Codex can now use the in-app browser to test your app at different viewport sizes! It will control the device tool bar and click through your app at different breakpoints to validate & iterate. If it's a long run, Codex will take screenshots at key moments during testing and... show them to you at the end of the turn, so you can verify its work. To speed up testing, Codex can hide the IAB to disable animations, and accelerate testing by 1-2x. Separately, we also made annotations send faster and consume less tokens. Hope you enjoy the updates and let us know what you think!show more

Shoutout to @jnpdx & @olivertepman who cooked on this update

Damn - I think the browser is codex’s best feature tbh

Thanks! Big fan

Much as I love the sauce you be cookin'.... but Codex browser preview still does some awful, horrendous behaviors that just kills its usefulness right in the face. Like, say you get one of these nice next errors, and then you try to click the oh so helpful copy button. Whatcho play'n at son!?

Yikes! Thanks for reporting the bug, we have a big upgrade coming next week that should make the core browser experiences much much better. Let me know if you still see it then

Other major one... and maybe it's user-knumbskulliness... but no way to see a dev console. I don't need the whole shebang, just a console would be super helpful. Fully realize a web browser in essentially a web view can be a barrel of fun.

Wow that's slick. I would say the biggest downside to these browser tools is how slow they are so that is really helpful. Thank you

Amazing! Could you make file uploads work in the browser? Makes testing many features hard for us

Yep, working on bringing full feature capability to the browser!

Amazing! Browser use and the chrome plugin work really well.

I should be allowed to set my viewports in config.toml and the iframes should render on an infinite canvas using translate() css props for panning and zoom. Then each iframe should have it's own subagent which try to accomplish the task so all sizes are tested in parallel.

Good feedback! We need to get better at parallelization for browser use. For saving viewports, thoughts on just saving that in a skill instead?

A skill would probably be fine. But I think codex browser usage tools are undocumented so writing a skill to prompt an undocumented feature triggers iOS app store rejection PTSD.

@jxnlco omg this is so hugeeee😭😭 no more me manually editing piece by piece and telling codex its looking at different view resolutions

why does all the codex generated UI look the same. its getting boring. seriously need to get better frontend.

Say what? @VraserX

Can you please make codex detect .localhost domains 🙏 Current treats them as remote and doesn’t open them in the app browser like localhost:port or other ips.

@OpenAIDevs Feature is working well - seeing the screenshots as they come through was very neat. Validation of testing in progress is a nice feature

Hey @cursor_ai team this is something that you lost going to v3 that desperately needs to come back.. viewport size changing in the embedded browser.

@OpenAIDevs yes! lfg

it doesn't matter. codex suck at frontend taste

You are not wrong on this lol. Lots of folks working on this

Browser in codex still very underrated, damn, love it!

As someone who has done QA and workflow testing, this feels important. A lot of product work is not just “write the code.” It is checking layouts, breakpoints, screenshots, edge cases, and whether the app actually behaves like users expect. Coding agents entering this layer is a big shift.

Can it attach to external browser? Or detach from the Codex shell? Access to full dev tools would be nice, no? Am I missing something? I find the in-app annoying.

@OpenAIDevs Can it replace test cases already written in playwright selenium??

@OpenAIDevs It probably can? Try it and let me know!

@OpenAIDevs Sure,, will share the results soon.

this is exact update i am waiting for. thank you

Hey James, is the in-app browser supposed to work on external websites? On my system it works for locally hosted sites but it can't operate on external ones freely.

Not yet! In app browser is targeted for local for now, but we are making upgrades to make it good for both

Good to know thanks!

@OpenAIDevs Discovered this yesterday! Great work, I was waiting for this. Thnx for cooking 👏

Wen native mobile version of this

hiding animations to save tokens feels like peak 2026 engineering already

@ajambrosino does it work with remote Linux connections?

@ajambrosino Try it! I have seen people do some crazy remote stuff with the browser

vibes absolute vibes

codex app和 cli是相同的节奏吗？我更优先关注cli

Breakpoint regression testing has been the manual chore that keeps eating my Sunday evenings. Codex driving the device toolbar and snapshotting at key moments is exactly the loop I have been wanting.

@OpenAIDevs Wasn't it already doing for the past several days? I even saw screenshots it captured and saved during the QA cycles.

The screenshots are the key detail. Responsive bugs are rarely one assert; you need viewport metadata, the exact interaction path, and visual evidence so another agent can reproduce the failure instead of guessing.

I thought you guys were already doing this for a long time.

this is the threshold everyone feels when they see it live. once the agent can check the breakpoints itself, it stops being codegen theater and starts feeling like delegated work.

My kaspersky av blocks this, what s the solution, beside turning it off

How can I use it? It's not available in Chrome app store for me.

This is just the in-app browser, it does not require the chrome extension. For the chrome extension, if you are in the EU/UK, it is not available yet!

I hope he closes the session after he finished his work

Very cool. I’m having problems to use microphone with in browser app when I’m testing. Any help here?

@vidalnext Known bug! we are working on web permissions support