Antigravity Adds ‘Remote Control’ Feature For AI Coding Agents

Google has introduced Remote Control for Antigravity, a new capability that allows developers to connect to and control their Antigravity 2.0 coding-agent sessions from a web browser. The feature is designed for long-running development tasks that may continue after a developer leaves their desk, allowing users to monitor progress, review changes and provide additional instructions remotely.

The update reflects a broader shift in AI-assisted software development from simple coding assistance toward autonomous agents capable of handling larger tasks. Google says Antigravity agents can work on activities such as full-subsystem refactoring, extensive test runs and complex dependency migrations, which can take significant amounts of time. Remote Control is intended to let developers keep those workloads running on their existing machines rather than recreating the development environment elsewhere.

Google Antigravity Remote Control At A Glance

Particular	Details
Company	Google
Product	Antigravity
Feature	Remote Control
Current version	Antigravity 2.0
Access method	Web browser
Supported environments	Laptops, desktops and servers
Main purpose	Remotely control Antigravity sessions
Local files	Retained
Workspaces	Retained
Build tools	Retained
Credentials	Retained
Environment variables	Retained
Multi-machine control	Yes
Push notifications	Available through web app on mobile devices
Rollout	Gradual
Priority access	Google AI Ultra plans

Google’s documentation says Remote Control is currently rolling out, with Google AI Ultra subscribers receiving priority access.

What Is Google Antigravity Remote Control?

Remote Control essentially turns a web browser into a window for controlling an Antigravity development environment running somewhere else.

Developers can have Antigravity running on a desktop, laptop or server and then connect to that machine remotely.

Developer's Main Machine
        ↓
Antigravity 2.0
        ↓
AI Agent Working On Task
        ↓
Remote Control
        ↓
Web Browser
        ↓
Developer Can Monitor
Review And Give Instructions

The important part is that the development environment remains on the original machine.

Google says users retain access to their files, workspaces, build tools, credentials and environment variables through the remote session.

Developers Can Control Multiple Machines

One of the main features Google highlights is multi-instance management.

Developers may use different computers for different purposes. For example, a local workstation could be dedicated to interface development while a Linux server handles backend or server-side workloads.

Remote Control lets users switch between these Antigravity instances from the browser.

Multi-Machine Workflow

Machine	Possible Workload
Main workstation	UI development
Laptop	Mobile or lightweight development
Linux server	Backend workloads
Cloud environment	Long-running tasks
Dedicated machine	Testing and builds

Google’s Remote Control interface provides an instance switcher that allows users to select the machine they want to control.

                Remote Browser
                     │
          ┌──────────┼──────────┐
          ↓          ↓          ↓
      Laptop      Desktop     Server
         │           │           │
         ↓           ↓           ↓
     Antigravity  Antigravity  Antigravity
         │           │           │
         └───────────┼───────────┘
                     ↓
              Multiple Agents

This could be particularly useful for developers who maintain several environments.

AI Agents Can Keep Working While Developers Are Away

The feature is designed around a simple problem: AI coding tasks can take time.

A developer might ask an agent to refactor a large subsystem, run an extensive test suite or perform a complex dependency migration.

Instead of waiting at the computer, the developer can leave the workstation and check the session later.

Start Task
    ↓
AI Agent Begins Work
    ↓
Developer Leaves Desk
    ↓
Agent Continues Working
    ↓
Notification
    ↓
Developer Opens Browser
    ↓
Reviews Progress
    ↓
Provides More Instructions

Google describes this as “untethered productivity”, allowing developers to move away from their physical workstation without stopping their development work.

Remote Control Preserves The Local Development Environment

A major advantage of Google’s approach is that developers do not need to recreate their entire development environment on another device.

For example, if a project depends on specific libraries, environment variables, build tools or local files, the agent can continue operating in the original environment.

Traditional Remote Workflow Vs Antigravity

Traditional Approach	Antigravity Remote Control
Move to another machine	Open a browser
Recreate environment	Keep existing environment
Transfer files	Files remain local
Reinstall dependencies	Existing dependencies remain
Reconfigure tools	Existing configuration remains
Resume work manually	Continue existing session

Google says this allows developers to continue sessions without attempting to recreate the environment on another workstation.

Why Local Context Matters For AI Coding Agents

AI coding agents often need access to much more than source files.

A working software environment can include:

Source code
Dependencies
Build systems
Environment variables
Credentials
Local databases
Development tools
Test environments
Configuration files

Recreating all of that on another computer can take considerable effort.

