Antigravity (Google)

Antigravity is an agentic software development platform from Google, launched in public preview on 18 November 2025 alongside the release of gemini 3 pro.^[1][2] Built as a heavily modified fork of Visual Studio Code, Antigravity combines a conventional AI-powered editor with an "agent manager" surface that orchestrates autonomous agents across an integrated code editor, terminal, and Chromium browser.^[1][3] It was built by the core engineering team behind windsurf, whose CEO Varun Mohan, co-founder Douglas Chen, and top researchers Google hired in a US$2.4 billion licensing-and-talent deal in July 2025.^[13] The platform is Google's direct response to AI-first development tools such as cursor, Windsurf, and claude code, and is positioned as the showcase environment for Gemini 3's agentic and coding capabilities.^[4][2] The original IDE supports model choice across Google's Gemini 3 family, Anthropic's claude sonnet 4 5, and OpenAI's GPT-OSS, and launched free for individuals during public preview.^[1][3] At Google I/O 2026 on 19 May 2026, Google announced Google Antigravity 2.0 as a new, separate desktop command-center application without an IDE.^[16][17] Google also announced a CLI, an SDK, a managed Antigravity Agent in the Gemini API, and Gemini Enterprise access as distinct surfaces in the wider Antigravity ecosystem.^[5][6][25]

Overview

The original Antigravity IDE is built around what Google calls an "agent-first" interface model.^[1] Rather than treating an AI assistant as a sidebar inside an editor, it places autonomous agents at the center of the developer workflow and presents two primary views: an Editor view, which resembles a conventional IDE with tab completions and inline commands, and a Manager view, a dashboard-style surface where developers queue tasks, spawn agents, and monitor parallel work.^[1][7] Each task delegated to the platform produces a set of verifiable outputs that Antigravity calls Artifacts, including task lists, implementation plans, screenshots, and browser recordings.^[1][8]

The original IDE is available for macOS, Windows, and Linux as a native desktop application.^[1][3] Its underlying editor is a fork of Visual Studio Code, similar to the architecture used by Cursor and the Windsurf editor.^[3][7] Google Antigravity 2.0 is also available on those desktop platforms, but it is a separate command center with no embedded editor. Google recommends using it alongside an IDE when developers want both direct editing and asynchronous agent orchestration.^[16][17]

Google's current pricing page describes Antigravity as generally available. The US$0 Individual plan includes basic weekly rate limits, unlimited tab completions, unlimited command requests, and access to listed Google, Anthropic, and OpenAI models.^[18] The historical public preview and its rate limits remain relevant to the original November 2025 launch, but they no longer describe the product's current release status.

History

Origins and launch

Google announced Antigravity on 18 November 2025, the same day it released gemini 3 pro across the Gemini API, Google AI Studio, and vertex ai.^[2][1] The launch was timed to demonstrate the new model's agentic capabilities in a developer-facing product. Google's blog post introducing Gemini 3 specifically called out Antigravity as the showcase environment for the model's tool-use behavior and pointed to a Terminal-Bench 2.0 score of 54.2% as evidence of the model's ability to operate a computer via a terminal.^[2]

A Developers Blog post titled "Build with Google Antigravity, our new agentic development platform" introduced the product on the same day.^[1] The post described the platform as "an agentic development platform that combines a familiar, AI-powered coding experience with a new agent-first interface" and framed its central design idea with the line, "Agents shouldn't just be chatbots in a sidebar; they should have their own dedicated space to work."^[1]

The product was made available at antigravity.google with downloads for macOS, Windows, and Linux. Coverage by The New Stack, DevClass, and other outlets noted that Antigravity arrived in the same week as several other Gemini 3-aligned product updates, positioning the IDE as the centerpiece of Google's late-2025 developer push.^[9]

Who built Antigravity? The Windsurf connection

Antigravity was built by the engineering team behind Windsurf, the AI coding editor formerly known as Codeium. On 11 July 2025, Google agreed to pay US$2.4 billion for a non-exclusive license to Windsurf's technology and to hire the startup's CEO, Varun Mohan, its co-founder Douglas Chen, and several of its top researchers into google deepmind.^[13] The arrangement, a so-called "reverse acquihire," followed the collapse of a reported US$3 billion acquisition of Windsurf by openai.^[13] A Google spokesperson, Chris Pappas, said the company was "excited to welcome some top AI coding talent from Windsurf's team to Google DeepMind to advance our work in agentic coding."^[13] Most of Windsurf's roughly 250-person staff did not move to Google and continued the product independently; the remainder of the company was subsequently acquired by Cognition AI, the maker of the Devin coding agent.^[13] This shared lineage explains why Antigravity, like Windsurf, is a fork of Visual Studio Code and carries over agentic workflow concepts from the Windsurf editor.^[13]

