Skip to content

Choose your Antigravity IDE #29

Closed
Closed

Description

@zskbot

Choose your Antigravity

Antigravity delivers native agentic assistance across purpose-built product surfaces, with dynamic reasoning controls, multi-agent orchestration, and enterprise-grade security. Evolved from the Agent Manager into a standalone desktop command center (Antigravity App) alongside native terminal, editor, and programmatic SDK interfaces, Antigravity supports both synchronous interactive pair-programming and asynchronous background agent workflows independent of a single IDE. Select any surface below to explore live interface previews, core workflows, and quickstart paths:

Surfaces

v2.5+

Antigravity AppDesktop application

agy · v1.2+

Antigravity CLITerminal client

6 IDEs

For IDEsEditor extensions

Python

Antigravity SDKProgrammatic runtime

New Conversation

Conversation History

Scheduled Tasks

Projects

checkout-service

Discount Code Validationnow

Stripe Webhook Handler2h

inventory-api

Low Stock Alerts1d

web-storefront

Order Summary Drawer

Settings

checkout-service/Discount Code Validation

Add promo code validation and expired coupon checks to the checkout endpoint.

Worked for 11s

I updated checkout-service to validate discount codes before creating a payment intent:

  • PromoValidator: Check expiration dates and minimum order totals.
  • CheckoutController: Return 400 Bad Request with a clear message for expired codes.

Shall I run the checkout test suite to verify?

Ask anything, @ to mention, / for actions

Gemini 3.8 FlashMedium

Model

Gemini 4 Argon _High_Gemini 3.8 Flash _Medium_Claude Sonnet 5.5 _Thinking_Claude Opus 5.5 _Thinking_GPT-OSS 120B Medium

TerminalsProject

dev@workstation checkout-service % npm test

PASS src/discounts.test.ts (14 tests)

✓ rejects expired promo codes (9ms)

Standalone Terminals

Conversations

Discount Code Validation

zsh

Artifacts2 files

Implementation Plan

Promo code validation & unit tests

Verification Report

Checkout test suite results

Proposed Changes

+ if (promo.isExpired()) return res.status(400);

+ const discount = applyPromo(cart.subtotal, promo);

Browserlocalhost:3000

http://localhost:3000/checkout/preview

Checkout API · Order Summary

FieldValue

Status200 OK

Promo CodeSPRING20 (-$24.00)

Order Total$96.00

A standalone desktop application for launching, orchestrating, and inspecting autonomous agent teams with visual diff and plan inspection.

Get started with Antigravity AppDesktop features guide

Antigravity CLI

Antigravity CLI 1.2.14

developer@example.com (Google AI Pro)

Gemini 3.8 Flash (High)

~/projects/checkout-service

>Run the test suite and fix the failing discount rounding test

▸Thought for 2s, 114 tokens

Running unit tests to reproduce the rounding failure

Bash(npm test -- src/discounts.test.ts)(ctrl+o to expand)

▸Thought for 3s, 142 tokens

Updating currency rounding in calculateDiscount

Edit(src/pricing/discounts.ts)(ctrl+o to expand)

>

? for shortcutsGemini 3.8 Flash · high

A keyboard-first terminal client that runs Antigravity workflows in local shells, remote cloud environments, and SSH sessions.

Get started with Antigravity CLICLI reference manual

spring-petclinic

ANTIGRAVITYVS Code

spring-petclinic / Project Analysis And Reporting

Please review this project. Investigate how it works and build a report showing important files and data flows.

  • BaseEntity, NamedEntity, and Person serve as base classes.
  • Owner, Pet, Visit, Vet, Specialty, and PetType define business models.

Key System Controllers:

  • OwnerController: Manages owner creation, lookup, and pagination.
  • PetController: Handles pet updates/creation using PetValidator.
  • VetController: Provides HTML vet list and cached JSON endpoint.

Project ReviewComprehensive review report for Spring PetClinic application detailing architecture...

