page: load it, disconnect, keep using it.
What about the AI features?
They are the one exception and they are opt-in by opening the tool. If your browser has a built-in on-device model it is used. Otherwise the model's weights download once from huggingface.co and are cached. That is a download to you; your document is not part of it, and nothing happens at all unless you open Summarise, Translate or a tool that removes backgrounds.
Where do my files actually live?
In this browser, on this device, only if you press Save to Library. Bytes go in the origin's private filesystem and metadata into a local SQLite database. Clearing site data removes both. Nothing syncs anywhere, which also means a file saved on your laptop is not on your phone.
Is it as good as the server version?
For the work here, generally better: there is no upload wait, no queue and no size cap beyond your own memory. The real engines are compiled to WebAssembly, so compression is Ghostscript and OCR is Tesseract, not approximations of them. A very large file is limited by your machine rather than by a plan.
What is the catch?
Three, stated plainly. Browser storage can be cleared by you or evicted by the browser, so download anything you must keep. The first visit downloads more than a thin client would, because the engines come with it. And AI models are large, so the first run of those tools takes a while on a slow connection.
Your device. Your files. Your privacy.
Nothing to sign up for and no server to trust. Keep what you want in the Library, lock it with a passphrase, and clear all of it whenever you like.