Google Launches Antigravity 2.0: Unified Desktop Application, SDK, and Multi-Agent CLI Debuted at I/O 2026
At the Google I/O 2026 developer conference, the company officially debuted Antigravity 2.0, a major upgrade to its agentic software engineering ecosystem. Originally launched last year as a direct competitor to popular agentic development suites like Cursor, Antigravity 2.0 expands the tool from a localized helper into a comprehensive desktop application, command-line interface (CLI), and highly customizable software development kit (SDK).
Unified Desktop Orchestration and Parallel Agent Execution
The center of the release is the new Antigravity desktop application, which enables developers to orchestrate multiple independent agents and run complex engineering operations concurrently. Users can design sophisticated subagent workflows and configure background tasks to execute on automated schedules.
Designed for deep ecosystem synergy, the desktop app integrates natively with Google AI Studio, Android development suites, and Firebase backends. Under the hood, this entire platform is powered by Google's new Gemini 3.5 Flash model, which was co-developed using the Antigravity ecosystem itself to accelerate research and engineering pipelines.
Terminals, Runtimes, and Extensible SDKs
For developers who prefer command-line workflows, Google has introduced the native Antigravity CLI. This new terminal interface allows engineers to spin up, configure, and monitor automated agents without leaving the terminal. Google has announced that the older Gemini CLI is officially deprecated, urging active developers to migrate their existing project setups to the new Antigravity command line tool.
Additionally, the release introduces the Antigravity SDK, enabling enterprise developers and cloud customers to construct custom agentic architectures atop Google's core code-generation engines. To support this deployment:
- Google Cloud customers can now directly connect their cloud workspaces to Antigravity.
- Pre-configured agent templates are being rolled out inside Google AI Studio for rapid onboarding.
- A native export tool has been added to AI Studio, allowing developers to cleanly package cloud-based agentic projects and transition them into local environments.
Consumer Product Integrations
Google is also embedding Antigravity's code synthesis capabilities directly into consumer-facing products. Most notably, Google Search is receiving an upgrade where complex development queries will generate highly interactive, custom user interfaces in real time. This allows users to build, run, and explore micro-applications directly within their search results. Furthermore, natural voice commanding is being integrated directly into the Antigravity desktop suite, mirroring recent productivity updates across Google Workspace tools like Docs, Keep, and Gmail.
Restructured Premium Subscription Tiers
Alongside the software release, Google has restructured its subscription tiers to accommodate developers requiring large-scale computing throughput. This includes the launch of a new mid-range tier and a price reduction on the top-tier enterprise plan.
| Subscription Tier | Monthly Pricing | Antigravity Usage Limits | Key Features & Target Audience |
|---|---|---|---|
| AI Pro | $20 / month | Baseline (1x) | Standard developer tool limits, single-agent workflows, and standard model access. |
| AI Ultra (New) | $100 / month | 5x Pro Limits | Advanced concurrent multi-agent orchestration, custom subagent scheduling, and priority compute queues. |
| AI Ultra Max (Updated) | $200 / month $250 | 20x Pro Limits | High-throughput enterprise execution, custom workspace integrations, and zero-data-retention options. |
This tiered pricing structure aligns Google with recent commercial structures adopted by other major frontier labs, including OpenAI and Anthropic, who have rolled out similar $100 and $200 tiers to satisfy varying developer compute demands.
Developers calculating the hardware overhead and memory profiles required to run equivalent open-weights agentic models locally can consult our interactive AI VRAM Calculator to design optimal local hosting configurations.