Ask anything, @ to mention, / for actions

Gemini 3.8 Flash Medium

Model

Gemini 4 Argon HighGemini 3.8 Flash MediumClaude Sonnet 5.5 (Thinking)Claude Opus 5.5 (Thinking)GPT-OSS 120B (Medium)

Preview README.mdproject_review.md

Project ReviewReview ▾

Spring PetClinic - Architectural Review & Technical Analysis

This document provides an architecture review and technical investigation of the Spring PetClinic application (spring-projects/spring-petclinic).

1. Project Overview & Tech Stack

Spring PetClinic is a reference application built to showcase modern Spring Boot patterns, clean domain-driven architecture, and Java best practices.

  • Framework: Spring Boot 4.1.0, Spring MVC, Spring Data JPA, Spring Cache
  • Java Toolchain: Java 17 (build.gradle)
  • Database & Persistence: In-memory H2, MySQL, PostgreSQL

2. Core Domain Model & Entity Hierarchy

The domain model follows standard Object-Relational Mapping (ORM) principles with common base class inheritance.

README.mdMarkdown Preview

Spring PetClinic Sample Application

Built with Spring Boot, Spring MVC, Thymeleaf, and Spring Data JPA (spring-projects/spring-petclinic).

Running PetClinic Locally

Clone the repository and start the embedded server using the Maven or Gradle wrapper:

  • Maven: ./mvnw spring-boot:run
  • Gradle: ./gradlew bootRun
  • Local Endpoint: http://localhost:8080

Database Configuration

Uses an in-memory H2 database by default, with Docker Compose profiles for mysql and postgres.

main⊗ 0 ⚠ 2 ⓘ 1Java: Ready

Gradle: Build Error

Editor extensions that embed agentic capabilities, inline completions, and side-panel interactions directly in your IDE.

Get started with IDE extensionsSupported extensions guide

agent_pipeline.pyPython 3.10+

from google.antigravity import Agent, LocalAgentConfig
from google.antigravity.hooks.policy import allow, deny

# Configure stateful agent harness and safety policies
config = LocalAgentConfig(
    policies=[allow("view_file"), deny("run_command")],
    system_instructions="You are a senior software architect.",
)

# Execute multi-turn conversation in async context
async with Agent(config) as agent:
    response = await agent.chat("Audit the repository architecture.")
    print(await response.text())

A programmatic Python SDK for building, customizing, and deploying autonomous AI agents and harnesses in your own applications and CI/CD pipelines.

Get started with Antigravity SDKTools and skills API


Core agent capabilities

Every Antigravity surface runs on a shared, highly optimized agent harness co-trained with Gemini models, combining synchronous turn-by-turn code editing and tool execution with asynchronous background subagents, scheduled cron sidecars, and automated verification loops:

Frontier Gemini models

Gemini 4 & 3.8

Co-trained with Google DeepMind. Powering all local agents with frontier speed, configurable reasoning levels, and large context capacity across Gemini 4 Argon, Gemini 3.8 Flash, and Gemini 3.1 Pro.

Model specifications

Asynchronous subagents

Parallel execution

Delegate complex background tasks to concurrent subagents working across isolated Git worktrees without blocking your active terminal or editor session.

Subagents guide

Visual artifacts

Verification loops

Track and verify agent output with structured visual reports, including interactive implementation plans, unified multi-file diffs, and browser session recordings.

Artifact inspection

Security by design

Bounded access

Enforce local execution safeguards with granular tool approval gates, workspace directory scoping, local proxying, and corporate VPC Service Controls.

Permissions and security

Google platform integrations

Curated bundles of skills, MCP servers, and extensions engineered with product teams across Google—including Firebase, Android, Modern Web, Chrome DevTools, Flutter, Google Cloud Data, Maps, Science, and the Antigravity SDK.

View all bundles


Frontier models and dynamic reasoning

Antigravity leverages Google DeepMind’s frontier Gemini models with dynamic reasoning controls. Adjust reasoning effort per conversation turn or delegate long-running tasks to autonomous agent swarms:

Interactive model and thinking level selector

Experience the model selector directly below. Antigravity lets you adjust reasoning depth and model speed dynamically per conversation turn.

Model specs and limits

† Available with paid Google AI Pro plans or on trial with eligible premium devices. To switch to a paid Google AI Pro plan, submit a request.


