From Coding Helper to Multi-Agent AI Orchestration Platform
At Google I/O 2026, Antigravity 2.0 was repositioned from a coding-centric tool into a full Antigravity 2.0 platform for multi-agent AI orchestration. Google now pitches Antigravity as an “agent-first development platform” where developers can move from idea to production by coordinating teams of autonomous agents rather than chatting with a single assistant. The new standalone desktop app acts as a command center, grouping work into projects and running several agents in parallel across multiple repositories and tasks. Under the hood, Antigravity 2.0 runs on the Gemini 3.5 Flash model, which Google claims outperforms Gemini 3.1 Pro on coding and “agentic” benchmarks while powering more complex workflows. This puts Antigravity directly up against rival AI agent development tools like Claude Code and Codex, signaling that Google wants Antigravity to be a foundational layer for how developers build, orchestrate, and ship AI-powered software.
New Command Center, CLI, SDK and Managed Agents in Gemini API
The platform expansion arrives in three main forms designed to cover graphical desktops, terminals, and programmatic workflows. The Antigravity 2.0 desktop app is the hub: it can orchestrate many agents simultaneously, from code generation to product planning to asset creation, and structure work into longer, multi-step chains instead of one-off prompts. For terminal-focused users, Google introduced the Antigravity CLI, which inherits core Gemini CLI capabilities such as skills, hooks, subagents, and plugins, while adding better support for multi-agent environments. An Antigravity SDK exposes the same agent harness Google uses internally, co-optimized with Gemini 3.5 Flash. In the cloud, Managed Agents in the Gemini API let developers spin up stateful agents inside isolated Linux environments with a single call, persisting state and files across sessions. Together, these pieces turn Antigravity into a broader stack for multi-agent AI orchestration that stretches from local development to enterprise cloud deployment.
CodeMender, Security Automation and Enterprise AI Access Strategy
Alongside the orchestration upgrades, Google is pushing Antigravity deeper into security and enterprise AI access. A new feature branded as CodeMender focuses on automated security patching, using the same multi-agent harness to scan codebases, propose fixes, and apply patches as background tasks. This builds on Antigravity’s ability to schedule agents and run them asynchronously, potentially turning routine hardening and dependency updates into continuous, AI-driven maintenance. On the business side, Antigravity now plugs into the Gemini Enterprise Agent Platform so organizations can tie agents directly to their existing cloud projects and governance patterns. Google has also introduced new AI subscription tiers, including a USD 100 (approx. RM460) per month Google AI Ultra plan that offers higher usage limits inside Antigravity than the existing Pro plan. These moves underscore an enterprise AI access strategy where governance, capacity, and performance are explicitly packaged for paying corporate customers.
A Chaotic Rollout That Broke Developer Workflows Overnight
The ambitious vision came with a jarring execution. On May 19, Google pushed an automatic update that transformed the existing Antigravity install into the new agent-focused interface, effectively gutting the familiar code editor. Developers opened Antigravity to find terminals, file explorers, and editing tools gone, replaced by a chat-style agent console. Source control, local terminals, and remote connections disappeared from the default experience and were moved into a separate Antigravity IDE download that users had to discover and install manually. Forums and social channels quickly filled with complaints, with some developers calling the agent-first pivot a step backward that “reeks of non-technical people shipping code to production.” What was supposed to be a seamless upgrade into a more powerful multi-agent environment instead broke thousands of active projects, forcing teams to scramble just to restore their previous development setup before they could experiment with the new features.
Gemini CLI Sunset, Paywalls and the Future of Developer Experience
The disruption is compounded by a strategic shift in how Google exposes its AI agent development tools. As part of the consolidation, the open-source Gemini CLI is being deprecated in favor of the new, closed-source Antigravity CLI. On June 18, Gemini CLI and Gemini Code Assist IDE extensions will stop serving requests for Google AI Pro and Ultra users, as well as those relying on free individual access. Gemini Code Assist for GitHub will block new installs the same day and then cease handling requests. Only enterprise customers on Standard or Enterprise licenses and those using paid Gemini or Gemini Enterprise Agent Platform API keys will retain Gemini CLI access. Everyone else is pushed toward Antigravity CLI, which does not yet offer full feature parity. This shift toward enterprise AI access and tighter control raises pointed questions about Google’s platform strategy and whether prioritizing governance and monetization is coming at the expense of a stable, developer-friendly ecosystem.
