Rayzia AI Agent Vector Editor - Draw with AI. Edit by hand.
by•
Connect Claude code and watch it draw, recolour and edit directly on your canvas. Every shape remains fully editable.
Replies
Best
Maker
📌
Hi Product Hunt! 👋
I'm the maker of Rayzia — a vector editor that works, fully, hand-in-hand with an AI agent. And I believe it solves a problem the whole design world is stuck on.
Right now, almost every AI design tool does one thing: it generates a flat image and stops. What you get back is hard to change — you can't grab one shape, can't fix one curve, can't build on it. The AI hands you a picture, and you're on your own.
Rayzia solves exactly that.
Instead of generating an image, the AI agent works inside the editor, alongside you. Ask it to draw a logo — and it draws a real, editable logo. Then you step in and refine it by hand: move a node, change a color, reshape a curve.
Then hand it back to the AI. Then take over again.
It's a seamless loop — human and AI editing the same file, together, with no wall between them.
And the agent isn't limited to a few tricks: it can pick up every tool you can — pen, shapes, gradients, layers, effects, undo, redo — and draw with them directly on the canvas.
In the demo, Claude Code connects through MCP and draws directly on the canvas. It creates real vector shapes, paths, and objects — everything remains editable, just like something a human designer created.
AI doesn't replace the designer. It becomes a creative partner.
I built every part of this myself, from the editor engine to the AI integration. It's still early, and I know there are rough edges — that's why your feedback means a lot.
Thank you for taking a look. 🙏
Report
Tried it with Claude and watched it place shapes on the canvas, then I could grab any of them and tweak them like normal. The editable shapes part is the real win here, since most AI tools just spit out a flat image.
Report
Maker
@mariefisheou3r Thanks Marie! That was the whole point for us — an AI tool that hands you a flat PNG isn't a design tool, it's a screenshot. Everything Claude draws stays real vector layers you can grab, reshape and restyle like anything you drew yourself. Really glad the editable part landed for you 🙌
Report
The live connection with Claude is genuinely cool, and being able to keep shapes editable after the AI touches them is a huge plus. One thing that would make this even better for me would be a layer history panel so I can see and revert each AI-driven edit step by step instead of losing track of what changed when the agent makes several adjustments in a row.
Report
Maker
@sleymantalnkyt Appreciate this a lot, Süleyman 🙏 Quick tip in the meantime: there's already an undo-history panel (Edit ▸ Undo History) that lists every step and lets you click any one to jump straight back — so nothing's ever truly lost.
Where you're spot on is the granularity: when the agent fires several edits in one go they currently collapse into a single step, and the AI steps aren't labeled with what they touched. Making each AI action show up on its own — "moved Roof", "recolored Door" — so you can walk back through them step by step is exactly what we're building next. Thanks for putting your finger on it.
Replies
Hi Product Hunt! 👋
I'm the maker of Rayzia — a vector editor that works, fully, hand-in-hand with an AI agent. And I believe it solves a problem the whole design world is stuck on.
Right now, almost every AI design tool does one thing: it generates a flat image and stops. What you get back is hard to change — you can't grab one shape, can't fix one curve, can't build on it. The AI hands you a picture, and you're on your own.
Rayzia solves exactly that.
Instead of generating an image, the AI agent works inside the editor, alongside you. Ask it to draw a logo — and it draws a real, editable logo. Then you step in and refine it by hand: move a node, change a color, reshape a curve.
Then hand it back to the AI. Then take over again.
It's a seamless loop — human and AI editing the same file, together, with no wall between them.
And the agent isn't limited to a few tricks: it can pick up every tool you can — pen, shapes, gradients, layers, effects, undo, redo — and draw with them directly on the canvas.
In the demo, Claude Code connects through MCP and draws directly on the canvas. It creates real vector shapes, paths, and objects — everything remains editable, just like something a human designer created.
AI doesn't replace the designer.
It becomes a creative partner.
I built every part of this myself, from the editor engine to the AI integration. It's still early, and I know there are rough edges — that's why your feedback means a lot.
Thank you for taking a look. 🙏
Tried it with Claude and watched it place shapes on the canvas, then I could grab any of them and tweak them like normal. The editable shapes part is the real win here, since most AI tools just spit out a flat image.
@mariefisheou3r Thanks Marie! That was the whole point for us — an AI tool that hands you a flat PNG isn't a design tool, it's a screenshot. Everything Claude draws stays real vector layers you can grab, reshape and restyle like anything you drew yourself. Really glad the editable part landed for you 🙌
The live connection with Claude is genuinely cool, and being able to keep shapes editable after the AI touches them is a huge plus. One thing that would make this even better for me would be a layer history panel so I can see and revert each AI-driven edit step by step instead of losing track of what changed when the agent makes several adjustments in a row.
@sleymantalnkyt Appreciate this a lot, Süleyman 🙏 Quick tip in the meantime: there's already an undo-history panel (Edit ▸ Undo History) that lists every step and lets you click any one to jump straight back — so nothing's ever truly lost.
Where you're spot on is the granularity: when the agent fires several edits in one go they currently collapse into a single step, and the AI steps aren't labeled with what they touched. Making each AI action show up on its own — "moved Roof", "recolored Door" — so you can walk back through them step by step is exactly what we're building next. Thanks for putting your finger on it.