Product comparison

Compare primary capabilities and execution environments across Antigravity surfaces:

AllCore workflowExtensibilitySecurity & enterprise

| Capability |
App

|

CLI

|

IDE Ext.

|

SDK

|
| --- | --- | --- | --- | --- |
| Desktop application | | | | |
| Terminal user interface (TUI) | | | | |
| Editor extensions (VS Code, JetBrains, VS, Zed, Xcode, Eclipse) | | | | |
| Programmatic Python SDK | | | | |
| Remote control & headless execution | | | | |
| Workspace & Git worktree isolation | | | | |
| Multi-agent swarms (/teamwork) | | | | |
| Deep reasoning mode (/boost) | | | | |
| Interactive visual artifacts & diffs | | | | |
| Sandboxed Chromium browser automation | | | | |
| Custom skills (SKILL.md) & rule files | | | | |
| Model Context Protocol (MCP) clients | | | | |
| Scheduled tasks & cron sidecars | | | | |
| Corporate SSO & BYOID (Workforce Identity) | | | | |
| Regional data residency (US & EU) | | | | |
| VPC Service Controls (VPC-SC) | | | | |
| Request, response & telemetry logging | | | | |

Capabilities are tailored to each surface's execution target. Enterprise features under IDE Ext. apply to Antigravity IDE extensions (GA in VS Code & Visual Studio; Preview in JetBrains, Zed, and Xcode).

Features guide


User and enterprise tiers

Antigravity supports individual developers and enterprise engineering organizations with Google Cloud security, regional data residency, and VPC Service Controls:

Deployment tiers

IndividualEnterprise suitev2.5+

Individual developers

Designed for software engineers, researchers, and open-source contributors who need frontier autonomous coding assistance on their local workstation.

Free tier · Pro & Ultra plans

Desktop & terminal clients

Full access to the native Antigravity App desktop application and the high-speed agy command-line interface.

All IDE extensions

Integration with VS Code, JetBrains IDEs, Xcode, Zed, and Visual Studio with dedicated agent panels, interactive plans, and inline diff reviews.

Flexible individual plans & AI credits

Start at $0/month on the Free plan, or upgrade to Google AI Plus, Pro, or Ultra for higher rate limits and optional AI credit overages.

Skills, MCP & rules

Author project-level custom skills (SKILL.md), connect stdio/SSE MCP servers, and define behavior rule sets.

Get started freeView AI credits & plans

Enterprise engineering organizations

Engineered for organizations requiring Google Cloud governance, regional data residency, VPC Service Controls, and enterprise identity management.

Google Cloud ToS · VPC-SC

[

Corporate SSO & BYOIDGoogle Cloud & WIF

Authenticate with Google Cloud SSO, Bring Your Own Identity (BYOID) via Workforce Identity Federation (such as Okta), or Application Default Credentials.

](/docs/enterprise/)[

Regional data residencyUS, EU & Global

Control where model inference and data processing occur across global, US multi-region, and European Union multi-region endpoints.

](/docs/enterprise/)[

VPC Service ControlsGoogle Cloud

Enforce network service perimeters and private connectivity with Google Cloud VPC Service Controls (VPC-SC) and IAM role governance.

](/docs/enterprise/)[

Request & telemetry loggingCloud Audit

Configure project telemetry and full request and response logging directly within your Google Cloud project for security and compliance.

](/docs/enterprise/)[

Flexible enterprise billingLicenses or Pay-as-you-go

Choose consumption-based billing via the Gemini Enterprise Agent Platform API or Gemini Enterprise licenses with included quotas and managed overages.

](/docs/enterprise/)[

Zero model training guaranteeData privacy

Governed by the Google Cloud Terms of Service. Your prompts, code, and transcripts are never stored outside your private environments or used to train models.

](/docs/enterprise/)

Explore enterprise suiteArchitecture & case studies


Documentation pillars

Explore the documentation across six core architectural pillars:

[

Start Here

Get Started

Install the Antigravity App, set up the CLI, integrate the SDK, and configure IDE extensions.

](/docs/getting-started/)[

Concepts

Learn

Master foundational concepts including Workspaces, Git Worktrees, and Sessions.

