The Shift to Agent-First Development

Google's Antigravity 2.0 marks a strategic pivot away from traditional IDE-integrated AI assistance toward a standalone platform designed for multi-agent orchestration. The platform treats agents as persistent, parallelized automation pipelines rather than single-turn chat interfaces. Key features include dynamic subagents for parallel task execution, scheduled background tasks, and native voice command support, allowing developers to manage complex workflows that extend beyond simple code generation.

A Unified Developer Harness

To support this shift, Google has introduced a cohesive ecosystem that shares a common agent harness:

  • Antigravity CLI: A terminal-based surface that replaces the Gemini CLI, maintaining support for Agent Skills, Hooks, and plugins.
  • Antigravity SDK: Enables developers to define custom agent behaviors and host them on their own infrastructure, facilitating the embedding of agentic capabilities into internal tools or third-party products.
  • Managed Agents (Gemini API): Provides infrastructure-level isolation for agent execution. Each interaction creates a persistent Linux environment where files and state remain intact across sessions, eliminating the need to reinitialize context. These are powered by Gemini 3.5 Flash, which Google claims outperforms Gemini 3.1 Pro while running 4x faster.
  • Enterprise Support: The Gemini Enterprise Agent Platform allows organizations to connect Antigravity directly to Google Cloud projects, providing a deployment path for secure, enterprise-grade agent operations.

Expanded Ecosystem and Integration

Google is broadening the developer surface area to include mobile and workspace-native capabilities. A new mobile app allows for prototyping on the go, with a one-click 'Export to Antigravity' feature to transition work to desktop environments. Additionally, new Workspace integrations allow agents to programmatically interact with Docs, Sheets, and Calendar, while native Android and Google Play Console support enables developers to build and test applications entirely within the AI Studio environment.