Google Antigravity Release: Agent-First IDE on Gemini 3 Pro
November 19, 2025 — Today, Google Antigravity, an agent-first integrated development environment (IDE) that directly gives AI agents direct access over editor, terminal and browser applications in response to Google Antigravity
declaring that its approach will transform the creation of software. The platform, which was launched in open preview together with the new Gemini 3 family, claims to allow developers to coordinate a number of autonomous agents that write, test and verify code and generate auditable artifacts like task lists, screenshots and browser videos.
There are two user experiences that Antigravity is based on. The Editor view is similar and feels like the traditional IDE which has a sidebar containing agents which can be requested to execute a task or a test; the Manager View is more of a mission control, and it lets developers spawn, monitor and coordinate numerous agents across multiple workspaces. Google Antigravity is placing this two-view design as a transition between working code and more advanced task-oriented processes.
The core of the Google Antigravity is the newest multimodal model, the Gemini 3 Pro, by Google Antigravity
is said to reason, code, and use the tools better than the previous models. The announcement by Google made it clear that Antigravity is not a single agent assistant but an environment in which multiple specialized agents can be semi-autonomous, can learn through feedback and leave a clear audit trail through Artifacts a feature of the system designed to provide trust and transparency in the behavior of agents.
The direct responses of the developer and industry communities were moderate. Protagonists sang the vision of the vibe coding and agent coordinating -the ability that could dramatically hasten daily engineering endeavors, automated testing and end to end repeat. Opponents and skeptics also expressed concern that the power of the product casts doubt on reliability, job loss (especially of junior developers) and early bugs among preview users. A few of the earliest users have already reported problems with setting up as well as instability of the UI in the initial day of public preview.
In an indication of practical attitude toward model choice and interoperability, Google Antigravity claims that in addition to Gemini 3 Pro, other models available, including Sonnet 4.5 by Anthropic and open-source GPT-OSS are compatible with Antigravity. To allow developers around the globe to test the workflows of agents and offer feedback in the early release, the company is releasing the Antigravity as a free public preview on Windows, macOS and Linux.
Other than the productivity enhancement, Antigravity raises critical policy and industry issues. New guardrails will be required to ensure agents do not leak secrets or engage in unsafe web activities, engineering managers will have to determine when to trust the output of agent versus human review, and companies will require policies to report agent decisions to ensure compliance and debugging. According to Venture Beat and other sources, Google Antigravity focused on Artifacts specifically because it needed to address these governance issues.
The arrival of antigravity comes at a time when vendors of large language models are competing to create viable developer products that achieve the transformation of AI potential into actual business value. By integrating agentic capabilities into the IDE as well as visualizing them via artifact trails and a manager interface, Google Antigravity is claiming the next generation of developer tooling one, in which artificial intelligence acts as a collaborative team member instead of a line-by-line autocomplete. Analysts believe that the technology has potential to transform the employment process, work life and the software delivery economics in case it comes out as advertised.
To developers and the teams that want to test Antigravity, Antigravity site and the Gemini developer pages on Google Antigravity have downloads and instructions on a getting-started guide of the public preview. Everything should think iterative early adopters need to be prepared to make the experience iterative: there will be speedy innovation, active loops, and the standard tradeoffs of the preview period between new things and early stability problems.
Read Also This: Bloomberg Journalist Catherine Lucey Draws Attention After Trump called her Piggy