](/docs/features/)[

Extensibility

Extend

Author custom Skills (SKILL.md), connect MCP Servers, and define Lifecycle Hooks.

](/docs/skills/)[

Security

Enterprise

Corporate SSO & BYOID, Regional Data Residency, VPC Service Controls, and Cloud Logging.

](/docs/enterprise/)[

Ecosystem

Build with Google

Firebase, Chrome DevTools, Android CLI, Dart & Flutter, Science, and Data Agent Kit skills.

](/docs/build-with-google/)[

Specs

Reference

CLI command manuals, models specifications, settings schemas, and API references.

](/docs/cli/reference/)

Activity

  1. self-assigned this
    on Nov 8, 2026
  2. added
    documentationImprovements or additions to documentation
    wontfixThis will not be worked on
    on Nov 8, 2026
  3. added theissue type on Nov 8, 2026
  4. zskbot commented on Nov 8, 2026

    @zskbot
    MemberAuthor

    Authenticate

    Available on: Antigravity App, Antigravity CLI, IDE extensions

    Antigravity uses a unified authentication layer across Antigravity App, Antigravity CLI, Antigravity for IDEs, and the Antigravity SDK. Signing in on one local surface shares your credentials across your local environment.

    Need to install Antigravity first?

    Follow the Quickstart to download and install Antigravity App, Antigravity CLI, an IDE Extension, or the Antigravity SDK.


    Sign in by surface

    Select your surface below to authenticate your account, switch credentials, or configure headless and enterprise environments.

    Antigravity App

    Personal Google Account

    1. Launch Antigravity App.
    2. Click Sign in to open the browser authentication flow.
    3. Complete the browser-based OAuth flow using your personal Google Account (Free, Pro, or Ultra plan).
    4. Once authenticated, the browser redirects back to Antigravity App and initializes your workspace.

    Gemini Enterprise & Google Cloud Account

    Google Antigravity uses a single sign-on (SSO) flow for corporate business accounts and automatically detects your assigned Gemini Enterprise license tier:

    1. Start Antigravity App and select Sign in to open the browser authentication flow.
    2. Choose Business account (subject to the Google Cloud Terms of Service).
    3. Select Continue with Google Cloud (or select Advanced WIF Configuration and enter your organization’s WIF Configuration String for Bring Your Own Identity / Workforce Identity Federation).
    4. Complete authentication in your browser.
    5. Once authenticated, the License Selector displays your assigned licenses.
    6. Confirm the project linked to your license and select it. Alternatively, select Other to self-assign a license by entering your Google Cloud Project ID and selecting a location (global, us, or eu).

    Note

    Project Switching & Governance: To switch to a different Google Cloud project or location, log out of Antigravity App and log back in to select your new project or region. Ensure the Gemini Enterprise API (aiplatform.googleapis.com) is enabled in your project. See the Enterprise guide for full administrator setup, BYOID/WIF, and VPC Service Controls details.

    Application Default Credentials (ADC)

    For automated or headless workflows, you can authenticate Antigravity App using Google Cloud Application Default Credentials:

    1. Generate local Application Default Credentials for your project:

      gcloud auth application-default login --project {GCP_PROJECT}
    2. Enable ADC in Antigravity App using either option:

      • Launch from a terminal with export AGY_ADC_AUTH=true.

      • Or enable the enableAdc user setting in ~/.gemini/config/config.json:

        {
          "userSettings": {
            "enableAdc": true
          }
        }

    Antigravity CLI

    Antigravity CLI (agy) uses secure credentials and token profiles to communicate with the shared agent harness.

    Local silent keyring sign-in

    When launching agy on your local machine, the CLI attempts to access your operating system’s native secure keyring (such as Apple Keychain, Linux Secret Service/dbus, or Windows Credential Manager). If a valid token profile is found, the CLI authenticates your session silently without opening a browser.

    If no saved session is found:

    1. The CLI automatically launches your local default web browser.
    2. Sign in using your approved account credentials (personal Google Account, or choose Business account → Continue with Google Cloud / Advanced WIF Configuration for Gemini Enterprise).

