Google has launched Antigravity as an agentic improvement platform that sits on prime of Gemini 3. It isn’t solely an autocomplete layer, it’s an IDE the place brokers plan, execute, and clarify complicated software program duties throughout editor, terminal, and browser surfaces. Antigravity was launched on November 18, 2025, alongside Gemini 3 as a part of Google’s push towards agent centric developer instruments.
What Antigravity Really is?
Antigravity is described by Google as a brand new agentic improvement platform with a well-known AI powered IDE at its core. The objective is to evolve the IDE towards an agent first future, with browser management and asynchronous interplay patterns that permit brokers autonomously plan and execute finish to finish software program duties.
In apply, Antigravity seems and behaves like a contemporary AI editor however treats brokers as top quality staff. Brokers can break duties, coordinate with different brokers, edit recordsdata, run instructions, and drive a browser. The developer operates at a activity degree, whereas the system manages the low degree device interactions.
Underneath the hood, Antigravity is an Electron utility primarily based on Visible Studio Code. It requires a Google account register and ships as a free public preview for macOS, Linux, and Home windows.
Fashions, Pricing, And Runtime Setting
Antigravity exposes a number of basis fashions inside the identical agent framework. Within the present preview, brokers can use Gemini 3, Anthropic Claude Sonnet 4.5, and OpenAI GPT OSS fashions. This offers builders mannequin optionality inside one IDE as a substitute of binding them to a single vendor.
For particular person customers, Antigravity is accessible at no cost. Google describes the Gemini 3 Professional utilization as topic to beneficiant price limits that refresh each 5 hours, and notes that solely a small fraction of energy customers are anticipated to hit them.
Editor View And Supervisor View
Antigravity introduces 2 major work modes that match completely different neural fashions. Documentation and protection persistently describe these as Editor view and Supervisor view.
Editor view is the default. It seems like a normal IDE with an agent within the facet panel. The agent can learn and edit recordsdata, recommend adjustments inline, and use the terminal and browser when wanted.
Supervisor view lifts the abstraction from single recordsdata to a number of brokers and workspaces. That is the place the place you coordinate a number of agent runs somewhat than enhancing code line by line.
Artifacts, Not Uncooked Device Logs
A key design factor in Antigravity is the Artifact system. As a substitute of exposing solely uncooked device name logs, brokers produce human readable artifacts that summarize what they’re doing and why.
Artifacts are structured objects that may embody activity lists, implementation plans, walkthrough paperwork, screenshots, and browser recordings. They signify work at a activity degree somewhat than at an API name degree and are designed to be simpler for builders to confirm than dense traces of mannequin actions.
Google positions this as a response to a belief drawback in present agent frameworks. Many instruments both present each inner step, which overwhelms customers, or cover every thing and solely present the ultimate code diff. Antigravity tries to sit down within the center by surfacing activity degree artifacts plus sufficient verification indicators so {that a} developer can audit what the agent did.
4 Design Tenets And Suggestions Channels
Antigravity is explicitly constructed round 4 tenets, belief, autonomy, suggestions, and self enchancment.
Belief is dealt with by way of artifacts and verification steps. Autonomy comes from giving brokers entry to a number of surfaces, editor, terminal, and browser, to allow them to run extra complicated workflows with out fixed prompts. Suggestions is enabled by way of feedback on artifacts, and self enchancment is tied to brokers studying from previous work and reusing profitable procedures.
Antigravity permits builders to remark immediately on particular artifacts, together with textual content and screenshots. Brokers can incorporate this suggestions into their ongoing work with out discarding the present run. This allows you to appropriate a partial misunderstanding with out restarting the entire activity.
The platform additionally exposes a data characteristic the place brokers can retain snippets of code or sequences of steps from earlier duties. Over time, this turns into a reusable inner playbook that brokers can question, somewhat than rediscovering the identical methods for every new venture.
Key Takeaways
- Antigravity is an agent first improvement platform that turns the IDE right into a management airplane the place brokers function throughout editor, terminal and browser surfaces, as a substitute of a slender inline assistant.
- The system is a Visible Studio Code fork that runs as a free public preview on Home windows, macOS and Linux, with beneficiant Gemini 3 Professional price limits and optionally available use of Claude Sonnet 4.5 and GPT OSS.
- Antigravity exposes 2 major modes, Editor view for arms on coding with an agent sidebar and Supervisor view as a mission management interface to orchestrate a number of brokers and workspaces asynchronously.
- Brokers emit Artifacts, activity lists, implementation plans, screenshots, browser recordings and extra, which act as verifiable proof of labor as a substitute of uncooked device logs and allow asynchronous evaluation workflows.
- Suggestions and self enchancment are in-built, builders can connect Google Docs fashion feedback to artifacts throughout surfaces, and brokers incorporate this suggestions and be taught from a improvement data base with out restarting duties.
Google Antigravity is a practical step towards agentic improvement. It anchors Gemini 3 Professional inside an actual IDE workflow, exposes Editor view and Supervisor view for supervising brokers, and enforces activity degree visibility by way of Artifacts. The 4 tenets, belief, autonomy, suggestions, self enchancment, are grounded in verifiable outputs and chronic data somewhat than opaque traces. Total, Antigravity treats the IDE as a ruled atmosphere for autonomous brokers, not a chat window with code actions.
Try the FULL TECHNICAL DETAILS right here. Be at liberty to take a look at our GitHub Web page for Tutorials, Codes and Notebooks. Additionally, be at liberty to comply with us on Twitter and don’t neglect to hitch our 100k+ ML SubReddit and Subscribe to our Publication. Wait! are you on telegram? now you may be part of us on telegram as properly.
Michal Sutter is an information science skilled with a Grasp of Science in Knowledge Science from the College of Padova. With a strong basis in statistical evaluation, machine studying, and information engineering, Michal excels at reworking complicated datasets into actionable insights.