---
type: "post"
id: "google-antigravity-rolls-out-remote-agents-on-cloud-e143c6"
slug: "google-antigravity-rolls-out-remote-agents-on-cloud-e143c6"
title: "Google Antigravity rolls out Remote Agents on cloud VMs"
description: "Google is gradually rolling out Antigravity Remote Agents, which run agent sessions and tools in an isolated, Google-managed remote environment."
retrieval_nugget: "Google is gradually rolling out Antigravity Remote Agents, which run agent sessions and tools in an isolated, Google-managed remote environment. For builders, the practical change is that eligible tasks can continue after a laptop sleeps or a browser closes, without keeping a local machine online. Access is currently limited to personal Google AI Pro and Ultra accounts."
published_at: "2026-11-06"
updated_at: "2026-11-07"
record_date: "2026-11-06"
date_kind: "published_at"
topics: ["google"]
entities: ["Google","Antigravity","Google AI Pro","Google AI Ultra","Google Workspace","Google Cloud"]
related_patterns: ["harness-architecture-outlives-model-choice"]
novelty_label: "same_event"
prior_refs: []
source_url: "https://antigravity.google/docs/remote-agents"
source_urls: ["https://antigravity.google/docs/remote-agents"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:google-antigravity-rolls-out-remote-agents-on-cloud-e143c6"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/google-antigravity-rolls-out-remote-agents-on-cloud-e143c6/","markdown":"https://newruntime.com/posts/google-antigravity-rolls-out-remote-agents-on-cloud-e143c6.md","json":"https://newruntime.com/posts/google-antigravity-rolls-out-remote-agents-on-cloud-e143c6.json"}
markdown_url: "https://newruntime.com/posts/google-antigravity-rolls-out-remote-agents-on-cloud-e143c6.md"
---

# Google Antigravity rolls out Remote Agents on cloud VMs

## Retrieval answer

Google is gradually rolling out Antigravity Remote Agents, which run agent sessions and tools in an isolated, Google-managed remote environment. For builders, the practical change is that eligible tasks can continue after a laptop sleeps or a browser closes, without keeping a local machine online. Access is currently limited to personal Google AI Pro and Ultra accounts.

## Summary
Google is gradually rolling out Antigravity Remote Agents, which run agent sessions and tools in an isolated, Google-managed remote environment. For builders, the practical change is that eligible tasks can continue after a laptop sleeps or a browser closes, without keeping a local machine online. Access is currently limited to personal Google AI Pro and Ultra accounts.

## What happened
Google’s Antigravity documentation describes Remote Agents as a way to run agent sessions and tools in an isolated, Google-managed remote environment rather than on a local machine. The stated benefit is continuity: long-running tasks and command execution can continue when a user closes a browser, puts a laptop to sleep, or switches between desktop and mobile devices. The documentation presents this as a distinct execution mode, not simply a way to view or control a local session remotely.

Remote Agents is being rolled out gradually to subscribers on Google AI Pro and Google AI Ultra who sign in with a personal Google Account. The documentation says it is not available on the free plan, Google AI Plus, or business accounts using Google Workspace or Google Cloud. It does not provide a price for Remote Agents as a separate product; the listed access condition is an eligible subscription. Users can launch and manage sessions from the Antigravity desktop app, the Antigravity CLI, or a browser at antigravity.google.com. In the browser, a user can start a standalone remote conversation or create a project and add resources.

A remote project can include GitHub repositories, Google Drive files, and files uploaded from a computer. The app can create a project around an existing GitHub repository, and the CLI can run conversations remotely. The supplied documentation excerpt also says users can unblock permission requests from a phone. These options make the remote environment useful for repository tasks and background work, while allowing access from more than one device.

Google distinguishes Remote Agents from Remote Control. Remote Agents execute in Google’s isolated remote environment and do not require a local machine to stay powered on. Remote Control connects to a user’s own desktop workstation or server, which must remain powered on and online. Remote Control is intended for work needing local hardware, local development servers, local toolchains, or uncommitted local checkouts. The distinction matters: a cloud session needs remote resources, while local work remains tied to the host machine.

