Tight!
Design and code,together.
The IDE for frontend teams. Design with AI on a canvas next to your code, and ship in the stack you already use.
Your AI
Bring the AI you already pay for
Sign in with your Claude Pro or ChatGPT plan, or add an API key. New accounts start with free Velork credits.
- Claude Code
- Codex
- Cursor
- GitHub Copilot
- OpenRouter
Agents
Your agent works on the canvas
Ask in plain words. It finds the frame, makes the change and leaves you a note.
The agent you already pay for. Claude Code, Codex, Cursor or Copilot run inside Velork, on your plan or your own key.
Hands on the real canvas. Agents read frames, find layers and change them where you can watch.
Notes you can answer. It leaves comments on the frame. Reply, resolve or ask for another pass.
The canvas
A design tool first
Auto layout, components, variables and a real pen tool. The AI works with them; you can always take over by hand.
Auto layout. Padding, gap and grid that work the way CSS does.
Components. Variants in one set. Edit the main and every instance follows.
Variables and modes. One design, light and dark, bound to the tokens in your code.
Snapping and measurements. Hold Option to read the distance to anything.
Pen and vectors. Bezier paths, boolean shapes and masks. Paste any SVG.
Typography. Any Google or local font, text styles and OpenType features.
Export anything. Images at any scale, and motion as video or Lottie.
Shortcuts you know. V, F, T, P and Cmd+K work the way your hands expect.
Performance
Fast with real files
A GPU canvas built for design systems with thousands of layers, not a web page pretending to be one.
- Pan and zoom

120 fps
With 10,000 layers on screen. On a MacBook Air M2.
- Open a file
0.8 s
To open a 400 MB design system. Cold start, local file.
- Images
5,000
Full-resolution images on one page. Loaded as you scroll to them.
Measured on Velork 0.9 with the public Atelier design system file.
Built in
Everything you'd leave for, built in
Browser, 3D, Figma and your team on one canvas. The AI works with all of them.
3D on the canvas. Ask the AI for a scene, or import a model. Light it and fly the camera.
Built-in browser. Open any site or your localhost. The AI clicks through, tests and fixes.
Captures keep their motion. Capture any site, or let the AI do it. The animations come along too.
Paste from Figma. Paste a frame from Figma. The AI builds it, layers and styles intact.
Live canvas sharing. Share a link to the canvas. Your team and their AI edit it live.
Design to code
From the frame to running code
Mention a frame and ask for code. Velork builds it with your components and opens the running page.
Draw on a real canvas. Layers, auto layout and components. Not a picture of an interface.
Shipped as your code. Components in the framework your repo uses, with the tokens it already has.
See it running. The built-in browser opens the real page next to the frame.
Only in Velork
Things no other canvas does
Built into the canvas, no plugins needed.
3D on the canvas. Build scenes, import models, light them and fly the camera, next to your 2D frames.
Export stills, video or the scene itself.
Motion to video. Keyframes and easing on a timeline, exported as MP4, ProRes or Lottie for your app.
MP4 · ProRes · Lottie · GIF
Captures that keep their motion. Bring any live site onto the canvas as layers, with its animations and hover states.
Paste a URL, or let the AI capture it for you.
Any AI, plus free credits. Claude Code, Codex, Cursor, Copilot, OpenRouter or your own key. New accounts start with credits.
Switch providers per chat. Keep your own keys.
Builders
What builders are saying
Notes from people who design and ship with Velork. Pick one up and move it around.
this is awesome! congrats to the team
3 minutes?? I spend longer naming the repo
Love the idea of having figma like design within my AI enabled IDE
It’s a great product, and I really appreciate it!, I’m just a bit torn between yours and Figma’s MCP, which I can directly plug into my Cursor and start using instantly. I’m eager to see what exciting differences you’re bringing to the table.
Looking good! Mind if I DM?
Looking good
Bro it's so clean, great work!
looks neat
It’s a great product, and I really appreciate it!, I’m just a bit torn between yours and Figma’s MCP, which I can directly plug into my Cursor and start using instantly. I’m eager to see what exciting differences you’re bringing to the table.
Love the idea of having figma like design within my AI enabled IDE
3 minutes?? I spend longer naming the repo
this is awesome! congrats to the team
Changelog
Shipped lately
Velork updates itself, so these are already on your machine. Full changelog
- November 18, 2026
Sign-in card after signing in. Signing in from the chat now takes you straight into the chat, with no second sign-in.
Read the notes - November 17, 2026
Velork AI stays in the model menu. Velork AI now stays in the model menu next to your own AI, and the model you are chatting with is never swapped for another one behind your back.
Read the notes - November 15, 2026
Spacing on the canvas. You can now change spacing right on the canvas: drag an auto layout frame's gaps and padding, or click a value and type a new one. Double-click takes you to the layer you meant, and Shift+A adds auto layout. Canvas scrolling on Windows and Linux now feels like the Mac. Old chats no longer undo newer work in your files when the app starts.
Read the notes
Your code
Where your work lives
Plain answers to what your security team will ask.
- Is my code uploaded anywhere?
- No. Velork is a desktop app that opens your repo where it already is. Agents read and edit files on your machine.
- Which AI sees my files?
- Only the provider you pick, on your plan or your key. Velork credits run through Velork's own API with no training on your data.
- What gets shared when I send a link?
- The canvas page you share, nothing else. Links can be private to your team, and you can turn them off at any time.
- Can I leave?
- Yes. The code is yours and already in your repo. Designs export as SVG, PNG and code.