Early adopter reception

Reception in the days following launch was mixed. DevClass reported that some early adopters exhausted their free credits within roughly 20 minutes of use, with no in-product purchase path available at that point, and that the service produced "Agent taking unexpectedly long to load" and "Agent terminated due to error" messages citing model provider overload.^[9] Other early hands-on coverage praised the parallel-agent workflow and the browser automation features, which removed the need for separate browser-control tooling.^[9]

Security disclosure and patch

Researchers at Pillar Security privately reported a prompt-injection vulnerability in Antigravity's find_by_name file-search tool to Google on 7 January 2026, and disclosed it publicly on 20 April 2026.^[10][11] The tool passed the value of its Pattern parameter into the underlying fd command-line utility without argument termination, allowing an attacker to inject command-line flags. By supplying the -X (exec-batch) flag, the exploit could escalate a routine file search into arbitrary command execution, and it could be triggered through indirect prompt injection from a document the agent was asked to read.^[10][11] Because find_by_name was classified as a native tool rather than a shell command, the existing Secure Mode protections did not block the invocation; Pillar reported that the exploit worked "with both Secure Mode enabled and Auto Execution disabled," the most restrictive configuration available.^[10] The researchers wrote that "native tool invocations are executed by the agent directly, before Secure Mode's command-level restrictions are evaluated," so "the security boundary that Secure Mode enforces simply never sees this call."^[10] Google marked the issue fixed on 28 February 2026 and awarded a bounty through its AI Vulnerability Reward Program.^[10][11] The case became an early reference point for discussions of prompt injection in agentic IDEs.

Antigravity 2.0

At Google I/O 2026 on 19 May 2026, Google announced Google Antigravity 2.0, a standalone command-center application for macOS, Linux, and Windows.^[16][17] Despite the version number, it is not a refreshed edition of the original VS Code-derived IDE. It is a separate app with no embedded editor, built around the asynchronous task-management ideas of the original Agent Manager.^[16] Existing IDE installations automatically updated to 2.0 and offered users a choice to keep the IDE. Google said both products would remain available for the time being, with a later plan to remove Agent Manager from the IDE and leave it as an agent-powered editor.^[16]

The 2.0 app added dynamic subagents, asynchronous background tasks, scheduled tasks, Projects that can span several folders, JSON lifecycle hooks, native Git worktree support, slash commands, and live voice transcription.^[16][19] A subagent inherits its parent agent's tool configuration and security permissions and cannot gain broader authority. Antigravity can also create isolated worktrees automatically for subagents so that parallel changes do not share one working tree.^[19] Scheduled tasks used Gemini 3.5 Flash at launch, and live voice transcription was available across subscription tiers.^[19]

Google introduced several related but separate surfaces at the same event: Antigravity CLI, an Apache-licensed Python SDK released as a research preview, a managed Antigravity Agent in the Gemini API, and enterprise access through the Gemini Enterprise Agent Platform.^{[5][20][21][22]} These products share parts of the Antigravity agent harness, but they are not modes contained inside the 2.0 desktop app. The CLI is a terminal interface, the SDK embeds the harness in Python applications, and the managed API surface runs an agent in a Google-hosted Linux environment.^[20][21][23]

The 2.0 announcement also introduced a Google AI Ultra subscription priced at US$100 per month, with priority access and five times the Antigravity capacity of Google AI Pro at launch.^[5][6][25] The default Antigravity 2.0 model at launch was Gemini 3.5 Flash, which Google positioned as a faster option than the previously default Gemini 3 Pro for agentic tasks.^[5][14]

Transition from Gemini CLI

On the same day as Antigravity 2.0, Google published a transition notice announcing that gemini cli would stop serving requests for consumer (Google AI Pro, AI Ultra, and free) users on 18 June 2026, with users redirected to Antigravity CLI.^[12] Enterprise customers with Standard or Enterprise licenses retained access to Gemini CLI alongside the new tooling.^[12] Antigravity CLI was rewritten in Go, designed for asynchronous workflows, and shared an agent harness with the Antigravity 2.0 desktop application.^[12][6] Existing Gemini CLI features including Agent Skills, Hooks, Subagents, and Extensions were carried over as Antigravity plugins.^[12]