    Remote SSH OAuth flow

    When running over SSH, the CLI detects the remote connection environment. Because it cannot launch a local web browser, the CLI initiates a manual URL loop:

    1. Launch agy in your remote terminal session.
    2. The CLI detects the SSH environment and prints a unique, secure authorization URL.
    3. Copy this URL and paste it into a web browser on your local machine.
    4. Sign in with your approved credentials and complete the authentication.
    5. The browser displays a unique alphanumeric authorization code.
    6. Copy this code, return to your remote SSH terminal, and paste it into the prompt.

    Using a Gemini API key

    Run Antigravity CLI with your own Gemini API key instead of a signed-in Google account. Model requests go directly to the Gemini API, and the CLI never establishes an account session. This suits headless and CI runs, where no browser is available to complete a sign-in. Create a key in Google AI Studio.

    To use a Gemini API key, you have to set a provider and an environment variable with the API key. Only setting a GEMINI_API_KEY environment variable on its own has no effect.

    Enable the Gemini API key
    1. Set modelProvider to gemini in ~/.gemini/antigravity-cli/settings.json:

      {
        "modelProvider": "gemini"
      }
    2. Export your key as GEMINI_API_KEY:

      export GEMINI_API_KEY="your-api-key"

      This applies to the current shell only. Add the same line to your shell profile, such as ~/.zshrc or ~/.bashrc, to persist it across sessions.

    3. Start the CLI:

      agy

    The CLI skips the sign-in screen and opens the main interface directly. The header shows Gemini API key instead of your account email:

    Antigravity CLI authenticated with a Gemini API key, with "Gemini API key" shown in the header in place of an account email

    Note: When you use the authentication with a GEMINI_API_KEY, /logout has no effect because there is no stored session to clear.

    Point the CLI to a custom endpoint

    To send model requests to a different Gemini-compatible endpoint, set the GOOGLE_GEMINI_BASE_URL environment variable:

    export GOOGLE_GEMINI_BASE_URL="https://your-endpoint.example.com"
    Revert to default authentication

    If you want to revert back to using the default account based authentication:

    1. Remove modelProvider from ~/.gemini/antigravity-cli/settings.json.
    2. Restart the CLI to sign in to your account.

    Note: The CLI cannot start if you unset the GEMINI_API_KEY environment variable, but still have the modelProvider set to gemini.

    Troubleshooting API key authentication
    Symptom Cause Resolution
    The CLI exits at startup reporting that GEMINI_API_KEY is not set modelProvider is gemini but the key is missing from the environment Export GEMINI_API_KEY, or remove modelProvider to use the default authentication
    The CLI signs in normally and ignores the setting modelProvider holds an unrecognized value gemini is the only accepted value. Check the spelling and restart the CLI
    A key set through GOOGLE_API_KEY or a .env file has no effect The CLI reads the credential only from GEMINI_API_KEY in the environment, and does not load .env files Export GEMINI_API_KEY in your shell or shell profile
    Requests fail mid-session with a generic model error The key is invalid, revoked, or lacks access to the requested model Verify the key in Google AI Studio. The CLI checks only that the key is non-empty at startup, so an unusable key only surfaces on the first conversation

    Application Default Credentials (ADC)

    For headless enterprise environments and automated terminal workflows, Antigravity CLI supports Google Cloud Application Default Credentials (ADC):

    1. Generate local ADC credentials for your project:

      gcloud auth application-default login --project {GCP_PROJECT}
    2. Enable ADC in the CLI by setting AGY_ADC_AUTH=true (and optionally GOOGLE_CLOUD_QUOTA_PROJECT and GOOGLE_CLOUD_LOCATION):

      export AGY_ADC_AUTH=true
    3. To sign out of ADC on the CLI, unset the environment variable (unset AGY_ADC_AUTH) and restart your terminal session.

    Managing your session (logging out)

    Terminating your session clears active credentials and local cache directories. To disconnect your account and purge saved authentication profiles from your operating system’s keyring, run the following command in the CLI prompt box:

    /logout
    

    Installation flags

    When executing the CLI installation scripts (see Quickstart), you can append the following customization flags:

    • --skip-aliases: Bypasses shell profile alias purging (prevents the script from purging or updating legacy agy or antigravity shell aliases).
    • --skip-path: Bypasses shell profile PATH appending (prevents the script from modifying your shell profile’s dynamic environment variables).

