The screenshot folder
You find a pricing section you love. You screenshot it. Three months later you have a folder of 4,000 PNGs named "Screenshot 2026-04-11 at 14.22.03.png," and no idea which site any of them came from, what fonts they used, or how the layout actually worked. You know the feeling. Everyone who has ever redesigned anything knows the feeling.
That folder is the whole reason Nisaba exists.
What it actually is
Nisaba is a desktop browser, but it's not trying to be your daily driver. It's the one you open when your reference library needs to grow. You browse normally, and every capture keeps its source URL, its DOM, its CSS, its fonts, its design tokens, and its provenance. The library gets more useful the longer you use it, and when you want a pattern back, an agent turns it into real code in your own stack.
Four things it does, in practice:
- Browse — every site opens in its own fully isolated view. No Node, no preload bridge, every permission denied by default. It's a browser for looking at other people's work, so it treats that work as hostile by default.
- Capture — viewport, full scrollable page, a dragged region, or one picked element. Annotate with arrows, highlights, blur, callouts — all stored as editable vectors so the original screenshot is never touched.
- Extract — hover the page and the DOM outlines itself. Arrow keys walk parent, child, sibling. Click, and Nisaba keeps the selector, the sanitized HTML, computed styles, CSS variables, fonts, palette, assets, and a framework guess with the evidence behind it.
- Audit — click your way down a live page or localhost, pin a note to anything that needs fixing, and it remembers exactly where that element lives and greps your workspace for the file that renders it. Export the whole audit as a stack of tasks an agent can just work through.
Why I built it this way
I didn't want another screenshot tool. I wanted something that made the next project faster because of everything I'd already looked at. A normal screenshot is a dead end — a flat image with no memory of where it came from. Nisaba's whole bet is that browsing should leave something behind: not a picture of a website, but a small structured fact about how it was built.
It took about a dozen point releases to get here — notarization headaches, a Rosetta trap, icon geometry that looked fine until it didn't, dialogs that used to shout instead of just showing up. None of that is interesting on its own. What's interesting is that none of it touched the core idea, because the core idea was right from the start: capture should compound.
Where it's at
Nisaba just hit v1.0.9 and it's macOS-native, local-first, no account required. It ships with an agent-driven audit flow, a design-token profiler, and an update pipeline that finally works on the first click instead of the third. If you're the kind of person with a folder of unlabeled screenshots — and you know if you are — it's worth a look.