Google accelerates agent-native software program growth with expanded Antigravity platform – SiliconANGLE

Google Cloud is enhancing its “agent-first” coding platform for builders with the launch of Antigravity 2.0, a brand new standalone desktop utility that permits a full “agent-optimized” person expertise.

It comes alongside the debut of Managed Brokers within the Gemini utility programming interface and native Android vibe coding in Google AI Studio, which had been additionally introduced right this moment. The updates arrived throughout Google I/O right this moment, constructing upon November’s launch of the unique Antigravity platform that’s designed to help synthetic intelligence automation at the next, task-oriented stage.

With Antigravity, AI evolves from an assistant to a real collaborator that’s in a position to work by itself initiative and with minimal supervision. The platform takes the looks of a typical coding setting, nevertheless it’s peppered with AI brokers that may entry the whole lot from the editor to the terminal and even an built-in net browser. This implies they’ve the entire instruments required to autonomously plan and execute advanced programming duties.

The unique Antigravity was powered primarily by Gemini 3 Professional, later upgraded to Gemini 3.1 Professional, and from right this moment, customers will now have entry to Google’s most superior Gemini 3.5 Flash model. In a weblog submit, Google DeepMind Director of Software program Engineering Varun Mohan famous that 3.5 Flash outperforms Gemini 3.1 Professional on nearly all coding benchmarks whereas working 4 occasions quicker than some other frontier mannequin, giving builders the “high-speed engine wanted for real-world agentic workflows.”

Antigravity 2.0 is supposed to behave as a central house for agentic interactions, enabling builders to orchestrate a number of AI brokers that may carry out duties in parallel. It additionally introduces “dynamic subagents,” that are for parallelized workflows, scheduled duties for background automation and new integrations with Google AI Studio, Android and the Firebase cell growth platform.

Builders might choose a conventional terminal-style expertise when the embark on agentic growth, and so they’re additionally catered to with Antigravity CLI. It’s a brand new and light-weight but high-velocity product floor the place brokers could be created immediately with out the necessity for a graphical person interface. Lastly, Google additionally revealed the brand new Antigravity SDK, which supplies programmatic entry to the identical brokers present in Antigravity, enabling customers to outline customized agent behaviors and host them on third-party infrastructure.

Managed Brokers and extra

Sticking with agentic growth, Google stated it’s updating the Gemini API for integrating AI capabilities into software program functions with the introduction of Managed Agents. Mohan defined that Managed Brokers made it attainable to spin up new brokers that may purpose, use third-party instruments and execute code in remoted Linux environments with a single API name.

These Managed Brokers are all powered by the Antigravity agent harness and Gemini 3.5 Flash, and could be accessed by the Interactions API and thru Google AI Studio. As a result of they use Antigravity’s harness, it implies that person’s customized constructed brokers shall be based mostly on the identical know-how and infrastructure as Google’s personal brokers. Builders can get began constructing customized brokers utilizing Google’s new agent templates, and also will be capable of lengthen the capabilities of their Managed Brokers with personalized directions and abilities utilizing markdown recordsdata.

One other manner for builders to entry the newest capabilities of Gemini 3.5 Flash is thru the Google AI Studio, which is a web-based prototyping setting that makes it easy to experiment with AI, refine prompts and generate API keys.

Google AI Studio is now powered by Antigravity’s coding brokers too, and the corporate stated it’s increasing the methods builders can use the platform. As an example, there’s a brand new cell utility opening for registration now that sacrifices deeper performance for comfort, permitting builders to “seize an thought on the go and have a working prototype prepared by the point you get to your desk,” Mohan stated.

Moreover, Google AI Studio is now built-in with Google Workspace, which suggests AI brokers will be capable of name related Workspace APIs and embed them instantly into apps constructed utilizing the platform. Ought to builders determine they want extra superior coding performance, they will now export their whole Google AI Studio tasks to the locally-hosted Antigravity platform with a single click on.

Lastly, there’s a brand new integration that makes it attainable to construct native Android functions with a single immediate. As soon as glad, builders will be capable of publish their apps instantly within the take a look at monitor inside the Google Play Console.

The final main dev-focused replace pertained to the Gemini Enterprise Agent Platform, previously generally known as Vertex AI, which is the corporate’s full-stack growth platform for constructing enterprise-grade AI brokers and autonomous functions that may plan and execute advanced, multistep duties.

The principle replace right here is all about securing agent-generated code. It’s called CodeMender, a newly built-in AI code safety agent that was first created by Google DeepMind researchers.

By leveraging Agent Platform’s capabilities and essentially the most superior Gemini fashions, it autonomously searches for and identifies any vulnerabilities in newly created code, together with that generated by different brokers. As soon as a bug is recognized, CodeMender will suggest a exact repair, apply it on command after which securely take a look at the code to make sure there are not any extra loopholes or weaknesses.

Picture: Google Cloud