Better feedback = Better code. Artifacts are the agent’s proof of work. Review plans, screen recordings, and code changes. Then, leave feedback to keep the agent on track. Try it today:

This is really great, but can you increase limits plz, its all a little pointless if we just keep hitting the limits after just a few prompts. I find myself going back to cursor everytime I hit a limit but would ideally want to try agy as my daily driver

Hey @antigravity , Gemini 3 Pro (Thinking) is not working in Antigravity, it's showing 'Error' I refreshed, restarted, still same issue, and I highly doubt it's my quota, please look into it team

Will you guys at Opus 4.5? 🙏

@mooohul show the remote control cars in the next tutorial

Does it hook into @Firebase to provision storage within the IDE? Cause it 100% should.

Hi, I think the developer community for this software is small, and even so underdeveloped that developers can no longer access beta versions of this software! Or a place to give feedback and submit bugs! So what is QC doing?

It's great when starting a project from scratch, but there's a lot of feedback that it breaks things when editing an existing project. If you do a little research, you can see this. So you need to fix this flawed way of working.

The best IDE Where have you been all this while

You Google Just merge my Gemini Pro credits with Antigravity. Nobody is touching your editor with a 2-prompt limit.

Let me just disable this 2 things and I can move out of cursor: - this panel that always open. - this message on empty lines.

This is what I love most about @antigravity — it doesn’t just generate code, it truly *understands* my project like a seasoned pair programmer. One prompt → a complete, production-ready Next.js contact page with perfect metadata, Tailwind styling, and copy that matches my voice exactly. No generic templates. No manual className wrestling. No hours lost on spacing and typography. Just clean, human-quality code that feels like I wrote it myself on my best day — because the agent studied my existing pages and respected the established patterns. This isn’t automation. This is augmentation. Grateful to be shipping faster and thinking deeper instead of debugging pixels. Thank you @antigravity for giving developers their most precious resource back: time. #coding #antigravity

The 'plan' review step is so key. Is the feedback structured, like approving/rejecting individual steps, or more free-form natural language? Finding the right level of abstraction for feedback has been the hardest part for me.

Please add the new OPUS model

Imagine this website done by @antigravity

@antigravity I reallylike the tool, but when it is debugging with the browser it seems like it is always confident it is correct (one time it broke the rendering, resulting in a fully black screen, but it didn't notice it). Also MMB+dragging isn't caught while in DOM. Any advice?

add claude 4.5 opus

Add option to use openrouter models

I haven’t fully switched to Antigravity because of the limit, please increase it.

Proof of work isn’t just for blockchain, serious agents show up with their receipts. People always underestimate how feedback loops shape the outcome until they actually need a win.

And your programmers are still clueless about the Windows version being broken! Why didn't you see the update? The Windows version is completely broken!

Kindly be adding more 3rd party models like Cursor and Windsurf

Really impressive direction. Clear artifacts + tight feedback loops are exactly what make autonomous agents actually usable in real workflows.

L'expérience est intéressante, dommage qu'il soit si bogué (fuite mémoire, blocage fréquent)

Agents only improve when you close the feedback loop. Artifacts make the work visible. Review drives precision.

Unusable rn

Love this framing — artifacts turn agents from a black box into something you can inspect, critique and improve. When humans review plans, recordings and diffs, the agent isn’t just “fixing code”, it’s learning through a real feedback loop.

Hi guys, thank you for the valuable work you are doing. Antigravity is an amazing tool! Is there any chance that for Gemini PRO users, the usage limit for the Gemini 3 Pro model (within Antigravity) will be raised?

but .. dont the agents have agency??? if da agents have agency.. no need for Karen mode.. no need to check artifacts.. check code blyat...

In the last 24 hour it barely works, always hangs on prompts

Absolutely! Artifacts are game-changers—turning AI agents into code-crushing machines. Reviewed, feedbacked, iterated. Who's ready to level up their dev game? 🚀 Try it now! #AICoding #BetterCode#AICoding #BetterCode

Add a simple copy feature for chat, thinkin, and prompt and also for the plan which agent creates. And make the given prompt shrunk in chat it takes longer scroll.

When can I pay for higher limits?

Question! I have had the Google AI Ultra since its inception, and I keep hitting rate limits on AntiGravity very quickly. What is the $250/month for? The way I see it, Google saves tokens if I don't keep entering my entire codebase into Gemini chat repeatedly, so why wouldn't you guys link it with our Ultra subscription?

The limits make this all a pointless exercise. 2 prompts in antigravity and straight back to the normal ide

I've been using it all day today, this one looks really good!