Google Antigravity makes the IDE a control plane for agent coding

Google launched Antigravity as an agent development platform on top of Gemini 3. It’s not just an autocomplete layer, it’s an IDE where agents can plan, execute, and interpret complex software tasks across editor, terminal, and browser interfaces. Antigravity launches with Gemini 3 on November 18, 2025, as part of Google’s push for agent-centric development tools.

What exactly is antigravity?

Google describes Antigravity as a new agent development platform with a familiar AI-powered IDE at its core. The goal is to evolve IDEs into an agent-first future, allowing agents to autonomously plan and execute end-to-end software tasks through browser control and asynchronous interaction modes.

In practice, Antigravity looks and behaves like a modern AI editor, but treats agents like first-class workers. Agents can interrupt tasks, coordinate with other agents, edit files, run commands, and drive browsers. Developers operate at the task level, while the system manages low-level tool interactions.

Essentially, Antigravity is an Electron application based on Visual Studio Code. It requires a Google account login and is available as a free public preview for macOS, Linux, and Windows.

Models, pricing and runtime environment

Antigravity exposes multiple base models within the same agent framework. In the current preview, agents can use Gemini 3, Anthropic Claude Sonnet 4.5, and OpenAI GPT OSS models. This provides developers with the ability to make model choices within an IDE, rather than tying them to a single vendor.

For individual users, Antigravity is available free of charge. Google describes Gemini 3 Pro usage as being subject to a generous rate limit of refresh every 5 hours, noting that only a small percentage of premium users are expected to hit those limits.

Editor view and manager view

Antigravity introduces 2 main working modes to match different neural models. Documentation and reports consistently describe them as editor view and manager view.

Editor view is the default view. It looks like a standard IDE, with an agent in the side panel. The agent can read and edit files, suggest changes inline, and use terminals and browsers when needed.

Manager views lift the abstraction from a single file to multiple agents and workspaces. This is where you can coordinate multiple agent runs rather than editing code line by line.

Artifacts instead of raw tool logs

A key design element of Antigravity is the Artifact system. Rather than just exposing raw tool call logs, agents generate human-readable artifacts summarizing what they are doing and why.

Artifacts are structured objects that can include task lists, implementation plans, walkthrough documents, screenshots, and browser logs. They represent task-level work, rather than API call-level work, and are intended to be easier for developers to verify rather than dense traces of model operations.

Google is positioning this as a response to trust issues in the current proxy framework. Many tools either show every internal step, which can overwhelm the user, or hide everything and only show the final code differences. Antigravity tries to sit somewhere in the middle by presenting task-level artifacts plus enough validation signals so that developers can audit what the agent is doing.

Four design principles and feedback channels

Antigravity is explicitly built around 4 principles: Trust, Autonomy, Feedback and Self-Improvement.

Trust is handled through artifacts and verification steps. Autonomy comes from giving agents access to multiple interfaces, editors, terminals, and browsers so they can run more complex workflows without constant prompting. Feedback is achieved through comments on artifacts, and self-improvement is tied to the agent learning from past work and reusing successful procedures.

Antigravity allows developers to directly comment on specific artifacts, including text and screenshots. Agents can incorporate this feedback into their ongoing work without abandoning the current run. This allows you to correct partial misunderstandings without restarting the entire mission.

The platform also exposes a knowledge feature where agents can retain code snippets or sequences of steps from earlier tasks. Over time, this becomes a reusable internal playbook that agents can consult, rather than rediscovering the same strategies for each new project.

Main points

1. Antigravity is an agent-first development platform that turns the IDE into a control plane, with agents operating over the editor, terminal, and browser interfaces instead of narrow inline helpers.
2. The system is a fork of Visual Studio Code that runs as a free public preview on Windows, macOS, and Linux, with generous Gemini 3 Pro rate limits and the option to use Claude Sonnet 4.5 and GPT OSS.
3. Antigravity exposes 2 main modes: Editor view (for manual coding using the agent sidebar) and Manager view (as a task control interface for asynchronously orchestrating multiple agents and workspaces).
4. Agents emit artifacts, task lists, implementation plans, screenshots, browser logs, and more that act as verifiable evidence of work instead of raw tool logs and enable asynchronous audit workflows.
5. Feedback and self-improvement are built-in, developers can attach Google Docs-style comments to artifacts on various surfaces, and the agent will incorporate this feedback and learn from the development knowledge base without restarting the task.

Google Antigravity is a pragmatic step towards agent development. It anchors Gemini 3 Pro in a real IDE workflow, exposing editor views and manager views for supervisory agents, and enforcing task-level visibility through Artifacts. The four principles of trust, autonomy, feedback, and self-improvement are based on verifiable output and lasting knowledge, not opaque traces. Overall, Antigravity treats the IDE as a controlled environment for autonomous agents rather than a chat window with code manipulation.

Check Full technical details here. Please feel free to check out our GitHub page for tutorials, code, and notebooks. In addition, welcome to follow us twitter And don’t forget to join our 100k+ ML SubReddit and subscribe our newsletter. wait! Are you using Telegram? Now you can also join us via telegram.