Google shipped Antigravity 2.0 on May 19 as a five-surface agent platform, not just a desktop app refresh, introducing a command-line tool, Python SDK, hosted agent infrastructure, and enterprise tier alongside a rebuilt desktop application. The launch represents the most complete competitor to Claude Code and Cursor that Google has ever released, yet it also triggered a mandatory migration that wiped terminals, file explorers, keybindings, and project history for thousands of developers overnight.

What Actually Shipped With Antigravity 2.0?

The confusion around Antigravity 2.0 stems from the fact that Google released five distinct products simultaneously under one name, each serving different developer workflows. Understanding what shipped matters because a developer who only wants a terminal tool can skip the desktop app entirely, while teams building internal tooling can access hosted agent infrastructure through a single API call.

• Desktop App: A rebuilt multi-agent orchestration hub with a visual Agent Manager dashboard that lets developers spawn multiple agents running in parallel and approve or reject actions before they execute.
• CLI Tool (agy): A Go-based, terminal-native command-line interface running the same agent runtime as the desktop, installable with a single command on macOS, Linux, and Windows.
• Python SDK (google.antigravity): A library for building custom agents on Google's infrastructure with support for MCP (Model Context Protocol) server connections, allowing integration with the same tool ecosystem as Claude Code or Cursor.
• Managed Agents API: Serverless agent hosting inside the Gemini API with per-run billing instead of per-token pricing, eliminating the need to manage containers or configure scaling.
• Gemini Enterprise Agent Platform: An enterprise-tier offering with service-level agreements and managed deployment for organizations requiring guaranteed uptime and support.

The lowest-friction entry point is the CLI tool, which requires only a curl command to install and a Google account to authenticate. Once set up, developers gain access to the full agent runtime, including multi-step reasoning, multi-file editing, MCP tool connections, and conversation history resumption.

Why the Multi-Agent Visualization Sets Antigravity Apart?

The desktop app's defining feature is its visual multi-agent orchestration interface, which no other consumer-tier agent tool currently offers. The Agent Manager dashboard enables what Google calls "Review-Driven Development," where the main agent breaks a task into subtasks, sub-agents work those concurrently, and the system surfaces consolidated output when complete. Critically, the agent moves autonomously but pauses for human approval before significant changes take effect.

Independent testing from XDA Developers across all three major agent tools found that Antigravity has the best speed and multi-agent visualization currently available. This distinction matters for teams whose workflows involve parallel task execution and human oversight, though whether it justifies switching from Claude Code or Cursor depends on specific use cases.

How to Get Started With Antigravity 2.0

• Start With the CLI: Install the agy command-line tool first, as it represents the lowest-risk surface and runs the full agent runtime without requiring the desktop app that caused most launch chaos.
• Wait on Desktop Deployment: Give the desktop app several patch cycles before committing live projects to it, since the forced migration caused significant data loss for early adopters.
• Monitor Managed Agents API: Watch the Managed Agents API for teams wanting hosted agentic infrastructure, but wait until Google publishes the compute-unit pricing table so cost models become transparent and predictable.

The Pricing Advantage and the Billing Problem?

On cost, Antigravity 2.0 wins decisively. Annual costs for a team of 10 developers run approximately $2,400 compared to $18,000 for Claude Code and $4,800 for Cursor. However, this pricing advantage comes with a significant caveat: Antigravity 2.0 uses compute-based quota instead of daily prompt limits, which is theoretically more accurate, but Google has not published a conversion table explaining how much a given agent run will cost before developers start it.

The /usage command only refreshes after quitting and reloading the application, making real-time cost tracking impossible. Multiple users reported seven-day lockouts after less than 30 minutes of use, indicating the quota system lacks transparency and safeguards. This is not a minor user experience oversight; it represents a material problem for any team running automated workflows that could unexpectedly exhaust their quota.

The Launch Execution Problem That Undermines Trust?

The forced update on May 19 did not go cleanly. For a significant number of users, Antigravity 2.0 wiped terminals, file explorers, keybindings, and project history without warning. Active work became inaccessible, and the migration was mandatory with no rollback option. While TechCrunch covered the launch announcement, developer forum threads documented the damage, revealing a pattern of insufficient attention to the developer trust equation.

This mirrors a recognizable Google pattern: technically impressive product launch paired with insufficient consideration for how the changes affect existing users. The five-surface architecture, visual multi-agent orchestration, and MCP-compatible SDK represent genuine progress in agent platform design, yet the execution failure creates legitimate hesitation among developers evaluating whether to adopt the tool.

For teams already productive on Claude Code or Cursor, the multi-agent orchestration is good enough to warrant serious evaluation if parallel task execution maps to a real workflow problem. Whether it justifies switching entirely remains an open question, and for most teams, the honest answer is not yet.