Codex can now take on more of your browser dev work. With the new Chrome plugin in the Codex app, it can test web apps, gather context across tabs, use web DevTools efficiently in parallel, and keep results organized without taking over your browser.

Oh yay, more trash. This is why your company is failing. You literally have the best model ever made and what are you doing with it? Oh yeah, you’re keeping it away from your users. Just give us back 4o. Instead of listening to bots in your comments, try listening to your actual customers #keep4o

not available in the UK or EU, i managed to download it just with VPN in US

When will codex get an iOS app?

Codex team cooked! Great and very useful plugin! 🔥🔥🔥

😢

Updated this morning but can't seem to find the plugin, will it be in the next release?

Gathering context across tabs is lowkey one of the most useful parts here 👀

already doing this with Claude Code cron loops — auto-triage issues, draft fixes, await human approval. the hard part isn't spawning agents per issue, it's teaching them when NOT to touch something. half the value is an agent that says 'this needs a human decision' instead of shipping a wrong fix.

I can’t find Chrome extension on my codex app.

Humans leak patterns.”

Browser-resident coding agents pressure the IDE-as-control-plane assumption. Endpoint security teams are about to have a lot of fun reasoning about agents with full DOM access in prod.

Just to verify, if I have 1 codex project chat doing something in a tab. I can have another codex project chat doing something else in a different tab and I don’t need to worry about them “bumping into each other” or cross contamination between there processes?

chrome plugin agents finally usable looking, claude version was too slow for me to keep around. let’s see if the codex one might actually live in my workflow

No Chrome for Codex for repressed group of people. @sama why are you letting your team advertise things without telling the whole truth?

Crying in corner

codex browser work is scarily fast when it uses scripts.. much faster then i or anone else can click click around, lol

AI coding agents are moving far beyond autocomplete. They are now testing apps, inspecting DevTools, and gathering context across tabs like real developers working inside a workflow. This is beginning to resemble the work of a junior developer more than a traditional coding assistant, @OpenAIDevs.

Amazing Dominik!

AI coding agents are getting much closer to handling real browser workflows end-to-end. Testing apps, inspecting DevTools, collecting context across tabs… This is starting to look less like autocomplete and more like an actual junior developer workflow.

The big win is not just clicking around the app. It is keeping the test tied to real operating context: schema, scripts, sample records, permissions, and what the browser observed. For FileMaker/WebViewer work, that catches the stuff a unit test never sees.

link?

Thank you for thinking of Windows users – that’s really kind of you!

how does this differ from the built in /browser?

in app browser in codex works just fine

only for macOS users??

@mikepat711 bro codex can use chrome now

Super useful update from the Codex team! The Chrome plugin sounds perfect for real browser testing without the usual headaches. Parallel DevTools + tab context is exactly what agent workflows have been missing. Huge for anyone shipping web apps. Can’t wait to try it. 🔥

so its like playwright-cli or similar but via an extension? does this save context ?

Codex Chrome plugin just turned your browser into a full dev sidekick — testing apps, pulling tab context, parallel DevTools, all without hijacking your session. Browser workflows just got 10x smoother for builders. Who’s shipping their first web app with it? 🚀

This is awesome! One thing I love about Claude is the chrome extension and this essentially can do exactly that now but better and directly in Codex.

the browser is becoming the new IDE faster than anyone expected

Bringing testing and context gathering into the browser workflow like this makes development more continuous and less fragmented across tools

Codex with DevTools access changes browser-based agent testing fundamentally. Parallel tab inspection + console scraping = real frontend QA loops, not just "click this button." Curious how it handles auth state isolation across parallel browser sessions — that's where most browser agents leak.