Paperwork, ready to send.
Merge documents, tidy pages and shrink scanned PDFs for email.
Explore PDF toolsWhy Omnikit
Useful tools should make life easier. Fix your PDF, prepare your photo or build a file workflow, with no account and clearly labelled local or server processing.
Practical by design
Merge documents, tidy pages and shrink scanned PDFs for email.
Explore PDF toolsRemove backgrounds, set dimensions and choose a useful image format.
Explore image toolsKeep compatible tools together and pass the result straight to the next step.
Explore workflowsBrowser-local processing
Most tools process files in your browser. Word to PDF and Excel to PDF use temporary server uploads. Some need a first-use engine or AI model download; that brings software to your device and does not upload your file.
Saved workflows, favourites and recent tools stay in this browser. Optional reliability metrics record page views and tool outcomes, never file contents or filenames. The network monitor below shows requests observed in this tab; it is a practical readout of browser activity.
Live measurement
Verified live in this browser: not a single network request has been made since watching began.
These are downloads — code, fonts and on-device engines coming to your browser. None of them carries a file body away from it.
Nothing at all has been fetched since watching began. Not one request.
Your browser keeps its own log of every file a page fetches — the Resource Timing API. This panel reads that log and lists what it finds. On top of it, Omnikit wraps the four ways a page can push data outward (fetch, XMLHttpRequest, sendBeacon and WebSocket) so it can also report the HTTP method and the exact size of anything placed in a request body.
That is why the number above is a measurement rather than a promise. If a tool ever did post your file somewhere, a row would appear here with the method and the byte count, and the panel would turn amber.
What this is not. It is evidence from inside the page, not a formal security audit. It sees this tab only — not browser extensions, not other tabs, not requests your browser or network make on their own. A cross-origin server that does not send a Timing-Allow-Origin header can withhold its size figures, in which case we print “not disclosed” instead of guessing. Requests made before monitoring started are still listed, but their HTTP method is shown as unobserved.
The test you can run yourself: switch off your Wi-Fi, reload a tool you have opened before, and use it. If it still works, the processing was never happening anywhere but on your machine.
Know before you start
There is no daily usage allowance. Per-file safety limits keep accidental oversized selections from overwhelming a tab. Your browser, available memory and the complexity of the file also affect what can be processed.
| File type | Maximum per file |
|---|---|
| PDF document | 150 MB |
| image | 75.0 MB |
| audio file | 250 MB |
| video | 500 MB |
| text file | 25.0 MB |
| JSON file | 25.0 MB |
| CSV file | 50.0 MB |
| spreadsheet | 75.0 MB |
| document | 100 MB |
| archive | 250 MB |
Background removal exports PNG and preserves dimensions up to 4096 pixels on the longest side. Larger photos are resized. Scanned-PDF compression re-encodes pages as images, so text is no longer selectable. Check each tool’s explanation before processing.
Previously downloaded local tools and engines can work offline. Server conversion requires a connection. A tool or model you haven’t downloaded yet needs a connection first. Keep the tab open while a job runs, and save the result before closing it.