From Coding Sandbox to Multi-Agent AI Command Center
Antigravity 2.0 marks a decisive shift in Google’s AI strategy: instead of a single-agent coding helper, it is now pitched as an “agent-first development platform” for building and managing teams of autonomous AI agents. The new standalone desktop app functions as a command center where multiple agents can run in parallel, coordinating tasks like code generation, brand asset creation, and product architecture planning in one place. Projects can span multiple repositories and include scheduled routines, pushing Antigravity beyond a simple AI coding tool into a broader multi-agent AI development environment. Under the hood, the suite is optimized for Gemini 3.5 Flash, which Google positions as its primary workhorse model for Antigravity. The goal is clear: turn what was once an AI-assisted IDE into a central hub for enterprise AI agents and complex, long-running workflows.
New Tools: Antigravity CLI, SDK, Managed Agents and CodeMender
To support this new role, Google is rolling out a full stack of Google AI coding tools around Antigravity 2.0. The Antigravity CLI consolidates and eventually replaces the Gemini CLI, carrying over core features like skills, hooks, subagents, and plugins while emphasizing multi-agent orchestration from the terminal. A new Antigravity SDK exposes the same agent harness Google uses internally, letting developers programmatically define behaviors and integrate agents into their own services. In the cloud, Managed Agents in the Gemini API can spin up full agents that reason, call tools, and execute code in persistent Linux environments, with state and files surviving across calls. On the security side, Google is introducing CodeMender, a specialized agent focused on security patching and remediation. Collectively, these Antigravity 2.0 features reposition the product from a single editor to an extensible enterprise AI agent platform.
A Sudden Rollout That Broke Developer Workflows
The transition has not been smooth. Google pushed an automatic Antigravity update that effectively gutted the existing code editor, leaving many developers staring at a chat-centric interface where terminals, file explorers, and editing tools once lived. Antigravity was split into separate downloads, and the default install became the new agent-focused Antigravity 2.0. Developers who relied on the original all-in-one environment now need to install Antigravity IDE separately to regain source control, local terminals, and remote connections. Forums and social channels quickly filled with complaints as active projects were disrupted within hours of launch. For teams in the middle of deadlines, the forced pivot from a stable coding environment to an experimental multi-agent platform felt less like an upgrade and more like unplanned downtime, raising concerns about Google’s deployment readiness and change-management discipline.
Gemini CLI Wind-Down and a New Access Gate
Antigravity 2.0’s rise coincides with the sunset of the more open Gemini CLI, and this is where access friction intensifies. As of June 18, Gemini CLI and Gemini Code Assist IDE extensions will stop serving requests for most individual and free users, including those on Google AI Pro and Ultra tiers. New installations of Gemini Code Assist for GitHub will be blocked, and existing access will be phased out soon after. Enterprises, however, are largely exempt: organizations with Gemini Code Assist Standard or Enterprise licenses, or Gemini Enterprise Agent Platform API keys, will retain Gemini CLI access. For everyone else, the path forward runs through the closed-source Antigravity CLI and paid Gemini API access. This shift raises equity questions about who can meaningfully use multi-agent AI development tools and whether core capabilities are being nudged behind enterprise paywalls.
Enterprise Ambitions vs. Developer Trust
Strategically, Antigravity 2.0 aligns tightly with Google’s enterprise AI agents narrative. Integration with the Gemini Enterprise Agent Platform allows corporate teams to connect Antigravity workflows to internal systems, data sources, and governance layers. Demos of large-scale autonomous development, with dozens of subagents coordinating over billions of tokens, underscore Google’s ambition to make Antigravity a default operating layer for complex software and automation projects. Yet the way this shift has been executed may slow adoption. Developers are weighing the appeal of powerful multi-agent orchestration, persistent cloud agents, and CodeMender’s security patching against the loss of open tools, sudden environment changes, and incomplete feature parity in the new CLI. Until Google provides clearer guarantees around stability, openness, and long-term Gemini API access models, Antigravity 2.0 will be seen as both a bold platform bet and a cautionary tale in developer platform transitions.