Architecture

Editor and agent surfaces

The original Antigravity IDE presents two primary surfaces.^[1][7] The Editor view is a conventional VS Code-derived IDE with file tree, multi-pane editing, tab completions, and inline command boxes. It is intended for synchronous work where a developer wants to drive cursor placement and review each change. The Manager view is a separate workspace that lists active agents, their queued tasks, and their generated Artifacts.^[1][7] Up to five agents can run in parallel by default, with each agent assigned a workspace and a model.^[7]

The Manager view is described in Google's documentation as "Mission Control" for the original platform.^[1] From it, developers can dispatch new agents, inspect the state of running agents, attach feedback to in-progress Artifacts, and pause or resume specific agents.^[1][7]

Antigravity 2.0 command center

Antigravity 2.0 turns the management surface into a standalone app. Its central view starts and monitors asynchronous tasks while code can remain open in a separate editor.^[16][17] Projects group folders, settings, resources, and permissions, and one Project can span several filesystem roots.^[19] A main agent can create subagents dynamically, while native Git worktrees let those agents work in isolation and later return results to the parent task.^[19]

What are Artifacts?

Artifacts are the central mechanism Antigravity uses to mediate trust between the developer and the autonomous agent.^[1][8] Rather than emitting only a final diff, an agent working in Antigravity produces a sequence of intermediate, human-readable outputs as it proceeds with a task:

• Task lists outlining what the agent intends to do before it begins.
• Implementation plans describing the structure of upcoming code changes.
• Screenshots produced by the platform's built-in Chromium browser.
• Browser recordings showing end-to-end UI walkthroughs.
• Diff summaries describing what changed and why.

Developers can comment on any Artifact, similar to a document review, and the agent incorporates the feedback into its ongoing run rather than restarting.^[8][7]

Integrated browser and terminal

Antigravity bundles a Chromium-based browser into the platform so that agents can navigate, click, fill forms, and capture screenshots without a separate browser-control tool.^[1][7] This makes the platform an integrated ai browser agent environment in addition to a code editor. The included terminal allows agents to install dependencies, run build commands, start dev servers, and inspect logs in the same session.^[1] Google's documentation describes the typical loop as "write code for a new feature, use the terminal to launch the application, and then use the browser to test and verify that the new component is functioning as expected."^[1]

Secure Mode and isolation

The original Antigravity IDE exposes a Secure Mode that restricts network access and limits writes outside the workspace.^[10][11] As the 2026 Pillar Security disclosure demonstrated, native tool invocations were not subject to the same checks as shell commands in the initial release, which prompted Google to revise its execution model.^[10][11]

Local and hosted Antigravity surfaces have different security boundaries. For the local agent, Google's settings documentation says default filesystem access covers Project folders and ~/.gemini/antigravity; access outside those locations is an explicit opt-in.^[26] Permission rules can allow, deny, or ask about operations, with denial taking precedence, and an operating-system sandbox can add another local boundary.^[27] Antigravity 2.0 subagents inherit the main agent's tools and security permissions and receive no additional rights.^[19]

The resumable, sandboxed Linux environment described at I/O belongs to the managed Antigravity Agent in the Gemini API, not to local Antigravity 2.0 execution.^[23] That managed surface can keep an environment across interactions, but it remains a preview API and should not be treated as evidence that local desktop tasks run in a Google-hosted sandbox.

Knowledge base

Antigravity includes a persistent knowledge base that agents can write to and read from across sessions, used to store project context, code snippets, and remembered instructions.^[1] The knowledge base is workspace-scoped and is intended to reduce the cost of long-running multi-session tasks where the agent would otherwise need to rediscover project structure.^[1]

Features

Multi-file edits and planning

Antigravity agents can plan and execute changes that span multiple files, generating an implementation plan as an Artifact before applying edits.^[1][7] This pattern of plan-then-edit is shared with claude code and Cursor's Composer mode, but Antigravity makes the plan itself a first-class, commentable Artifact rather than an inline message.^[7][8]

Terminal and browser control

Through the integrated terminal, agents can install packages, start build pipelines, and run tests. Through the integrated browser, they can navigate to a local dev server, click through user flows, and capture screenshots or recordings as evidence.^[1][7] This combination is the platform's most distinctive feature compared to editor-only AI tools and is part of what makes Antigravity an agentic workflow platform rather than a code completion product.^[7]

