From coding companion to multi-agent AI orchestration hub
With Antigravity 2.0, Google is reframing its AI development tools as an “agent-first” platform rather than a coding sidekick. The new Antigravity 2.0 platform centers on a desktop command hub where multiple autonomous agent teams can work in parallel—one agent generating code, another designing brand assets, a third planning architecture. This move pushes Antigravity beyond a single IDE into a broader stack for multi-agent AI orchestration. Google pairs the desktop app with a new SDK and Managed Agents in the Gemini API so developers can programmatically spin up persistent agents in isolated Linux environments, with state and files surviving across calls. Co-optimization with Gemini 3.5 Flash, positioned as Antigravity’s primary workhorse, underlines the company’s intent: turn Antigravity into the default control plane for autonomous agent teams, not just a convenient AI code helper.
CodeMender and security-focused AI development tools
Alongside the Antigravity 2.0 platform, Google is pushing a security narrative with new AI development tools. The flagship addition is CodeMender, a feature designed to automatically analyze, propose, and apply security patches across codebases. Combined with Antigravity’s SDK, CLI, and Gemini API integration, the goal is to embed Google CodeMender security workflows directly into day-to-day development. Agents can run in the background on schedules, scan repositories, and issue or even implement fixes governed via JSON-defined hooks and policies. This shifts Antigravity from a reactive coding assistant to a proactive security and automation layer. For enterprises weighing AI development tools, the pitch is clear: a multi-agent AI orchestration platform that not only builds software but continuously hardens it. Yet buyers are already signalling they want detailed governance, deployment, and security guarantees before treating Antigravity 2.0 as a formal, long-term platform choice.
A jarring rollout that broke existing developer setups
The technical vision collided with reality when Google auto-updated existing Antigravity installs to Antigravity 2.0. Many developers opened their machines to find familiar terminals, file explorers, and code editors gone—replaced by an agent-first chat interface. Source control, local terminals, and remote connections were removed from the default install and moved into a separate Antigravity IDE download. Active projects suddenly became unusable, triggering frustration across forums and social channels and leaving some to describe the move as a non-technical ship decision. At the same time, Google split the product line into three pieces: Antigravity 2.0 for agents, Antigravity IDE for traditional editing, and Antigravity CLI for terminal workflows. In theory this clarifies roles; in practice, the abrupt changeover underscored how brittle AI-centric tooling can be when core development workflows are altered without clear migration paths.
Gemini CLI deprecation and the shift to a more closed Antigravity CLI
Google’s consolidation strategy extends to the command line. The company is deprecating the open Gemini CLI in favor of the new Antigravity CLI, which inherits core capabilities like skills, hooks, subagents, and plugins and adds better support for multi-agent environments. However, access and openness are narrowing. After June 18, Gemini CLI will stop serving most individual and free users; ongoing access is largely reserved for customers with paid Gemini or Gemini Enterprise Agent Platform API keys and certain enterprise licenses. For everyone else, it is effectively “Antigravity CLI or nothing,” and Google has already warned there will not be full feature parity at launch. That shift, combined with Antigravity’s closed-source direction, has angered developers who relied on the flexibility of Gemini CLI and raises questions about how much control Google expects to retain over the emerging AI development command line.
Competing with enterprise AI suites while testing developer trust
Underneath the turmoil, Antigravity 2.0 clearly targets the same territory as enterprise AI development suites like Claude Code: a unified environment for orchestrating autonomous agent teams across code, design, and operations. Antigravity’s project grouping, multi-repository workflows, scheduled tasks, and managed cloud agents position it as a full-stack AI development platform rather than a collection of experiments. The Gemini 3.5 Flash integration and large-scale demos—with dozens of subagents coordinating on complex builds—signal a serious platform play. But the rushed rollout, missing features, and restricted Gemini CLI access highlight a maturity gap between Google’s ambitions and developers’ expectations. Enterprise buyers will scrutinize governance, security assurances, and migration stability before committing. For now, Antigravity 2.0 looks simultaneously like a powerful multi-agent AI orchestration system and a cautionary case study in how not to ship disruptive platform changes to working developer environments.