From Coding Companion to Multi-Agent AI Command Center
With the Antigravity 2.0 release, Google is repositioning its AI coding environment as a multi-agent AI platform for building and managing teams of autonomous agents. Instead of a single assistant helping with code snippets, Antigravity now acts as a central hub where multiple agents can run in parallel—one generating code, another designing assets, and a third planning product architecture. At Google I/O, DeepMind leaders described Antigravity as an “agent-first development platform” that can take an idea to a production-ready app. The new desktop app groups work into projects, spans multiple repositories, and supports scheduled background tasks, turning everyday development into a coordinated workflow stack rather than a chat-driven IDE. Backed by the Gemini 3.5 Flash model, which Google claims outperforms earlier Gemini versions and rival frontier models, Antigravity 2.0 is clearly designed as a foundation for autonomous agent orchestration rather than a simple coding assistant.
CodeMender and Managed Agents Push Deeper Automation
Antigravity 2.0 is not just a new interface; it is a broader ecosystem for autonomous agent orchestration. A key addition is CodeMender, an automated security patching system meant to plug directly into developer workflows and continuously fix vulnerabilities without manual intervention. Alongside the new SDK, developers gain programmatic access to the same agent harness Google uses internally, co-optimized for Gemini 3.5 Flash and geared toward long-running, multi-agent jobs. Managed Agents in the Gemini API extend this model into persistent Linux environments in the cloud, where an agent can reason, call tools, execute code, and maintain state across many requests. In a live demo, Google showcased Antigravity coordinating 93 subagents across billions of tokens to build a full operating system from scratch, underscoring how far the platform has moved from autocomplete-style assistance to end-to-end software creation and maintenance.
A Rapid Rollout That Broke Thousands of Developer Setups
The ambitious shift came at a steep cost for many existing users. On May 19, Google pushed an automatic Antigravity update that effectively gutted the familiar development environment. Developers opened Antigravity to find terminals, file explorers, source control integration, and local editing tools missing—replaced by an agent-first chat interface. Within hours, community forums and social channels filled with reports that active projects had become unusable, with some engineers calling the rollout a “massive step backward.” The change stems from Google splitting Antigravity into three installers, with Antigravity 2.0 dedicated to multi-agent work while the traditional editor now lives in a separate Antigravity IDE download. For developers who had come to rely on Antigravity as a full-featured IDE, the unannounced removal of core tools felt less like an upgrade and more like having their primary workspace unexpectedly dismantled.
Gemini CLI Deprecation Tightens Access to Google AI Coding Tools
In parallel, Google is consolidating its command line tooling, deprecating the open-source Gemini CLI in favor of the new Antigravity CLI. Officially, the shift is framed as unifying Google AI coding tools under the Antigravity brand and improving support for multi-agent environments. In practice, most individual users lose access. After June 18, Gemini CLI and related IDE extensions will stop serving requests for free tiers and many Pro users, with new installations blocked for popular integrations. Continued access requires enterprise-grade Gemini Code Assist licenses or paid Gemini and Gemini Enterprise Agent Platform API keys. While Antigravity CLI preserves core features such as skills, hooks, subagents, and plugins, Google warns there will not be 1:1 feature parity at launch. For many developers, the move signals a pivot from open, broadly accessible tooling toward a more closed, enterprise-gated multi-agent AI platform.
Balancing Visionary Multi-Agent AI with Developer Trust
Taken together, Antigravity 2.0, CodeMender, the new SDK, and the Antigravity CLI paint a picture of Google betting heavily on multi-agent AI platforms to reshape software development. The system’s ability to coordinate dozens of agents, persist long-running tasks, and automate security patching points to a future where teams orchestrate autonomous systems as much as they write code themselves. Yet the rollout also highlights the risks of aggressive, top-down product strategy. Automatic updates that strip away core IDE features, coupled with the retirement of an open-source CLI in favor of a closed, credential-gated tool, have eroded confidence among many of the very developers Google hopes will adopt autonomous agent orchestration. The ultimate success of Antigravity 2.0 may depend less on raw capability and more on how quickly Google can rebuild trust with a user base shaken by disruptive change.