Parallel agents

The original IDE's Manager view allows the developer to queue several tasks that run in parallel, each with its own agent.^[7] Antigravity 2.0 added dynamic subagents, where a primary agent can spawn helper agents for bounded subtasks and aggregate their results.^[16][19] The helper inherits the primary agent's tool configuration and permission boundary. Native Git worktree support can give it an isolated checkout, reducing collisions between parallel changes.^[19]

Multimodal input

Antigravity can accept code, natural language descriptions, images, and API responses as input.^[1][3] A developer can supply a screenshot of a desired UI and have the agent translate it into front-end code, an interaction that overlaps with Google's separate work on multimodal Gemini models.^[1]

Plugins and skills

The platform supports plugins and skills, with Antigravity 2.0 carrying over Agent Skills, Hooks, Subagents, and Extensions concepts previously used by gemini cli.^[12] Skills are markdown instruction packages with associated tools and can be installed globally or scoped to a workspace.^[19] The 2.0 app also supports MCP connections and JSON hooks that can invoke local shell scripts before or after tool calls, model calls, and other lifecycle events.^[17][19] The June 2026 desktop releases added a built-in Antigravity Guide skill that answers questions about the platform itself.^[15]

Scheduled tasks and voice input

Scheduled Tasks let users define recurring or later work from the command center. At launch, these tasks ran with Gemini 3.5 Flash rather than the model selected for an ordinary interactive task.^[19] Live voice transcription converts spoken instructions to text and was made available on Free, Pro, and Ultra tiers.^[16][19] These features operate in the standalone 2.0 app and do not turn it into an IDE.

What models does Antigravity support?

Default models

At launch, Antigravity defaulted to Gemini 3 Pro for most agentic tasks, with Gemini 3 Flash available for faster, lighter actions.^[1][2] Gemini 3 Pro was reported by Google to score 54.2% on Terminal-Bench 2.0, a benchmark that "tests a model's tool use ability to operate a computer via terminal," and to top the WebDev Arena leaderboard with an Elo score of 1487 at launch.^[2]

Antigravity 2.0 changed the default to Gemini 3.5 Flash, which Google released on 19 May 2026 and made the default across the Gemini app, AI Mode in Search, and Antigravity.^[14][5] Google said the model was tuned for speed in agentic workflows, stating that "when looking at output tokens per second, it is 4 times faster than other frontier models."^[14] Google reported that Gemini 3.5 Flash outperformed the intermediate Gemini 3.1 Pro model on coding and agentic benchmarks:

Scroll sideways for more →

Google positioned the model at both developers and analysts, saying that "what used to take a developer days or an auditor weeks, 3.5 Flash can now help complete in a fraction of the time, often at less than half the cost of other frontier models."^[14] These are provider-reported model benchmarks, not end-to-end evaluations of Antigravity 2.0 task completion, permission behavior, security, or resistance to prompt injection.

Third-party models

Antigravity has supported third-party models since its initial release. It offered Anthropic's claude sonnet 4 5 and OpenAI's open-weight GPT-OSS at launch, and later added claude sonnet 4 6 and Claude Opus 4.6.^[1][3] The current Individual plan lists Gemini 3.5 Flash, Gemini 3.1 Pro, Gemini 3 Flash, Claude Sonnet 4.6, Claude Opus 4.6, and gpt-oss-120b as included models.^[18] Model selection in the desktop experience is per agent or task.^[7]

Model routing

The managed Antigravity Agent in the Gemini API is a separate preview surface with explicit model configuration. As of July 2026, its documentation identifies antigravity-preview-05-2026 as the base agent, uses Gemini 3.6 Flash by default, and lists Gemini 3.6 Flash, Gemini 3.5 Flash, and Gemini 3.5 Flash-Lite as supported models.^[23] This managed-agent default does not establish the default model in the local 2.0 app.

Custom managed agents also have preview limitations. Current documentation says their schemas may change, they have no versioning or rollback, and they cannot nest subagents. The documented limit is 1,000 saved managed agents, and antigravity-preview-05-2026 is the only supported custom-agent base.^[24] These API limits do not apply to dynamic subagents in the local 2.0 command center.

How much does Antigravity cost?

