Antigravity CLI: Gemini CLI Alternative for Agent-first Development

Antigravity CLI is Google’s new CLI AI coding agent for working with Google’s new Antigravity agent. It enables you to start, monitor, and control AI coding agents from the command line while using the same core agent harness as Antigravity 2.0.

The CLI is especially important for Gemini CLI users. On May 19, 2026, Google announced that Antigravity CLI is available and that Gemini CLI will stop serving requests for free individual users and Google AI Pro or Ultra users on June 18, 2026. Enterprise customers and users with supported paid API key access have a separate path, but most personal Gemini CLI users should plan to migrate.

Antigravity CLI is a part of Google’s new agent-first development ecosystem, which includes Antigravity 2.0 for a desktop agent experience and Antigravity CLI for terminal-first workflows. It is built for developers who work inside repositories, remote SSH sessions, and local terminals where a fast keyboard-driven interface is more practical than a full visual workspace.

With Antigravity CLI, you can ask agents to inspect a codebase, edit multiple files, run terminal tools, manage background tasks, and delegate work to subagents. It also supports familiar developer extension points such as skills, hooks, MCP servers, plugins, permission controls, and slash commands.

Antigravity CLI Docs

Gemini CLI Migration Timeline

May 19, 2026: Google announced the transition from Gemini CLI to Antigravity CLI.
Now available: Antigravity CLI is available as the new terminal interface for Antigravity agents.
June 18, 2026: Gemini CLI and Gemini Code Assist IDE extensions will stop serving requests for free individual users and Google AI Pro or Ultra users.
Enterprise access: Organizations using Gemini Code Assist Standard or Enterprise, Gemini Code Assist for GitHub through Google Cloud, or supported paid API key access are not affected in the same way.

Antigravity CLI does not have one-to-one feature parity with Gemini CLI at launch. The most important Gemini CLI concepts are carried forward, including Agent Skills, Hooks, Subagents, and Extensions, with extensions moving into the Antigravity plugin model.

Features

Terminal-first agent workflow: Work with Antigravity agents from your existing command-line environment.
Shared Antigravity agent harness: Use the same core agent system as Antigravity 2.0, with shared settings and future harness improvements across both surfaces.
Subagents: Delegate background work to concurrent agent sessions and monitor them from the /agents panel.
Asynchronous workflows: Run larger coding, research, or refactoring tasks in the background without tying up the main terminal conversation.
Slash commands: Use commands such as /resume, /rewind, /permissions, /model, /skills, /mcp, /tasks, and /logout.
Plugins, skills, hooks, and MCP: Add custom workflows, rules, MCP servers, hooks, and subagents through the Antigravity CLI plugin system.
Configurable permissions: Adjust agent autonomy, command approvals, and workspace access through CLI settings and permission controls.
Terminal sandbox support: Enable sandbox behavior for local shell commands when you want stricter containment for agent-run terminal actions.
Remote session support: Use browser-based authentication flows that work on local machines and remote SSH environments.

Use Cases

Explore a codebase: Open a repository and ask the agent to explain architecture, locate important files, summarize dependencies, or identify risky areas before you edit.
Implement coding tasks: Ask the agent to draft a fix, update tests, refactor a module, or prepare a migration plan from inside the project folder.
Run parallel work: Use subagents for background research, build checks, test runs, or file inspections while you keep the main conversation focused.
Manage terminal-heavy workflows: Use it on local projects, remote machines, and SSH sessions.

How to Install Antigravity CLI

Install Antigravity CLI and then use Google’s sign-in flow when the CLI asks you to authenticate.

macOS and Linux:

curl -fsSL https://antigravity.google/cli/install.sh | bash

Windows PowerShell:

irm https://antigravity.google/cli/install.ps1 | iex

Windows CMD:

curl -fsSL https://antigravity.google/cli/install.cmd -o install.cmd && install.cmd && del install.cmd

On a local machine, the CLI can open a browser-based Google Sign-In flow. In an SSH session, it can print an authorization URL for you to open in a local browser, then ask you to paste the authorization code back into the terminal.

Basic Workflow

Install Antigravity CLI.
Sign in with your Google account or connect the required Google Cloud project for enterprise use.
Open a terminal in the project folder you want the agent to work on.
Start the Antigravity CLI session and describe the task in natural language.
Review tool requests, file edits, and terminal commands before approving actions.
Use /tasks and /agents to monitor background work, and use /resume when you need to continue a previous session.

Example prompts:

“Explain this repository’s architecture and identify the main entry points.”
“Find the failing tests, propose a fix, and wait for approval before editing files.”
“Create a migration plan for moving this project to the latest framework version.”
“Use subagents to inspect the API layer, database layer, and test suite in parallel.”

