Why we build client-side first
If a tool can run in the user's browser, it probably should. Here's the rule, and the four times we've broken it.
We have one rule we keep coming back to: if the tool can run in the user's browser, it probably should.
That rule is why most Toolips utilities never touch our servers. You drop a PDF, we shrink it in your tab, you download it. No upload. No round-trip. No "your file will be available for 24 hours, lol."
The four reasons it's better
1. Privacy by default. Your tax return doesn't need to visit Estonia to get a watermark.
2. Speed. No upload time, no queue, no server roundtrip. Cold cache to compressed PDF in 800ms on a M2.
3. Cost. Egress bandwidth is the killer. Doing work on the client means we can charge $6/mo for Pro instead of $40.
4. Reliability. Your tab is up. Our server may not be. The tool still works.
When we break the rule
Four times. Be honest about it.
- AI tools. Claude doesn't run in your browser yet. So Pro AI features (paraphraser, hooks, captions) call our server through Efolusi Relay. We don't keep the prompts.
- Heavy ML. OCR via Tesseract works in-browser, but Whisper does not. STT goes to the server.
- Network scans. Your browser can't open a TCP socket. DNS, SSL, ports: server-side.
- Video downloads. YouTube blocks browser-only yt-dlp flows. So the downloaders were built to run as a job on our worker, upload to product-scoped private object storage, and hand you a signed URL. They are paused for now.
That's it. Everything else runs in your tab.
How to tell which is which
Look at the tool card. If it's labeled FREE, it almost certainly runs locally. If it's PRO and in the "AI" or "Security" categories, it's server-side.
We'll add a tiny icon for this in 2.1.