From Single-Agent IDE to Antigravity 2.0 Platform
Antigravity began as an AI-assisted coding environment, but Google is now pitching it as an agent-first development platform designed for multi-agent AI development. At Google I/O, executives framed Antigravity 2.0 as a central hub for building and managing teams of autonomous agents rather than a simple code editor. The new standalone desktop app functions like a command center: one agent can generate code, another produce brand assets, while a third plans architecture, all running in parallel. This shift is reinforced by a broader suite of Google AI coding tools that pull editor-based workflows, terminal execution, and agent-building into a single branded stack. With this move, Antigravity 2.0 platform is less about an IDE and more about autonomous agent orchestration, signaling Google’s ambition to standardize how developers move from an idea to a production-ready, agent-powered application.
New CLI, SDK, and Managed Agents for Multi-Agent AI Development
To support its agent-centric vision, Google introduced an expanded toolchain around Antigravity 2.0. The new Antigravity CLI consolidates and supersedes the earlier Gemini CLI, bringing core capabilities like skills, hooks, subagents, and plugins into a single command line experience tailored for multi-agent AI development. Alongside it, the Antigravity SDK exposes the same agent harness used internally at Google, co-optimized for Gemini 3.5 Flash, which the company says outperforms Gemini 3.1 Pro on most benchmarks and runs significantly faster than rival frontier models. In the cloud, Managed Agents in the Gemini API allow developers to spin up full-featured agents in isolated Linux environments that persist state and files across calls, enabling longer-running coding and automation workflows. Together, these pieces turn Antigravity 2.0 into a multi-surface platform spanning desktop, terminal, and API-driven autonomous agent orchestration rather than a single, self-contained coding tool.
CodeMender and Automated Security Patching as a New Pillar
Beyond general-purpose coding and orchestration, Google is positioning Antigravity 2.0 as a backbone for secure, automated software maintenance. The company introduced CodeMender, a system focused on automated security patching that plugs into the same Antigravity ecosystem. While the details are still emerging, Google frames CodeMender as part of a broader push toward production-ready, enterprise-grade workflows where agents can monitor codebases, apply patches, and coordinate with other agents handling testing or deployment. Demonstrations of Antigravity’s underlying capabilities highlight this scale: Google showcased the platform running 93 subagents simultaneously to build a complete operating system in 12 hours, processing billions of tokens across thousands of model requests before booting Doom on stage. CodeMender slots into this vision as a specialized agent capability, suggesting that future Google AI coding tools will treat security and lifecycle management as first-class, autonomously orchestrated tasks rather than afterthoughts.
A Chaotic Rollout That Broke Thousands of Developer Setups
The technical expansion came at a steep usability cost. On the day Antigravity 2.0 launched, Google pushed an automatic update that effectively gutted many existing development environments. Developers reported opening Antigravity to find terminals, file explorers, and editing tools missing, replaced by a chat-centric interface optimized for agents. Source control, local terminals, and remote connections moved into a separate download, Antigravity IDE, while Antigravity 2.0 itself focused on agent work. The abrupt split into three downloads and the default shift toward an agent-only experience triggered immediate backlash on forums and social media. Some users described active projects becoming unusable overnight and criticized the move as an “agent-first pivot” that prioritized Google’s platform ambitions over existing workflows. For many, Antigravity 2.0’s promise of advanced autonomous agent orchestration was overshadowed by the disruption of everyday coding tasks.
Deprecating Gemini CLI and Locking Enterprise Access Behind API Keys
Google’s consolidation strategy also means a forced migration path. The company is deprecating the open source Gemini CLI in favor of the more closed Antigravity CLI, a move that has frustrated developers who relied on the older tool’s flexibility. After June 18, Gemini CLI and Gemini Code Assist IDE extensions will stop serving most non-enterprise users, including those on free tiers, while new installations of Gemini Code Assist for GitHub will be blocked shortly afterward. Enterprise customers with existing Standard or Enterprise licenses, or those accessing tools through their cloud accounts, retain access via paid Gemini and Gemini Enterprise Agent Platform API keys. Everyone else is nudged toward Antigravity CLI, which Google admits will not have one-to-one feature parity at launch. The result is a platform that clearly targets enterprise development workflows and agent-friendly infrastructure, but raises questions about accessibility and openness for the broader developer community.