    For IDEs

    Unified sign-in across IDE extensions

    Antigravity uses a unified authentication system across all IDE extensions (Visual Studio Code, IntelliJ-based IDEs, Visual Studio, Zed, Xcode, and Eclipse), the Antigravity CLI, and Antigravity App.

    • Individual developers: Click the Antigravity icon in your editor’s activity bar or tool window (or select Antigravity and use the … button under Xcode > Settings > Intelligence in Xcode) and click Sign In to complete authentication in your default web browser.
    • Enterprise teams & organizations: Click Sign In and select Sign in with Gemini Enterprise (or choose Business account → Continue with Google Cloud / BYOID), then select your target Google Cloud Project and Region (global, us, or eu). Gemini Enterprise authentication for JetBrains IDEs, Zed, Xcode, and Eclipse is currently in Preview.

    Agent configuration (auth.type)

    In editors that configure external agents via JSON settings (IntelliJ-based IDEs, Zed, and Xcode), configure your authentication method using the following settings:

    Google Accounts

    For individual Google AI subscription plans (including Free, Pro, and Ultra):

    {
      "auth": {
        "type": "oauth-personal"
      }
    }

    Gemini Enterprise

    Reach out to your cloud administrator for your GCP Project ID and Region (Enterprise support for JetBrains, Zed, Xcode, and Eclipse is currently in Preview):

    {
      "auth": {
        "type": "oauth-business"
      },
      "gcp": {
        "project": "<YOUR_GCP_PROJECT_ID>",
        "location": "<YOUR_GCP_REGION>"
      }
    }

    Gemini API

    Authenticate using a Gemini API key:

    {
      "auth": {
        "type": "gemini-api-key"
      }
    }

    API / ADC

    Authenticate with Gemini Enterprise using an API Key or Application Default Credentials (ADC) (Preview). When using ADC, specify your GCP Project ID and Region:

    {
      "auth": {
        "type": "agent-platform"
      },
      "gcp": {
        "project": "<YOUR_GCP_PROJECT_ID>",
        "location": "<YOUR_GCP_REGION>"
      }
    }

    Antigravity SDK

    Programmatic authentication

    The Google Antigravity SDK (google-antigravity) supports the following authentication methods for Python 3.10+ applications (see SDK Overview):

    1. Gemini API Key: Set your Gemini API key in the environment before initializing Agent:

      export GEMINI_API_KEY="your_api_key_here"
    2. Gemini Enterprise (Google Cloud / Vertex AI): Set vertex=True in LocalAgentConfig alongside your GCP project and location to route requests through the Gemini Enterprise API (aiplatform.googleapis.com):

      from google.antigravity import Agent, LocalAgentConfig
      
      config = LocalAgentConfig(
          vertex=True,
          project="your-gcp-project",
          location="us-central1",
      )

      Or configure Gemini Enterprise via environment variables and Application Default Credentials:

      export GOOGLE_GENAI_USE_VERTEXAI=True
      export GOOGLE_CLOUD_PROJECT="your-gcp-project"
      export GOOGLE_CLOUD_LOCATION="us-central1"
      gcloud auth application-default login
    3. Local models (offline, no API key required): Run agents directly on-device using LiteRTAgentConfig or connect to a local OpenAI-compatible server (Ollama, LM Studio, vLLM) using LocalOpenAIAgentConfig. See Local models.


    Next steps

    [

    Get Started

    Quickstart

    Install Antigravity and complete your first end-to-end agent workflow.

    ](/docs/getting-started)[

    Models

    Choose a model

    Configure Gemini 4 Argon, Gemini 3.1 Pro, Gemini 3.8 Flash, Claude, and reasoning effort.

    ](/docs/models)[

    Cloud & APIs

    Build with Google

    Connect Vertex AI, Cloud Run, Firebase, and Google Cloud MCP tools.

    ](/docs/build-with-google)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

documentationImprovements or additions to documentationgood first issueGood for newcomerswontfixThis will not be worked on

Type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions