You can now drag images directly into your live prototype in MagicPath while in edit mode.
This is a first for AI made apps and websites, we are very excited about this feature!
Simply enter edit mode and drag any image you want on a frame like you see here!
And it plays nice with your agents. Build from Codex, Cursor, or Claude Code, then fine-tune everything visually, like a Figma file that ships.
Available on all plans today.
GPT-5.6 Sol is now available in the MagicPath model selector.
We decided to go with the smartest model, Sol.
This model represents a step change in front-end design. We think you’re all going to love it!
Anthropic's Sonnet 5 is now available in MagicPath to choose from in our model selection.
We found this model has a very similar taste to the bigger Opus, but faster and cheaper, meaning more generations for you.
Codex can import UI from your repo into MagicPath and make it editable.
It understands your components, design system, and interactions — then recreates the experience on the canvas.
Here’s our full left-side panel, fully functional...
It's now easier to import your GitHub repository in MagicPath!
Update the MagicPath skill for your agent (Codex,Cursor, Claude Code) to import and build from an existing GitHub repository in MagicPath 👇
Time to watch experts in their craft go AI native and showcase their design workflows of choice (Codex, Claude Code, Cursor) all side by side with MagicPath.
Match 1 (hosted by Josh Loh of The Arena) starts today @ 10am PDT
Watch live in 30min below...
GPT-5.5 is now available in MagicPath via model selection.
From our testing, this model is incredible at complex interaction and creating/manipulating 3D objects.
@OpenAI most powerful model yet, GPT-5.4 Thinking, is now available in MagicPath.
Select it from the model switcher in the chat toolbar to start using it right away.
Gemini 3.1 Pro is now available in MagicPath as a model to choose as a design agent.
We find this model to be really good at design, particularly at recreating code from images, probably the best at this task right now, like in this example.
Claude Sonnet 4.6 is now available in MagicPath.
This model is particularly strong when editing designs, using mentions, or for generating variations and flows.
Our MagicPath Chrome extension that lets you copy and paste real components from the web to further design or remix just crossed 10,000 users!
Thanks everyone for downloading it.
So many updates to come.
With Prompt Styles, you can explore a much wider range of design ideas. It also doubles as an educational tool, helping you learn how to prompt AI correctly for design.
We all know prompting is hard, so we make it easier for you.
Introducing Prompt Styles
Set the creative direction before you type a single word. Prompt Styles let you choose from 32 curated design aesthetics, and every component you generate follows that visual language.
Starting the week with some amazing new updates into our manual editor.
1) Scale
You can now press “K” when selecting an element to edit to bring up the scaling option. You can manually select the dimension or choose from a dropdown. 🧵
To start:
1) Open any project and click Import from Figma, or start from the landing page
2) Connect your Figma account.
3) Copy any design with ⌘L on Mac or Ctrl + L on Windows
4) Paste it into MagicPath with ⌘V
Images, typography, colors, and layout are all preserved.
No MCP servers. No plugin hell. No "close enough" conversions.
Just copy your design in Figma (⌘L) and paste it into MagicPath (⌘V).
Every pixel. Every spacing detail. Every color value. Preserved
We just added completion sounds to MagicPath.
When a design finishes, you’ll hear a sound 🔊
Enable or disable it in settings and choose from 4 different effects.
Grab this amazing component library made by @adrianoreis, packed with all the latest @tailwindcss components.
It’s incredible to see so many real, interactive components all in one place, ready to use. 👇
To start, just drag images onto the canvas, or hit the image button in the toolbar.
Just like components, you can reference images in chat (and mix them with components too). We’re building the interface for working with code, design and AI on a single canvas.
Small but powerful update ✨
Introducing text gradients in MagicPath.
Now, in the editor, any text can be filled with a gradient and if AI generates gradient text, you can edit it too.
Our Chrome extension that captures web UI elements and turns them into pixel-perfect React components just crossed 4,000 users 🎉
We’re already working on the next update, what features would you love to see?
Then use Nano Banana Pro (we are using @GoogleAIStudio here), but it will work anywhere you can access the model.
Then attach your exported image with this prompt:
We are currently testing which model is better at recreating code from images in our app.
Which one do you think is Gemini 3 Pro and which one is Opus 4.5?
Original image in the comments.