Remote Control instead keeps the agent inside the environment where the task was originally started.

Existing Environment
       ↓
Files
Workspaces
Tools
Credentials
Variables
Dependencies
       ↓
Antigravity Agent
       ↓
Remote Browser Control

This could make remote agent workflows substantially more convenient for developers working across multiple machines.

Mobile Devices Can Receive Notifications

Google says developers can use the Remote Control web application on mobile devices.

Users can optionally install the web app on their phone’s home screen to receive push notifications when agents complete tasks or require input.

That creates a workflow in which a developer does not need to continuously monitor an AI agent.

Example Mobile Workflow

Step	Action
1	Start coding task on workstation
2	Agent begins execution
3	Developer leaves workstation
4	Agent continues working
5	Phone receives notification
6	Developer opens browser
7	Reviews agent status
8	Provides additional instruction if necessary

This is particularly relevant for tasks that may take extended periods to complete.

Remote Control Is Built For Longer AI Tasks

The feature makes more sense as AI coding agents become increasingly autonomous.

Earlier coding assistants were generally designed around short interactions:

Ask → Generate Code → Review

Agentic development can involve much longer workflows:

Plan → Modify Files → Run Tests → Inspect Errors → Fix → Test Again → Validate

Simple AI Coding

Prompt
 ↓
Code
 ↓
Review


Agentic Development

Task
 ↓
Planning
 ↓
Code Changes
 ↓
Build
 ↓
Testing
 ↓
Error Analysis
 ↓
More Changes
 ↓
Testing
 ↓
Validation

Google’s own documentation specifically cites large-scale refactoring, extensive test-suite runs and dependency migrations as examples of tasks that can run for extended periods.

Remote Control Could Change How Developers Use AI Agents

The feature changes the relationship between the developer and the coding agent.

Instead of sitting beside the AI throughout the entire process, the developer can increasingly act as a supervisor.

Traditional Coding Assistant

Developer
    ↓
Prompt
    ↓
AI
    ↓
Developer Reviews
    ↓
Next Prompt


Agentic Workflow

Developer
    ↓
Defines Goal
    ↓
AI Agent
    ↓
Plans + Executes
    ↓
Developer Monitors
    ↓
Intervenes When Needed

This is closer to an AI project worker model than a conventional autocomplete tool.

Google Is Expanding Antigravity Beyond A Traditional IDE

Remote Control comes as Google continues positioning Antigravity as an agent-first development platform.

At Google I/O 2026, the company described Antigravity 2.0 as a platform for orchestrating agents and introduced capabilities including the Antigravity CLI and SDK. Google said developers could use specialized subagents for complex workflows and programmatically control the agent harness.

The Remote Control feature extends that philosophy beyond the physical development machine.

Antigravity’s Broader Direction

Capability	Purpose
Agent-first IDE	AI-driven software development
Multiple agents	Parallel workflows
CLI	Terminal-based agent control
SDK	Programmatic agent control
Remote Control	Browser-based session control
Multi-machine support	Manage different environments

The result is an ecosystem where AI agents can operate across increasingly varied development workflows.

Google Is Also Bringing Antigravity To Enterprise Users

The Remote Control launch arrives alongside a broader expansion of Antigravity into enterprise environments.

Google announced that Antigravity is available as part of eligible Gemini Enterprise subscriptions, with administrative and spending controls. The company also introduced IDE extensions allowing developers to use Antigravity in IDEs including VS Code.

This matters because remote control of development environments can become more valuable when organizations operate multiple machines, servers and development environments.

Enterprise Antigravity Expansion

Area	Google Offering
Enterprise access	Eligible Gemini Enterprise subscriptions
Administration	Centralized controls
Spending	Budget and threshold controls
Usage monitoring	Usage metrics
IDE support	Extensions including VS Code
Agent development	Antigravity platform

Google says the enterprise controls are designed to consolidate governance, spending, security and usage management.

Security Becomes More Important With Remote Agent Control

Remote access to development environments also introduces an important security consideration.

Google describes Remote Control as a secure window into the workspace, while its documentation says the feature provides access to files, workspaces, build tools, credentials and environment variables.

That means organizations need to treat remote agent access as more than simply viewing a coding session.

Sensitive Resources Potentially Involved

Resource	Why It Matters
Source code	Intellectual property
Credentials	Access to services
Environment variables	May contain secrets
Build tools	Can execute development tasks
Workspaces	Project data
Server environments	Potential infrastructure access

The convenience of remote control therefore needs to be balanced with appropriate access controls and security practices.