Public preview

Antigravity launched in public preview on 18 November 2025 and was free for individual users, with what Google described as "generous" rate limits.^[1][3] That preview included the original desktop IDE, Manager view, parallel agents, browser and terminal integration, and Artifacts.^[1] The current pricing page labels Antigravity generally available rather than in preview.^[18]

The current US$0 Individual plan includes basic weekly rate limits, unlimited tab completions, unlimited command requests, and the listed Google, Anthropic, and OpenAI models.^[18] Thus, the old preview terms remain launch history but no longer describe the current release status or model entitlement.

Google AI Pro

Antigravity is included in Google AI Pro, the consumer tier above the free plan.^[5] The current plan provides more generous rate limits than Individual and a flexible pool of AI credits that can be allocated to Antigravity usage.^[18]

Google AI Ultra

Google introduced Google AI Ultra for Antigravity on 19 May 2026 at US$100 per month.^[5][6] At launch, the plan provided priority access and five times the capacity of the Pro tier.^[5][25] The current pricing page describes Ultra as the individual tier with the most generous rate limits and AI credit pool.^[18]

Enterprise

Organizations can access Antigravity 2.0, CLI, and SDK through the Gemini Enterprise Agent Platform and Google Cloud.^[22] The offering uses consumption pricing and Google Cloud project integration under Google Cloud Terms, rather than a public per-seat price.^[18][22] Google says enterprise prompts, responses, code, and telemetry are not stored outside the customer's private environment.^[22] Additional administrator controls and IDE extensions were described as coming soon at launch, so they were not part of the initial enterprise availability.^[22]

How does Antigravity compare to Cursor, Windsurf, and Claude Code?

Comparison to Cursor

cursor is the most direct predecessor in the AI-first IDE category and, like Antigravity, is built on a fork of Visual Studio Code.^[7] Cursor was first released in 2023 and emphasizes inline editing, fast tab completion, and a Composer mode for multi-file edits. Cursor allows the developer to select among GPT, Claude, and Gemini models per task.^[7] The original Antigravity IDE differs by making Agent Manager a peer to the editor, bundling a browser, and treating Artifacts as commentable, reviewable objects.^[7][8] Antigravity 2.0 goes further toward orchestration by removing the embedded editor and operating as a separate command center.^[16]

Scroll sideways for more →

Comparison to Windsurf

windsurf (formerly Codeium's editor) is also a VS Code fork and provides Cascade, an agentic mode that performs multi-file edits, runs terminal commands, and indexes large codebases.^[7] Windsurf has positioned itself as a strong option for monorepos and enterprise codebases, with automatic codebase context retrieval as a differentiator.^[7] Antigravity matches Windsurf's multi-file editing and terminal control but adds browser automation and a dedicated parallel-agent surface that Windsurf does not provide in the same form.^[7] Notably, the two products share engineering DNA: Google built Antigravity with the Windsurf leadership and researchers it hired in July 2025, while the Windsurf product continued independently under separate ownership.^[13]

Comparison to Claude Code

claude code is Anthropic's command-line and IDE-extension coding agent. It is a terminal-first agent rather than a fork of an editor, which means it runs inside whatever editor or shell the developer already uses.^[7] The original Antigravity product ships its own editor and browser, while Antigravity 2.0 removes the editor and keeps a dedicated command-center interface. Both product generations allow agents to drive a terminal and edit files; Claude Code's primary interface remains the command line. Anthropic's models are available inside Antigravity, but Claude Code remains the canonical Anthropic-first agentic experience.^[7]

Comparison to GitHub Copilot

github copilot began as a code completion tool and has expanded into agent modes including Copilot Workspace.^[7] Copilot is integrated into GitHub's product surface and Microsoft's editors and tends to emphasize integration with pull requests and code review. Antigravity does not aim at the same pull-request surface and instead emphasizes long-running parallel agents working on local repositories with browser and terminal control.^[7]

Reception

Initial reception in November 2025 was mixed but oriented toward cautious enthusiasm. Coverage by The New Stack and DevClass described the agent-first interface as a genuine product idea rather than a repackaging of existing IDE tooling.^[9] Reviewers praised the parallel-agent Manager view, the bundled browser, and the Artifacts model for making agent behavior more inspectable.^[7][9]

Negative coverage focused on stability and capacity issues. Early adopters reported credit exhaustion within minutes of starting non-trivial tasks, repeated "model provider overload" errors, and difficulty completing complex tasks without significant manual intervention.^[9] Some reviewers compared the experience unfavorably to the more mature, less ambitious Cursor at launch, although they tended to credit Google with offering a richer baseline feature set.^[9]

