Antigravity CLI
A multi-agent terminal AI programming tool introduced by Google as the successor to Gemini CLI. It supports a primary agent that coordinates multiple sub-agents to handle complex software tasks in parallel, sharing the same agent core with the Antigravity 2.0 desktop application.
Core Facts
Timeline (last 90 days)
The Antigravity CLI command is agy, and the first run requires OAuth authentication via a browser.
All 16 runs passed the checks, and all code reviews accurately found the 3 planted bugs with no false positives.
Claude Code cost $0.169 per passing task, while Antigravity CLI cost $0.544, making the former 3.2 times cheaper; total time was 318 seconds vs 3712 seconds, a gap of nearly 12 times.
Antigravity CLI migration points: GEMINI.md and AGENTS.md are read as usual; workspace skills migrate from .gemini/skills/ to .agents/skills/; MCP servers migrate from ~/.gemini/settings.json to ~/.gemini/config/mcp_config.json; agy plugin import gemini can convert old extensions into plugins.
In the code review, Claude Code not only pointed out the ReDoS risk but also measured the time taken to quantify the impact, whereas Antigravity CLI also found the flaw but did not verify it.
In the bug-fix task, Claude Code completed both runs within 30 seconds, with 4 to 6 tool calls; Antigravity CLI took 526 seconds and 672 seconds, with 50 to 64 tool calls.
In the refactoring task, agy's code quality was slightly better: the two agy runs additionally added TYPE_CHECKING imports and __dir__(), at the cost of 10 minutes and 92 tool calls, whereas Claude Code used only 1 minute and 7 calls.
Two additional single runs were done to separate the variables: dropping agy's model to gemini-3.8-flash-low still took 332 seconds and 67 tool calls; switching to claude-opus-5-5-high, agy used only 51 seconds and 7 tool calls, close to Claude Code's 29 seconds.
On 2026-10-13, SideHub conducted a comparative test of Claude Code and Antigravity CLI (Google's terminal agent) using four types of tasks, each task run twice, all in headless mode.
The actual test used Claude Code 2.1.296 (claude-opus-5-5) and Antigravity CLI 1.3.4 (gemini-3.8-flash-high), with default models and effort, running on a Google AI Pro account.
6 more timeline events
All Facts (19)
Gemini CLI is hosted on GitHub and uses the Apache 2.0 open-source license.
65%UnverifiedCLI形态的AI编程工具天然适合集成到CI/CD流水线（如GitHub Actions、Jenkins）中，实现代码审查、自动修复、文档生成等自动化任务
85%UnverifiedThe Antigravity CLI command is agy, and the first run requires OAuth authentication via a browser.
50%UnverifiedIn the bug-fix task, Claude Code completed both runs within 30 seconds, with 4 to 6 tool calls; Antigravity CLI took 526 seconds and 672 seconds, with 50 to 64 tool calls.
50%UnverifiedAll 16 runs passed the checks, and all code reviews accurately found the 3 planted bugs with no false positives.
50%UnverifiedClaude Code cost $0.169 per passing task, while Antigravity CLI cost $0.544, making the former 3.2 times cheaper; total time was 318 seconds vs 3712 seconds, a gap of nearly 12 times.
50%UnverifiedAntigravity CLI migration points: GEMINI.md and AGENTS.md are read as usual; workspace skills migrate from .gemini/skills/ to .agents/skills/; MCP servers migrate from ~/.gemini/settings.json to ~/.gemini/config/mcp_config.json; agy plugin import gemini can convert old extensions into plugins.
50%UnverifiedIn the code review, Claude Code not only pointed out the ReDoS risk but also measured the time taken to quantify the impact, whereas Antigravity CLI also found the flaw but did not verify it.
50%UnverifiedIn the refactoring task, agy's code quality was slightly better: the two agy runs additionally added TYPE_CHECKING imports and __dir__(), at the cost of 10 minutes and 92 tool calls, whereas Claude Code used only 1 minute and 7 calls.
50%UnverifiedTwo additional single runs were done to separate the variables: dropping agy's model to gemini-3.8-flash-low still took 332 seconds and 67 tool calls; switching to claude-opus-5-5-high, agy used only 51 seconds and 7 tool calls, close to Claude Code's 29 seconds.
50%UnverifiedThe actual test used Claude Code 2.1.296 (claude-opus-5-5) and Antigravity CLI 1.3.4 (gemini-3.8-flash-high), with default models and effort, running on a Google AI Pro account.
50%UnverifiedAntigravity CLI adopts a multi-agent parallel architecture, where the main agent is responsible for decomposing tasks and distributing them to multiple SubAgents for parallel background execution.
50%UnverifiedAntigravity CLI and the Antigravity 2.0 desktop client share the same agent core, with the terminal and desktop interconnected.
50%UnverifiedAntigravity CLI provides multiple extension entry points including MCP, Skills, Plugins, and Hooks.
50%UnverifiedIndividual free users, Google AI Pro, and Google AI Ultra users have now primarily migrated to the Antigravity CLI.
50%UnverifiedThe tools officially listed by Clippo as supported include Claude Code, OpenCode, Codex, CommandCode, Antigravity CLI, Kiro, and others.
50%UnverifiedAntigravity CLI is Google's official code assistant, through which developers can complete code generation and testing in VS Code using natural language instructions.
50%Unverified使用普通 Google 账号登录 Antigravity 可免费调用 Gemini 3 Pro、Gemini 3.5 Flash、Claude Opus 4.6 Thinking 和 Sonnet 4.6 等模型
50%UnverifiedOn 2026-10-13, SideHub conducted a comparative test of Claude Code and Antigravity CLI (Google's terminal agent) using four types of tasks, each task run twice, all in headless mode.
50%