Remote Control Could Make Dedicated AI Workstations More Useful

One interesting consequence is the possibility of using dedicated machines primarily for AI agents.

A developer could leave a powerful workstation or server running continuously and access it remotely when necessary.

Powerful Workstation
       ↓
Antigravity Agent
       ↓
Long-Running Tasks
       ↓
Developer Away
       ↓
Browser / Phone
       ↓
Monitor + Intervene

This could allow developers to separate their everyday computing device from the machine responsible for heavy or long-running development tasks.

Cloud And Local Environments Can Work Together

Google specifically highlights the ability to use different machines for different types of work, including cloud environments for server-based tasks.

This could lead to hybrid workflows.

                 Developer
                     ↓
              Remote Browser
                     ↓
          ┌──────────┼──────────┐
          ↓          ↓          ↓
       Laptop     Desktop      Cloud
          ↓          ↓          ↓
        UI Work    Testing    Backend
          └──────────┼──────────┘
                     ↓
               AI Agents

Instead of forcing all development onto one computer, developers can distribute workloads while maintaining centralized browser access.

Antigravity Remote Control Is Rolling Out Gradually

Google’s documentation currently describes Remote Control as a feature being rolled out, with Google AI Ultra plans receiving priority access.

That means availability may vary depending on account type and rollout stage.

Google’s documentation provides the following basic setup process:

Open Antigravity 2.0 settings.
Go to the Account section.
Enable Remote Control.
Optionally assign a nickname to the machine.
Open the Remote Control dashboard from a browser.
Sign in using the same Google Account.
Select the machine from the instance switcher.

What Developers Could Use It For

The feature is particularly suited to workloads where the agent needs time to operate independently.

Potential Use Cases

Use Case	Why Remote Control Helps
Large refactoring	Agent can work for longer periods
Test-suite execution	Developer can leave machine
Dependency migration	Long-running workflow
Build debugging	Monitor remotely
Backend development	Dedicated server environment
Cloud development	Access remote environment
Multiple projects	Switch between machines
Overnight tasks	Check progress later

Google specifically identifies subsystem refactoring, extensive tests and dependency migrations as examples of longer-running workloads.

Remote Control Fits The Shift Toward Agentic Development

The bigger significance of the feature is not simply remote access.

It reflects the changing role of AI in software development.

AI coding systems are moving from tools that generate individual pieces of code toward agents capable of handling complete workflows.

AI Coding Evolution

Autocomplete
    ↓
Chat-Based Coding
    ↓
Multi-File Editing
    ↓
Tool-Using Agents
    ↓
Autonomous Workflows
    ↓
Long-Running Agents
    ↓
Remote Agent Management

Remote Control is a natural extension of this progression because longer-running agents require new ways for humans to supervise them.

The Bigger Picture

Google’s Antigravity Remote Control represents another step toward software development in which AI agents can work independently for longer periods while developers act as supervisors rather than constantly interacting with the coding assistant. The feature allows users to connect to Antigravity 2.0 sessions running across laptops, desktops and servers from a web browser, while retaining access to the original files, workspaces, build tools, credentials and environment variables.

The multi-machine capability could be particularly useful for developers who maintain separate environments for frontend work, backend development, testing or cloud workloads. Google also supports mobile access through its web app, including push notifications when an agent completes a task or needs user input.

The launch comes as Google expands Antigravity beyond a traditional AI coding assistant. Antigravity 2.0 is positioned as an agent-first development platform, while Google has also introduced an Antigravity CLI and SDK for more advanced agent orchestration and programmatic control.

The enterprise expansion is another important part of the picture. Google has made Antigravity available through eligible Gemini Enterprise subscriptions and added administrative, security, spending and usage controls, while also introducing IDE extensions including support for VS Code.

Looking Ahead

Remote Control could become increasingly important as AI coding agents take on larger and more complicated software-development tasks. When an agent can spend substantial amounts of time modifying code, running tests, diagnosing failures and completing migrations, requiring a developer to remain physically next to the machine becomes a significant limitation. Browser-based control removes that constraint while keeping the agent inside its existing development environment.

The larger trend is toward a development model where humans define goals, monitor progress and intervene when necessary while AI agents handle more of the execution. Google Antigravity’s Remote Control feature is designed for exactly that transition: start the work on one machine, let the agent continue working, and take control from wherever you are when needed. As agentic development matures, remote supervision could become as important as the coding agent itself

Get the day’s top stories in your inbox

One concise email. No spam, unsubscribe anytime.