The Pillar Security disclosure in April 2026 added a security dimension to the reception of the product.^[10][11] Coverage by CSO Online, Dark Reading, and The Hacker News framed Antigravity as a case study for the systemic risks of agentic IDEs, in which a single indirect prompt injection inside a document could escalate to remote code execution despite Secure Mode being enabled.^[10][11] Google's prompt patching and acknowledgment through its Vulnerability Reward Program were viewed positively, but several commentators argued that the incident showed why sandbox-style execution isolation was preferable to input sanitization for agentic platforms.^[10][11]

By the time of Antigravity 2.0 at Google I/O 2026, coverage had shifted toward Google's multi-surface ecosystem strategy. Reports emphasized the separate no-IDE command-center app, the CLI and SDK, competition with ChatGPT Pro and Claude Max at the US$100 price tier, and Google's decision to sunset gemini cli for consumer users in favor of Antigravity CLI.^{[5][6][12][25]}

Integration with the Google ecosystem

Antigravity is closely tied to Google's broader developer stack. Google AI Studio can export an application into Antigravity, while integrations cover Firebase backend services and Android development targets.^[5] Enterprise access runs through the Gemini Enterprise Agent Platform and Google Cloud projects.^[22] These integrations connect separate product surfaces rather than placing every capability inside the 2.0 desktop app.

The managed Antigravity Agent exposes the agent harness through the Gemini API so an application can start background work and resume a persistent hosted environment across interactions.^[23] This is the surface that runs inside a Google-hosted Linux sandbox. The local 2.0 command center and original IDE operate against local workspaces under their own permission settings.^[19][26]

This ecosystem alignment is one of Antigravity's differentiators relative to Cursor, Windsurf, and Claude Code.^[7][25] For organizations already using Google Cloud, the enterprise offering provides project integration, cloud-governed terms, and consumption billing.^[18][22]

Is Antigravity generally available?

Google's current pricing page describes Antigravity as generally available.^[18] Antigravity 2.0 is available for macOS, Windows, and Linux as a separate command-center app, while the original Antigravity IDE remains available as its own product.^[16][17] The original desktop downloads list macOS Monterey 12 or later, Windows 10 or later, and 64-bit Linux distributions with glibc 2.28 or later.^[3]

The Antigravity CLI replaced Gemini CLI for consumer Free, Google AI Pro, and Google AI Ultra users after the transition announced for 18 June 2026; eligible enterprise customers retained Gemini CLI access alongside the newer tooling.^[12] The Python SDK was introduced as a research preview rather than a generally available library.^[21] The managed Antigravity Agent and custom-agent API also remain preview features, with documented schema, versioning, and subagent limitations.^[23][24]

Google continued to ship rapid point releases through mid-2026. The June desktop updates added audio rendering, PDF attachments to Gemini messages, improved substring file search, and a built-in Antigravity Guide skill.^[15] Access is tied to a Google account and plan. Individual, Pro, and Ultra tiers offer progressively higher limits, while the current free Individual tier includes the listed Anthropic and OpenAI models rather than requiring a separate model-provider subscription.^[18]

See also

• cursor
• windsurf
• claude code
• github copilot
• gemini 3 pro
• gemini 2 5 pro
• gemini cli
• google deepmind
• google ai studio
• google vertex ai
• openai
• agentic workflow
• ai browser agent
• prompt injection
• indirect prompt injection
• vibe coding
• claude sonnet 4 5

