Google Antigravity remote agents let your coding agent keep working in the cloud after you shut your laptop. The feature arrived with Antigravity 2.29, which started rolling out the week of November 6, 2026, for paid Google AI Pro and Ultra subscribers.
Here’s the short version before the details:
- Version: Antigravity 2.29, rolling out in stages from the week of November 6, 2026
- Remote agents: for Google AI Pro and Ultra subscribers
- Also new: a Voice Mode powered by Gemini Live and a redesigned app
- Platforms: the desktop app on Mac, Windows and Linux
Those four points cover most of the news. The rest of this post explains what each one means for how you actually work.
How Google Antigravity remote agents actually work
You create a remote project from a GitHub repo or a Google Drive folder. Antigravity then runs your agent’s tasks inside an isolated virtual machine in Google’s cloud, not on your own computer.
That’s the key difference from a local agent. If your laptop sleeps, loses Wi-Fi or gets packed into a bag, the task keeps going. According to 9to5Google, Google describes the payoff as coming back to passing tests and a clean pull request.
You can also check in from your phone at antigravity.google.com/remote. If an agent stalls waiting for a permission, you can unblock it from there instead of waiting until you’re back at your desk.
Talk to your agent with the new Voice Mode
Next to the existing microphone button, there’s now a separate Voice Mode built on Gemini Live. You speak a request and the agent picks it up, much like talking to the Gemini app.
You get several voice options, and you can set playback speed anywhere from 0.75x to 2x. If you spend long stretches reading code, this is handy for firing off small instructions without breaking your flow.
A redesign that borrows from the Gemini app
The prompt box moves to the bottom of the conversation page, with a greeting at the top. The app also switches to Google Sans and the Onyx theme.
Small workflow tweaks came along too. For single-repo projects, branch and Git worktree selection now share one menu, and the agent picker moved into the model menu. A new Automations page sorts your jobs into Scheduled, Always on and Paused lists, with Next run and Last run times.
Everything else in version 2.29
The built-in browser got several fixes people have asked for. Downloads now save straight to your downloads folder, and you can pick a default folder in Settings > General. The comparison below puts the main changes side by side.
| Change | What it does | Where to find it |
|---|---|---|
| Remote agents | Runs tasks in cloud VMs, even with your laptop closed | New remote project; antigravity.google.com/remote |
| Voice Mode | Speak requests using Gemini Live | Next to the microphone button |
| Automations page | Groups jobs as Scheduled, Always on or Paused | Automations in the app |
| Custom shortcuts | Lets you remap keys and reset to defaults | Settings > Shortcuts |
The browser’s address bar also gets history suggestions and clearer security indicators. Google says the app launches faster from a cold start too.
Who should use a remote agent instead of a local one?
Remote agents suit work that takes a while and doesn’t need your eyes: running a long test suite, chasing down a flaky build, or drafting a first pass at a refactor. You hand it off, go to lunch and review the result later.
A local agent still makes more sense for quick edits where you want instant feedback. It also fits cases where your code can’t leave your machine. Remote projects live in Google’s cloud and start from GitHub or Google Drive, so check your team’s data rules first.
The new Voice Mode and phone control make the two styles blend together. You can start a task by voice at your desk, then keep an eye on it from your phone while you’re out.
What we don’t know yet
The rollout is staged, so you may not see these features today even if you pay for AI Pro or Ultra. 9to5Google’s report also doesn’t mention usage limits for remote agents or which regions get them first.
That matters because cloud VMs cost Google real money. Check your plan’s terms before you hand an agent a huge, long-running job. If you’re weighing other coding agents, our posts on Cursor’s cloud agent that builds and hosts live apps and Android Studio Rabbit 2’s bring-your-own-agent option show how fast this space is moving.
Frequently Asked Questions
Who can use Antigravity remote agents?
Paid Google AI Pro and Ultra subscribers are first in line. Google hasn’t said whether free users will get access later.
Do remote agents keep running if I close my laptop?
Yes. Tasks run in isolated cloud virtual machines, so they continue while your computer is off or asleep.
What can I connect to a remote project?
You can start a remote project from a GitHub repository or a Google Drive folder.
Can I control agents from my phone?
Yes, through antigravity.google.com/remote. You can approve permission requests from there so a stuck task doesn’t wait for you.
Which computers does the Antigravity desktop app support?
Mac, Windows and Linux.
My take: remote agents are the part worth trying, because they fix the most annoying thing about local AI coding, which is babysitting a laptop. Start with a small, low-risk repo, review every pull request the way you’d review a new teammate’s work, and wait for usage limits to become clear before you rely on it for deadlines.
Leave a Reply