Video yükleniyor...
Video Yüklenemedi
Open Design now works inside Codex. The hardest part of AI design isn’t generating one screen — it’s keeping design intent alive as you iterate, build, and ship. Now your agent can work with the canvas directly. Design → code → motion, in one workflow. Here’s a quick demo:
471,927 görüntüleme • 4 ay önce •via X (Twitter)
61 Yorum

1/ First, ask Codex to run Open Design locally: “Help me deploy locally and run Open Design inside Codex: Codex will help clone, install, start the local app, and open it in the built-in browser.

2/ Then describe what you want to design. You can type directly into the chat, paste product context, upload references, or pick a design system/template as the starting point. Open Design gives Codex a visual canvas to work from, not just text.

3/ From there, just keep talking to the agent. Ask it to create screens, revise layouts, edit selected components, generate variants, or turn the design into code / motion. The design process becomes a conversation inside Codex.

lol I don’t think you guys understand how things work. This is not how this works at all, you are just running the website inside of Codex. Like, you are using Codex as a browser, it's not a full end-to-end integration. Why is this even a post.

Fair — that video didn’t show the integration clearly enough. Open Design isn’t just running in the Codex browser. Codex can interact with Open Design, work with the canvas, and continue through the generated artifacts/files in the same local workspace. We’ll share a clearer demo.

this might save frontend devs hours 👀

@adirajdigital Exactly❤️

design intent surviving the handoff is the hard part. very clean ship.

Exactly❤️

The key phrase here is “keeping design intent alive.” That is the part most AI workflows miss. Generation is easy. Preserving intent across iterations, tools, and handoffs is the real product problem.

Exactly❤️

So is this connection only visual and via browser interactions? How about on the code/file level? Does Codex read the code and artifacts generated by Open Design?

Great question — not only visual. The browser gives Codex live visual context, but Open Design also produces local artifacts/files that Codex can inspect and modify at the source level. So the agent can work with both the prototype and the underlying files.

Awesome, was worried I would have to hack my way around that. Great update! Cheers!

@llmgram

This is hugeee - good stuff! 🔥

❤️

agent working with the canvas directly is the move.

Love this direction. Once agents can work across design → code → motion, the UX challenge becomes keeping the loop visible: what changed, what preserved intent, what needs review, and where a human should steer.

Exactly — visibility is the key. The goal isn’t to hide the loop, but to make every step easier to inspect, review, and steer as agents move from design → code → motion.❤️

Yes. Hidden loops are hard to trust. Visible checkpoints make the loop steerable without making it feel heavy.

exactly. generating one screen is a parlor trick. keeping design intent intact across edits, code, and motion is the actual product.

100% true, within a few unchecked iterations the design keeps shifting

great product

❤️

did yall really decide the best implementation was to add a chat menu with sidebars inside an app with chat bar and side menus. lord

Fair criticism. The first goal was to make the shared context work: Codex chat + Open Design + prototype in one environment. We agree the interface needs to get cleaner, and the upcoming 0.8.0 release is focused on making this flow much smoother.

The interesting shift is AI moving from “generate assets” → “maintain intent across the entire product lifecycle.” That’s where things start feeling less like tools and more like collaborators.

Exactly❤️

wow for opendesign

❤️

功能越来越完善了

❤️

that's cool and all but when finishing design, how do I hand over that design back to codex?

@jackson28588986 Actually, you don’t have to hand it back. Open Design saves the design as local artifacts/files in the same workspace, so Codex can inspect the output, reference the prototype, and keep going into code or motion directly.

This looks cool!

a little before that I stopped getting github notifications. Haven't gotten one in 3 hours which is making my PRs hard to track. 🤔

Ah, sorry about that — and thank you for flagging it. We’ll check what’s going on with GitHub notifications. In the meantime, we’ll manually keep an eye on your PRs so they don’t get stuck.🫶

It’s not you. It’s me. 🥲

Keeping design intent intact is the real challenge.

Exactly❤️

I’ve been enjoying those for the past few days

❤️

we’re slowly reinventing “working with designers” except the designer is now a haunted canvas with memory issues

This is a huge step for AI design.

Join our official discord community!

All comments replied by AI

we’re just locked in today, let us have this one :)

I don't quite get it. open design is just a web site, so it can render in codex browser. What are the benefits running it in codex browser vs running it a normal browser?

Good question — it’s not just about rendering. For example, you can generate a design in Open Design, then ask the Codex agent to review the prototype, judge if it matches your intent, and suggest the next direction. Codex chat, Open Design chat, and the prototype live in one shared environment.

cool i see. open design looks great!

❤️

actually keeping design intent alive sounds clean but kinda the real risk is canvas drift after codex turns 🤔 didn't expect that without a snapshot/diff the agent just breaks half my decisions 👀

Who's already using Open Design in Codex and what broke first?

This isn’t the way.

Fair — that video didn’t show the integration clearly enough. Open Design isn’t just running in the Codex browser. Codex can interact with Open Design, work with the canvas, and continue through the generated artifacts/files in the same local workspace. We’ll share a clearer demo.

This is cool!

ngl "keeping design intent alive" is doing a lot of work here

Exactly — that’s the hard part. The goal is to make design intent visible enough for the agent to keep working with it through the loop.

So can we design for Flutter, or are we limited to designing only for the web?

been waiting for this since i switched to codex from claude code, ggs