References

1. ^Google Antigravity Team, "Build with Google Antigravity, our new agentic development platform", Google Developers Blog, 2025-11-18. developers.googleblog.com/...-development-platform Accessed 2026-05-25.
2. ^Google, "Gemini 3 for developers: New reasoning, agentic capabilities", The Keyword (blog.google), 2025-11-18. blog.google/...gemini-3-developers Accessed 2026-05-25.
3. ^Google, "Google Antigravity (product homepage)", antigravity.google, 2025-11-18. antigravity.google Accessed 2026-05-25.
4. ^Codecademy, "Agentic IDE Comparison: Cursor vs Windsurf vs Antigravity", Codecademy, 2026-02-15. codecademy.com/...rsor-vs-windsurf-vs-antigravity. Accessed 2026-05-25.
5. ^Google, "I/O 2026 developer highlights: Antigravity, Gemini API, AI Studio", The Keyword (blog.google), 2026-05-19. blog.google/...google-io-2026-developer-highlights Accessed 2026-05-25.
6. ^Ivan Mehta, "Antigravity turns into a full agentic development platform with desktop app, CLI, and SDK", The Next Web, 2026-05-19. thenextweb.com/...avity-2-desktop-cli-sdk-io-2026. Accessed 2026-05-25.
7. ^Codecademy, "Agentic IDE Comparison: Cursor vs Windsurf vs Antigravity", Codecademy, 2026-02-15. codecademy.com/...rsor-vs-windsurf-vs-antigravity. Accessed 2026-05-25.
8. ^Arjan KC, "Google Antigravity Agent Manager Explained: Deep Dive", arjankc.com.np, 2026-01-22. arjankc.com.np/...igravity-agent-manager-explained Accessed 2026-05-25.
9. ^Tim Anderson, "We take a look at Google's Antigravity: Agentic AI development but some frustrations for early adopters", DevClass, 2025-11-19. devclass.com/...but-frustrating-for-early-adopters Accessed 2026-05-25.
10. ^Pillar Security, "Prompt Injection leads to RCE and Sandbox Escape in Antigravity", Pillar Security blog, 2026-04-20. pillar.security/...-sandbox-escape-in-antigravity. Accessed 2026-07-08.
11. ^Lucian Constantin, "Prompt injection turned Google's Antigravity file search into RCE", CSO Online, 2026-04-22. csoonline.com/...ravity-file-search-into-rce.html. Accessed 2026-05-25.
12. ^Google Developers, "An important update: Transitioning Gemini CLI to Antigravity CLI", Google Developers Blog, 2026-05-19. developers.googleblog.com/...li-to-antigravity-cli Accessed 2026-05-25.
13. ^TechCrunch, "Windsurf's CEO goes to Google; OpenAI's acquisition falls apart", TechCrunch, 2025-07-11. techcrunch.com/...-openais-acquisition-falls-apart Accessed 2026-07-08.
14. ^Google, "Gemini 3.5: frontier intelligence with action", The Keyword (blog.google), 2026-05-19. blog.google/...gemini-3-5 Accessed 2026-07-08.
15. ^Google, "Google Antigravity Changelog", antigravity.google, 2026-06-25. antigravity.google/changelog. Accessed 2026-07-08.
16. ^Google Antigravity Team, "Introducing Google Antigravity 2", Google Antigravity Blog, 2026-05-19. antigravity.google/...ducing-google-antigravity-2. Accessed 2026-07-24.
17. ^Google, "Google Antigravity 2.0", product page. antigravity.google/...antigravity-2. Accessed 2026-07-24.
18. ^Google, "Google Antigravity Pricing", product page. antigravity.google/pricing. Accessed 2026-07-24.
19. ^Google Antigravity Team, "Google I/O 2026 feature deep dive", Google Antigravity Blog, 2026-05-19. antigravity.google/...e-io-2026-feature-deep-dive. Accessed 2026-07-24.
20. ^Google Antigravity Team, "Introducing Google Antigravity CLI", Google Antigravity Blog, 2026-05-19. antigravity.google/...cing-google-antigravity-cli. Accessed 2026-07-24.
21. ^Google Antigravity Team, "Introducing Google Antigravity SDK", Google Antigravity Blog, 2026-05-19. antigravity.google/...cing-google-antigravity-sdk. Accessed 2026-07-24.
22. ^Google Antigravity Team, "Google Antigravity for enterprises", Google Antigravity Blog, 2026-05-19. antigravity.google/...antigravity-for-enterprises. Accessed 2026-07-24.
23. ^Google, "Antigravity Agent", Gemini API documentation. ai.google.dev/...antigravity-agent. Accessed 2026-07-24.
24. ^Google, "Custom agents", Gemini API documentation. ai.google.dev/...custom-agents. Accessed 2026-07-24.
25. ^Kyle Wiggers, "Google launches Antigravity 2.0 with an updated desktop app and CLI tool at I/O 2026", TechCrunch, 2026-05-19. techcrunch.com/...ktop-app-and-cli-tool-at-io-2026 Accessed 2026-07-24.
26. ^Google, "Agent settings", Google Antigravity documentation. antigravity.google/...agent-settings. Accessed 2026-07-24.
27. ^Google, "CLI permissions", Google Antigravity documentation. antigravity.google/...cli-permissions. Accessed 2026-07-24.