Google released version 2.29 of Antigravity, its agentic development platform, on November 6, and the update gives the coding agent a voice — and a life beyond the developer’s laptop. Voice mode, powered by Gemini Live, lets developers talk to the agent while it works, and Remote Agents run tasks in isolated cloud machines that keep going after the laptop closes. The release also redesigns the conversation page, with the cloud agents rolling out to paid Google AI Pro and Ultra subscribers.
Talk to the agent instead of typing at it
Voice mode turns a conversation in the Antigravity App into a spoken, two-way exchange. Click Start Voice mode in the message box, pick one of nine voices, and set the speaking speed from 0.75x to 2x. The agent already has the context of everything typed before, so a spoken request lands in the same conversation.
While the developer talks, a panel above the message box shows what was said, then lights up the agent’s reply word by word as it speaks. The mode runs on the Gemini Live model, which hears the voice directly instead of a transcript, so a misheard word does not mean the agent misunderstood.
The developer can interrupt at any time just by talking, steer the work mid-task, or say “stop” to end the running task. Approvals can be spoken too, but destructive actions such as deleting files or force-pushing still need an on-screen click.
By default the agent works quietly and speaks only when it has an answer or needs input. Saying “keep me posted” brings a short update about every 30 seconds, or the developer can ask for a different interval. Voice mode works the same way in the Antigravity CLI, started with /voice live or the F6 key.
Voice mode is available in the Antigravity App and in CLI version 1.5.0 and later. A status row below the prompt shows how long voice mode has been on and whether the agent is listening, thinking, or speaking. When the session ends, everything the agent wrote, edited, or drew stays in the conversation, along with a transcript.
Remote Agents keep working when the laptop closes
The bigger change is Remote Agents, now rolling out to paid Google AI Pro and Ultra subscribers. A remote project is created in the cloud from a GitHub repository or a Google Drive folder, and the task runs in an independent cloud virtual machine.
The isolated environment keeps working even when the laptop closes, and the developer can unblock permissions from a phone at antigravity.google.com/remote. Remote Agents fit the same idea as voice mode: the agent keeps working while the developer steps away for coffee.
Fire off a task, step away, and come back to passed tests and a clean pull request, as Google puts it. The cloud machines run independently, so a long task no longer depends on the developer’s machine staying awake and online.
A redesigned conversation page
Version 2.29 also redesigns the conversation page, moving the prompt box to the bottom of the screen and adding a greeting at the top, in the style of the Gemini app. The app now uses a default Google Sans font and an Onyx theme.
Single-repo projects combine branch and Git worktree selection into one menu, and the agent picker now lives inside the model menu. The Automations page now organises automations into Scheduled, Always on and Paused lists with next-run and last-run times shown plainly.
The built-in browser saves downloads directly, suggests pages from browsing history in the address bar, and gains a redesigned address bar with clearer security indicators. Keyboard shortcuts are now customisable, and Antigravity launches faster on a cold start, and the sidebar animates smoothly.
Our Read
Voice is the natural interface for steering long agent tasks, because the moments that need a developer — a forgotten detail, a permission prompt, a progress check — are exactly the moments that pull hands back to the keyboard again.
Talking while watching the code keeps the developer’s eyes where the work is, and Gemini Live hearing the voice directly rather than a transcript removes a layer where meaning gets lost. That is the kind of detail that matters when the instruction is spoken, not typed.
Remote Agents are the larger shift, though. An agent that keeps working after the laptop closes turns the editor from a tool into a colleague that works the night shift — and Google putting it behind Pro and Ultra paywalls first is the commercial shape of that idea.
The risk is the same as with any always-on agent: work done out of sight needs the permission checks Google built in to stay visible. The redesign matters less than the two features, but it points the same way: Antigravity is being reshaped around conversations that run long, across machines and mornings.
Why It Matters
If developers can talk to their coding agent and leave it running in the cloud, the shape of a programming day changes: the editor becomes a place to supervise work rather than type it. For rival agent editors, the bar moves again — voice and cloud execution are now part of the contest, not extras.
For developers, the win is quieter: fewer trips back to the keyboard, and long tasks that finish while they are elsewhere. That is the promise of both features in this release — less typing, less waiting, and an agent that keeps working when the developer walks away.
Background
Antigravity is Google’s agentic development platform for Mac, Windows and Linux, built around agents that read code, edit files and run commands. Google shut down Jules, its earlier async coding agent, and moved its plugins to Antigravity, and has since used the platform to test ideas such as drug-discovery skills.
Source: Google’s Antigravity voice mode announcement and 9to5Google’s report
0 Comments