Introducing our updated MagicPath plugin for Cursor and Grok bot. Your agent gets a shared canvas for your code, builds interactive components, and now draws diagrams, flows, and wireframes you can edit after.

The agent reads your repository and generates components that match your codebase's style. Tweak them on the canvas, implement directly.

And with the new Figma import/export: Drop a Figma component in, your agent recognizes it as React code, make variations, send them back to Figma, or ship them.

In @bot just click on plugins in the left sidebar, then search for @MagicPathAI

For Cursor use this link: Install it, then try: "import my [component] into MagicPath and design 3 variations" and watch your agent taking control of the canvas.

Also works in Claude Code, in the terminal: claude plugin marketplace add MagicPathAI/magicpath-agent-plugin && claude plugin install magicpath@magicpath

Learn more about how to bring any third party agent to MagicPath here

Anywhere you need a visual layer, MP has got you covered.

This is super exciting!🙌

Thank you!

let's go!

A shared canvas the agent can actually draw on is overdue. Chat is a terrible place to review a flow. If the diagram is editable, you finally have a surface that is not another markdown dump.

an agent that actually draws the diagrams is huge. most enterprise architecture lives in a whiteboard photo from a 2023 meeting nobody remembers. the docs were always the part humans skipped.

will get my hands on it soon 🥹

Can't wait to hear your thoughts

The shared canvas concept with editable diagrams is such a smart upgrade

agent that codes and draws diagrams you can edit is huge

Does the agent read the diagrams back, or only draw them

Yes it reads it back!

Thats the useful half. drawing is easy, reading a wireframe back into components is where these usually stop

This is huge—having diagrams and flows directly on a shared canvas alongside code completely changes the game for spec-to-build workflows. Amazing update!

are you rendering those wireframes as editable SVG components on the canvas or parsing mermaid syntax under the hood?

Shared canvases make agent driven coding workflows much more visual

Drawing diagrams is a huge pain. How much manual prompting does it take to get a useful flow out of complex code? Or does it just know?