Gemini CLI vs Claude Code: differences, access and the Antigravity CLI switch

What is the difference between Gemini CLI and Claude Code?

Gemini CLI is Google's open source terminal agent for Gemini models; Claude Code is Anthropic's agentic coding tool for Claude models, available in the terminal, IDEs, a desktop app and the browser. Both read your code, edit files and run commands from a prompt, and both support context files, Agent Skills, subagents, hooks and MCP servers.

The bigger difference in 2026 is access. On 18 June 2026 Google replaced Gemini CLI with Antigravity CLI for unpaid and Google One users. So for many people the real comparison is now Claude Code against Antigravity CLI, with Gemini CLI left for enterprise licenses and paid API keys. This page covers all three, from the official docs as of 24 October 2026.

What changed for Gemini CLI on 18 June 2026?

The Gemini CLI repository now opens with this notice: "Unpaid tier and Google One users: Gemini CLI was replaced by Antigravity CLI on June 18th, 2026."

Google's announcement on the Google Developers Blog, dated 19 May 2026, gives the details:

• On 18 June 2026, Gemini CLI and the Gemini Code Assist IDE extensions stopped serving requests for Google AI Pro and Ultra subscribers and for people using Gemini Code Assist for individuals free of charge.
• Customers with a Gemini Code Assist Standard or Enterprise license keep access, and the post says their "access remains unchanged."
• Gemini CLI "will remain accessible via paid Gemini and Gemini Enterprise Agent Platform API keys."
• Antigravity CLI keeps "Agent Skills, Hooks, Subagents, and Extensions (now as Antigravity plugins)," though Google says "there won't be 1:1 feature parity right out of the gate."

The repository also says: "Gemini CLI has not been optimized for Gemini 4 Argon. To use Gemini 4 Argon, please use Antigravity." In short, Gemini CLI still works and is still open source, but Google's newest coding work is happening in Antigravity CLI.

How does Claude Code work?

Claude Code is "an agentic coding tool that reads your codebase, edits files, runs commands, and integrates with your development tools." You install it, open a project and run claude:

# macOS, Linux, WSL
curl -fsSL https://claude.ai/install.sh | bash

cd your-project
claude

The same engine runs in the terminal, VS Code, JetBrains IDEs, a desktop app and the web at claude.ai/code. For scripts and CI, claude -p "query" runs one task and exits, and there are GitHub Actions and GitLab CI/CD integrations.

Claude Code is configured with a CLAUDE.md file, skills in .claude/skills/, subagents in .claude/agents/, hooks, MCP servers and plugins. It needs a paid Claude plan, a Claude Console account or a supported third-party provider. As of October 2026, Claude's pricing page says "Claude Code is included in all paid plans," with Pro at $20 a month billed monthly and Max from $100 a month. The Claude Code pricing guide goes deeper.

How does Gemini CLI work?

Gemini CLI is an open source terminal agent distributed through npm. The README lists several ways to install it, and you start it with gemini:

npx @google/gemini-cli          # run without installing
npm install -g @google/gemini-cli
brew install gemini-cli         # macOS and Linux

You authenticate with a Google account (for a paid Code Assist license, set your Google Cloud project), a Gemini API key, or Vertex AI. The Gemini CLI docs cover the rest of the feature set:

• Context files: GEMINI.md in ~/.gemini/ and in your workspace and parent directories, with @file.md imports.
• Agent Skills: in .gemini/skills/ or .agents/skills/ for a workspace, and ~/.gemini/skills/ for your user.
• Subagents: Markdown files in .gemini/agents/ or ~/.gemini/agents/, plus built-in agents such as Codebase Investigator.
• Hooks, extensions and MCP servers, with MCP configured in ~/.gemini/settings.json.
• Headless mode for scripts, such as gemini -p "Explain the architecture of this codebase", with text, JSON or streaming JSON output, plus checkpointing and sandboxing.

It is licensed under Apache 2.0, so you can read, fork and modify the code.

How do Gemini CLI and Claude Code compare side by side?

How do they handle project instructions and skills?

This is where the tools overlap most, and where you can avoid writing everything twice.

Instructions. Claude Code reads CLAUDE.md. Its memory docs say it can also "read AGENTS.md as your project instructions," by default only when you have no CLAUDE.md in your working directory or above it. Gemini CLI reads GEMINI.md, and its context file docs show how to load other names with the context.fileName setting:

{
  "context": {
    "fileName": ["AGENTS.md", "CONTEXT.md", "GEMINI.md"]
  }
}

So one AGENTS.md at the repo root can serve both tools. The CLAUDE.md vs AGENTS.md guide walks through the setup, including the @AGENTS.md import for repos that keep a CLAUDE.md.

Skills. Claude Code's skills docs say its skills "follow the Agent Skills open standard," and Gemini CLI's skills page says its skills are "Based on the Agent Skills open standard." A SKILL.md folder with a name and description works in both. Claude Code adds its own fields, such as disable-model-invocation and context: fork, which other tools may ignore. See the Claude Code skills guide for the details.

What does the Antigravity CLI switch mean for you?

If you used Gemini CLI on a free or Google One account, Gemini CLI no longer serves your requests. You have three options: move to Antigravity CLI, keep Gemini CLI with a paid Gemini API key, or switch to another agent such as Claude Code.

Moving to Antigravity CLI is mostly mechanical. Google's migration guide says that when you first run agy in a folder with old configuration, it offers to convert it. The main changes:

• GEMINI.md and AGENTS.md: no changes. Antigravity CLI reads both in the active directory and ~/.gemini/GEMINI.md globally.
• Workspace skills: move .gemini/skills/ to .agents/skills/ by hand. Global skills move from ~/.gemini/skills/ to ~/.gemini/antigravity-cli/skills/.
• Extensions: run agy plugin import gemini to convert them to plugins. Old custom commands become skills.
• MCP servers: they move from ~/.gemini/settings.json to ~/.gemini/config/mcp_config.json globally or .agents/mcp_config.json per workspace, and remote servers use serverUrl instead of url or httpUrl.

Antigravity CLI installs with curl -fsSL https://antigravity.google/cli/install.sh | bash on macOS and Linux. If you plan to try Claude Code instead, the same AGENTS.md and SKILL.md folders carry over with little or no change.

Which should you pick?

Pick based on your account, your model preference and whether open source matters to you. Neither tool is right for everyone.

Claude Code tends to fit when:

• You already pay for a Claude plan, or want Anthropic's Claude models.
• You want one tool across terminal, IDE, desktop and browser, with the same CLAUDE.md, settings and MCP servers in each.
• You want CI integrations such as GitHub Actions out of the box.

Gemini CLI tends to fit when:

• Your company has a Gemini Code Assist Standard or Enterprise license, or you use paid Gemini API keys.
• You want an Apache 2.0 codebase you can audit or fork.
• You have an existing Gemini CLI setup with extensions you rely on.

Antigravity CLI is Google's path forward if you are on a free or consumer Google plan, or you want Gemini 4 Argon, which the Gemini CLI README says needs Antigravity. Feature parity is still catching up, so check the migration guide for anything your workflow depends on.

Whatever you choose, keep shared instructions in AGENTS.md and reusable workflows as Agent Skills. Both travel between these tools, so switching later costs you little.

Keep your prompts and skills portable

Tools change faster than the instructions you build up, as the Gemini CLI switch shows. Kipt keeps your prompts and Agent Skills as Markdown in a private GitHub repo you own, and installs every skill into Claude Code or Codex with one npx skills add command. If you have a prompt you keep pasting into a terminal agent, the free SKILL.md generator turns it into a skill folder in your browser.