Explainer · updated Oct 10, 2026
Gemini CLI is an open-source (Apache 2.0) AI agent from Google that brings Gemini into the terminal to query and edit codebases. The official site says unpaid-tier and Google One users were moved to Antigravity CLI on June 18, so check your access before you start.
Gemini CLI is an open-source AI agent from Google that runs in your terminal. It is released under the Apache 2.0 license, so you can read the code. You type prompts, and the agent works with your local files and directories, images or PDFs, and web pages it fetches through search. The vendor is Google, and the official site is geminicli.com, where the plans and documentation also live.
The official site lists its main uses as querying large codebases, editing files from the terminal, running shell commands, searching and fetching web content, and automating workflows through MCP. Outputs include edited files, shell command results, generated apps, and plans and todos. An account is required to use it. The docs describe it as a tool for developers who want to work from the command line instead of an editor.
You install it with `npm install -g @google/gemini-cli`, or through npx, Homebrew or MacPorts. That gives you a `gemini` command. Run it in your project folder and sign in with a Google account, an AI Studio API key, or Google Cloud credentials for Vertex AI. You can add a GEMINI.md file to give the agent project context. Optional steps like this one keep the first setup short. Project conventions can live in an AGENTS.md file, and external tools connect through MCP servers.
Built-in tools read, write and edit local files and run shell commands. Google Search grounding brings current web information into answers. The docs describe sandboxing, trusted folders, a policy engine and a .geminiignore file for controlling what the agent can touch. MCP servers and extensions add more tools, though setup is manual, and the docs also list Agent Skills and Plan Mode.
The tool itself is open source. What you pay for is access to the models. The plans page lists Google AI Pro and Google AI Ultra for individuals, the Google Developer Program for teams, and pay-as-you-go billing through an AI Studio API key or Google Cloud billing for Vertex AI. A free Google account option is listed at $0 per month. Plan details are on the Gemini CLI pricing and Google Gemini pricing pages.
The plans page shows no USD prices for the paid plans, so we cannot state a figure for any of them. Higher limits require a paid Google plan or API billing. Current free-tier limits are also unconfirmed, because the site says unpaid-tier users were moved to Antigravity CLI. For cost planning, get a quote from Google or check Google Cloud pricing.
Gemini CLI fits terminal-centric developers and people who already use Google AI, AI Studio or Vertex AI. It suits you if you want an open-source agent you can inspect, with several sign-in options and extension through MCP. Enterprises can use Vertex AI, which lists Google Cloud terms, an SLA and regional data residency options. The docs also mention enterprise configuration for organizations. For the closest rivals read what Claude Code is and what OpenAI Codex is, then Gemini CLI vs Claude Code, Gemini vs ChatGPT, Claude vs Gemini and Gemini vs GitHub Copilot; the AI coding tools guide covers the wider field.
Look elsewhere if you are on the unpaid tier or Google One, since the official site says those users were moved to Antigravity CLI on June 18. Also pause if you work with sensitive code on a free API key: the plans page says free-tier data may be used to improve products, while AI Studio pay-as-you-go offers privacy.
Our catalog lists Google Gemini, GitHub Copilot and Cursor as alternatives. Google Gemini is the broader Google assistant, while Gemini CLI is the terminal agent. GitHub Copilot and Cursor are other coding tools you can weigh against it if you prefer a different workflow or are affected by the Antigravity CLI change. We list names only where our catalog links them.
We have not run a hands-on benchmark of Gemini CLI against these tools, so the comparison rests on documented differences only. Gemini CLI's documented strengths are its open-source license, multiple authentication paths and MCP extensibility. For any switch, compare current plans and limits on each vendor's own pages, since Gemini CLI's free-tier access changed and its paid prices are not published in USD.
Install the package, open a terminal in your project folder and run `gemini`, then sign in. Start with a read-only request, such as asking for an overview of the project structure, before you ask for changes. When you request an edit, ask to see the diff before it is applied. Starting small makes it easier to judge how the agent reads your code before you let it change anything.
After the agent finishes, check the work yourself: run the tests and review the git diff. A prompt like the one below is a reasonable first task on an unfamiliar repository. For example: "Explain how authentication works in this repository, then add rate limiting to the login endpoint in src/api/login.ts, and write a test for it. Show me the diff before applying changes." Treat the result as a draft that you verify, not as finished work.
Open-source terminal AI agent from Google bringing Gemini to the command line; site notes unpaid and Google One users moved to Antigravity CLI.
Free stops at: Verify current availability
Verified geminicli.com (opens Gemini CLI website in a new tab)
Pricing US price unconfirmed
For unpaid-tier and Google One users, yes: the official site says it was replaced on June 18. Paid options remain through Google AI plans, the Developer Program, AI Studio or Vertex AI.
Not for new unpaid users. The free Google account tier was moved to Antigravity CLI on June 18, so Gemini CLI now runs through Google AI Pro or Ultra, the Google Developer Program, an AI Studio API key or Vertex AI. The plans page shows no USD prices.
Run `npm install -g @google/gemini-cli`. The site also lists npx, Homebrew and MacPorts. You need to authenticate before first use.
A Google account, an AI Studio API key, or Google Cloud credentials for Vertex AI.
Yes. The project README says it is released under Apache 2.0.
compare>
