Google expands Antigravity into agent management platform

Google is expanding Antigravity beyond its original role as a coding environment and positioning it as a platform for building and managing teams of autonomous AI agents. The company announced the shift ahead of Google I/O 2026, along with new desktop, command-line, SDK, and cloud offerings tied to the platform.

Koray Kavukcuoglu, CTO of Google DeepMind and Chief AI Architect at Google, said in a press briefing this week that Google is “expanding Antigravity beyond a coding environment and turning it into a platform to develop and manage teams of autonomous AI agents.” In a blog post, Varun Mohan, director of software engineering at Google DeepMind, and Logan Kilpatrick, a member of the technical staff there, described Antigravity as an “agent-first development platform” meant to take an idea and turn it into a production-ready app.

The centerpiece of the update is Antigravity 2.0, a standalone desktop application that acts as a control center for agent orchestration. Google said developers will be able to coordinate multiple agents at once instead of assigning a single agent to a single task. The company gave examples such as one agent generating code, another producing brand assets, and a third handling product architecture planning.

For developers who work in the terminal, Google is also launching the Antigravity CLI, a lightweight interface for starting agents without a graphical app. Google said existing Gemini CLI users are being encouraged to migrate to the new tool.

Google is also releasing the Antigravity SDK, which gives developers programmatic access to the same agent harness used in Google’s internal products. The SDK is co-optimized for Gemini models, especially Gemini 3.5 Flash, which Google said was developed alongside Antigravity 2.0 as the platform’s main workhorse model. The company said Gemini 3.5 Flash outperforms Gemini 3.1 Pro on almost all benchmarks and runs four times faster than competing frontier models.

The cloud side of the platform is getting an upgrade through Managed Agents in the Gemini API. Google said a single API call can now start a full agent that reasons, uses tools, and executes code inside a persistent, isolated Linux environment. State and files persist across follow-up calls, which means sessions can continue without being rebuilt from scratch.

Google is also tying Antigravity more tightly to its enterprise stack. The platform now integrates with the Gemini Enterprise Agent Platform, allowing Google Cloud customers to connect directly to their Cloud projects. The company is also introducing a new Google AI Ultra subscription tier priced at $100 per month, which it says offers five times higher usage limits in Antigravity than the Pro plan.

Google pointed to several enterprise customers as early validation. Nikunj Shanti, CTO of AirAsia Next, said more than half of the company’s production-ready code is now generated through Antigravity agentic workflows. Faruk Muratovic, US AI and Engineering Strategy and Services Leader at Deloitte, said the platform can run governed, autonomous engineering workflows that meet Deloitte’s enterprise security standards. Vikas Agarwal, PwC Advisory CTIO, said the shift moves “past simple AI code completion to true agent orchestration.”

WPP also said it has integrated Antigravity into WPP Open, its agentic marketing platform. Callum Anderson, Head of Engineering at WPP, said the system has streamlined workflows, automated repetitive tasks, and helped engineering teams deliver higher-quality solutions faster.

Google also used the announcement to expand CodeMender, an AI code security agent developed by Google DeepMind. Thomas Kurian, CEO of Google Cloud, said CodeMender can autonomously identify vulnerabilities, recommend fixes, test them in a secure environment, and apply patches across dependent systems, with developer approval.

Kavukcuoglu said CodeMender shows how AI can “actually patch code, not just find exploits.” Google is adding the tool to its Agent Platform as part of its AI Threat Defense offering, and a select group of security experts is being invited to test a CodeMender API now.

The company also announced an AI Content Detection API for Agent Platform, which is rolling out today and can identify AI-generated content from Google and third-party models. Taken together, the announcements show Google pushing Antigravity toward persistent agent systems while adding the security and governance tools needed to manage what those systems produce.