Antigravity CLI vs. Gemini CLI

Gemini CLI proved that the terminal can be a strong interface for AI-assisted development. Antigravity CLI keeps the terminal workflow but moves it into Google’s most powerful Antigravity platform.

The biggest change is architecture. Gemini CLI was a standalone terminal agent. Antigravity CLI connects to the same agent harness used by Antigravity 2.0, so the CLI and desktop app can share core settings, agent behavior, and future improvements.

The second major change is multi-agent work. Antigravity CLI is designed for subagents, background tasks, and asynchronous execution. This matters for larger refactors, research tasks, and test-heavy work where a single foreground conversation can become slow.

Pros

Good fit for terminal users: It keeps the AI agent workflow inside the command line.
Better path for former Gemini CLI users: Google is directing Gemini CLI users to Antigravity CLI, so this is the most important migration target.
Shared platform with Antigravity 2.0.
Useful for larger tasks.
Strong customization model: Plugins, skills, hooks, MCP servers, settings, and keybindings give advanced users room to adapt the CLI to their workflow.
Permission controls: Review modes, command approvals, and sandbox options help reduce the risk of unwanted file or terminal actions.

Cons

No full Gemini CLI parity at launch: Some Gemini CLI features may not transfer one-to-one immediately.
Migration work may be required: Skills, MCP servers, extensions, and settings should be reviewed after import.
Google account dependency: Most users need to sign in through Google’s authentication flow.
Internet connection required: Agent work depends on Google’s cloud-hosted models and services.
New permission model to learn: Users who relied on Gemini CLI habits should spend time checking Antigravity CLI permissions, sandbox settings, and plugin behavior.

Related Resources

Antigravity CLI Product Page: Official overview of the terminal-first Antigravity CLI experience.
Antigravity Download Page: Official installers for Antigravity 2.0, Antigravity CLI, and related tools.
Antigravity CLI Docs: Documentation for CLI concepts, setup, migration, and usage.
CLI Features: Reference for plugins, sandboxing, slash commands, subagents, and permissions.
Gemini CLI Migration Announcement: Google’s official announcement about the transition from Gemini CLI to Antigravity CLI.
7 Best CLI AI Coding Agents: Compare Antigravity CLI with other terminal AI coding agents.

More Free AI Tools from Google

Code Wiki: Google’s AI-Powered Solution for Code Understanding.
NotebookLM: AI Research Assistant for Students and Professionals.
Pomelli: Google’s Free AI Tool for SMB Marketing Campaigns
Mixboard: Free AI Moodboard Tool from Google Labs.
Google Stitch: Free AI Vibe Design Tool with Agent and Web Export
Producer.ai: Free AI Studio-Quality Music Generator by Google.
Google Antigravity: Agent-First Dev Platform from Google.

FAQs

Q: Is Antigravity CLI replacing Gemini CLI?
A: Yes. Google is directing Gemini CLI users to Antigravity CLI. Gemini CLI and Gemini Code Assist IDE extensions will stop serving requests for free individual users and Google AI Pro or Ultra users on June 18, 2026.

Q: Can enterprise users still use Gemini CLI?
A: Organizations using Gemini Code Assist Standard or Enterprise, Gemini Code Assist for GitHub through Google Cloud, or supported paid API key access are not affected in the same way.

Q: Does Antigravity CLI have every Gemini CLI feature?
A: Not at launch. There will not be one-to-one feature parity immediately, but key Gemini CLI concepts such as Agent Skills, Hooks, Subagents, and Extensions are part of the migration path.

Q: What is the difference between Antigravity CLI and Antigravity 2.0?
A: Antigravity CLI is the terminal interface. Antigravity 2.0 is the desktop application for visual orchestration, project management, and broader agent workflows. They share the same agent harness and common settings.

Q: Can Antigravity CLI use subagents?
A: Yes. Antigravity CLI supports subagents for parallel background work. You can monitor them through the /agents panel and use CLI controls to approve actions.

Q: Does Antigravity CLI support MCP?
A: Yes. The CLI includes MCP management through slash commands and the plugin system. Existing Gemini CLI MCP setups should still be reviewed after migration.

Q: Is Antigravity CLI only for coding?
A: It is built for developer workflows, but it can also help with codebase research, file operations, documentation work, test runs, project analysis, and terminal-based automation.

Q: Is it safe to let Antigravity CLI edit files and run commands?
A: You should review permissions before using it on important projects. Antigravity CLI includes approval controls and sandbox settings, but you should still inspect proposed file edits and terminal commands before approving them.