What changed
Bolt.new added Visual Edits, a click-and-change editing layer that lets you adjust text and styling directly in your running app instead of writing a prompt for every small tweak. Bolt.new is a browser-based AI app builder from StackBlitz: you describe what you want in plain English, and it writes and runs the code for you.
As of September 3, 2026, Visual Edits is live for every Bolt.new user on both free and paid plans.
The feature lives on the Select button you already use to click elements in your app preview, now marked “new.” Click it, select an element, and a toolbar appears above it. Elements with text get a bigger toolbar: font size by the exact pixel or from a size picker, styling options, list and link formatting, and in-place copy editing using the keyboard shortcuts you already know (like bold or italic). The color picker reads its options straight from your app’s own colors, so you’re always choosing from shades already in use instead of typing in a color value and hoping it matches. Selected elements also show plain-language names instead of raw code terms like “div” or “h1” (the generic building-block tags a browser uses to lay out a page), so you don’t need to read any code to know what you clicked.
Changes don’t fire one at a time. Each approved change stacks into a list instead of sending immediately. Re-select an element you already changed and adjust it again, and it updates that same entry rather than adding a second one.
Don’t like a change? Discard it and the preview snaps back. When you’re done stacking, hit save once, and the whole batch goes to the model together.
The feature covers text and styling only. You can’t drag elements to rearrange a layout, and applying one change across many elements at once isn’t reliable yet. The prompt box is still there for anything the toolbar doesn’t reach.
Why it matters
Small visual fixes, like a bigger headline, a different button color, or punchier subhead copy, used to mean the same three-step loop as any other change in Bolt: write a prompt, wait for the build, check the result. That loop makes sense for changes that need real judgment. It’s overkill for “make this text 24px” or “use the blue we’re already using on the button.”
Visual Edits gives the model exact styling values and a selected element. Instead of describing a color in words and hoping the model picks the shade you meant, you pick it directly from a palette pulled off your own app. Instead of explaining “the headline, not the subhead,” you click the exact element. Batching multiple changes into one save also means one round trip to the model instead of three or four separate prompts, each waiting on the last.
There’s a real cost to know about before you lean on this. Bolt says each saved batch “spends tokens the way a prompt does.” Tokens are the usage units AI models charge for, the same meter that tracks every prompt you send.
Bolt says a saved batch uses fewer tokens than describing each change separately in chat. It still consumes your usage allowance. Check the saved batch’s measured usage on your plan. If you’re watching your monthly usage allowance closely, stacking five small tweaks into one save still consumes tokens, including on the free plan.
Who should care
This matters most if you’re the kind of builder who does a pass of small polish after the big structural work is done, tightening copy, matching a client’s brand colors, bumping font sizes until something reads right. That’s exactly the kind of work that used to eat several slow prompt-and-wait cycles for changes that don’t need the model to think hard.
It matters less if most of your changes involve new logic, new data, or restructuring how a page is laid out. Visual Edits explicitly doesn’t handle rearranging elements, so anything beyond styling and text still goes through the regular prompt box.
What builders should do next
Use Visual Edits for small text and styling changes, and check usage if token cost matters.
End of article