The supplied material identifies the relevant Antigravity CLI command as `agy --remote-agent` and says this requires CLI version 1.4.12 or later. It also notes that the same v2.29 wave refreshed the interface, voice mode, and Jupyter notebook viewer. Those items are mentioned alongside the rollout, but the documentation excerpt does not specify their detailed behavior or availability separately. The central change documented here is remote execution for eligible agent sessions, with a gradual rollout and account restrictions.

## What the archive adds
A prior Field Note, [Cloud execution is coming to Antigravity soon Google will retire Jules on Dec.](https://newruntime.com/posts/cloud-execution-is-coming-to-antigravity-soon-google-814edc/), described a local/cloud execution toggle and a dev-only Running Jobs section as signs of remote background workflows moving into Antigravity. This rollout is a new fact on that developing story: the docs now specify access, execution, and the distinction from Remote Control. Earlier coverage of [Antigravity custom agents](https://newruntime.com/posts/google-antigravity-custom-agents/) is related context, not the same event.

## Trend signal
This weakly supports [harness architecture outlives model choice](https://newruntime.com/patterns/harness-architecture-outlives-model-choice/): the user-facing product boundary includes where an agent runs, which resources it can reach, and whether a local host must stay online. The documentation does not establish that the execution environment is model-independent or durable, so this is a limited operational signal rather than proof of the broader pattern.

## Why now
The archive’s earlier Antigravity note had already identified a local/cloud execution toggle and running-jobs interface as signs of a coming remote workflow. The current documentation turns that signal into a described capability, while leaving rollout gradual and eligibility restricted. The source does not state a broader launch rationale or deadline.

## Connections to other events
The earlier Antigravity coverage described custom agents as a way to package specialized work in the IDE and CLI. Remote Agents adds a different dimension: where those sessions execute and whether they depend on an active local host. It also resembles cloud-agent workflows described in the archive’s note on Cursor’s cloud agents, but the supplied context does not establish feature parity between the products.

## Why it matters
For a team that already uses Antigravity, the key operational choice is whether a task depends on the local development environment. Repository work that can use the remote resources listed in the docs may be able to continue without tying up a workstation. That can make overnight or mobile workflows more practical, and may reduce interruptions caused by a sleeping laptop or closed browser.

But moving execution off the laptop is not a universal upgrade. A task that needs a local server, hardware, a local toolchain, or uncommitted files still points toward Remote Control, where the host must remain online. Teams also need to consider which repository and Drive resources they authorize for a managed remote environment. The documentation describes isolated execution, but does not provide enough detail here to evaluate its security controls or compare its operational guarantees with a team's existing setup. Treating the two modes as different deployment contexts, rather than interchangeable buttons, is the useful starting point.

## What happens next
- **Remote Agents will become available to more eligible personal Pro and Ultra accounts** (3 months, 75%). Watch for: Google’s documentation changing the gradual-rollout language or listing additional availability.

## Limits and uncertainty
The source establishes the rollout, account eligibility, supported entry points, resource types, and broad difference between Remote Agents and Remote Control. It does not state a rollout schedule, regional availability, separate usage limits, compute specifications, task-duration limits, or pricing beyond requiring an eligible subscription. It also does not detail the permission model, data retention, isolation guarantees, or how remote environments handle secrets. The claim that work continues while a laptop sleeps describes the intended mode, not a quantified reliability commitment. The excerpt mentions a v2.29 wave and CLI 1.4.12+, but does not give a release date.

## Skill to build now
Classify a task before sending it to a remote agent: does it need only a repository or other listed remote resource, or does it depend on a local server, uncommitted checkout, hardware, or toolchain? Try one small repository task through Remote Agents and record any missing local dependency. That gives a concrete basis for choosing cloud execution or Remote Control for the next longer-running job.

## Reader takeaway
Antigravity now documents cloud-hosted agent execution for eligible personal Pro and Ultra subscribers. Use it when a task can work from remote resources and should outlast your active session; use Remote Control when the work still depends on your local machine. The availability limits and missing operational details are reasons to test a small task first.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Google](https://newruntime.com/topics/google.md): Explore the google topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)