What is Gemini CLI? Setup (Windows, Mac), is it still free, API key, MCP and moving to Antigravity CLI
You install Gemini CLI (with Node.js 20 or later, `npm install -g @google/gemini-cli`) and give it a task in a project folder; the agent reads files, changes them with your approval and runs commands. It is Google's open-source (Apache 2.0) terminal agent. Important change: on June 18, 2026, Google shut down sign-in with a personal Google account, so the old "1,000 free requests a day" figure no longer applies. You now need a paid Gemini API key, Vertex AI or an enterprise Code Assist license; Google is pointing consumer users to Antigravity CLI. Installing the tool is free. The interface is in English; there is no official source on Turkish support.
- What it does
- Google coding agent that reads and edits files in the terminal
- Sign-in
- Paid API key, Vertex AI, enterprise license
- Setup
- npm install -g @google/gemini-cli
- Where
- Terminal: Windows, Mac, Linux
- Turkish
- Interface in English; response language not documented
- Current version
- 0.64.0 (October 13, 2026)
- Free?
- Software yes; Google account sign-in has ended
What does Gemini CLI do? Can I use it without code?
!Since June 18, 2026, you can't sign in to Gemini CLI with a personal Google account (including free Code Assist, Google AI Pro and Ultra). Google is pointing consumer users to Antigravity CLI. Gemini CLI keeps working with enterprise Code Assist licenses and paid API keys. The README and quota docs on GitHub still describe the old free quota; don't rely on them.
No. Gemini CLI is a command-line tool: it runs in the terminal, the window where you give the computer commands as text. It has no cloud version or interface of its own. If you're looking for a visual editor, Google's announcement mentions the Antigravity app.
You open Gemini CLI in a project folder and write what you want in plain sentences. The tool works like an agent: instead of giving a single answer, it reads files, proposes changes and runs commands. For every step that changes a file or runs a command, it shows you first and asks for approval.
- Built-in tools: file operations, shell commands, web page fetching and Google Search (README).
- Extending it: MCP servers, extensions and agent skills.
- Headless use: running it with a single command inside a script or CI.
- Project context: a GEMINI.md file in the folder becomes the tool's standing instructions for that project.
Who it's for: developers comfortable with the terminal, and people doing website and automation work. If you have never used a terminal, consider a visual tool first.
How do you install Gemini CLI? Windows, Mac and Linux
First you need Node.js (the free software runtime the tool needs to run), version 20 or later. The docs list the recommended systems as macOS 15+, Windows 11 24H2+ and Ubuntu 20.04+; the shell can be Bash, Zsh or PowerShell. An internet connection is required.
› npm install -g @google/gemini-cli› gemini› brew install gemini-cli› sudo port install gemini-cliFor restricted environments: you install Node.js in a separate environment and install the tool inside it.
› conda create -y -n gemini_env -c conda-forge nodejs
conda activate gemini_env
npm install -g @google/gemini-cli› npx @google/gemini-cliIf Docker or Podman is installed, you can run the tool's commands in an isolated container. This is not installing the tool with Docker; it isolates the commands of a tool that is already installed. Your working folder is mounted into the container at the same path.
› gemini -s -p "analyze the code structure"- Install Node.js. In the terminal, confirm with node --version that your version is 20 or later.
- Install the tool. Pick one of the tabs above.
- Confirm it's installed. gemini --version prints the version.
- Start it and choose a sign-in method. Type gemini in the project folder. For sign-in options, see the "Is it still free?" section.
!Windows doesn't have Unix commands like chmod; if the tool tries to run one, it errors. The docs suggest using the icacls command or WSL (a Linux environment on Windows) instead.
4 GB+ RAM for light use; 16 GB+ RAM for long sessions and large projects (the docs' recommendation). The docs don't mention a graphics card; the model runs on Google's servers.
The latest stable release is 0.64.0 (October 13, 2026). There are weekly stable, weekly preview and daily nightly channels; stable is recommended for most people. To update: npm install -g @google/gemini-cli@latest
The software is free. What you pay for is model access; you don't need to run a server, but installation, updates and key management are on you.
If you want to remove it
- If you installed with npm: npm uninstall -g @google/gemini-cli
- Homebrew: brew uninstall gemini-cli. MacPorts: sudo port uninstall gemini-cli.
- If you tried it with npx, nothing stays installed; to clear the cache, delete npm's _npx folder.
- Your settings and history live in the ~/.gemini folder; delete it manually if you don't want to keep them.
Source: Gemini CLI installation, uninstall and FAQ docs; npm registry; October 15, 2026.
Gemini CLI or Antigravity CLI? Moving your old settings
On May 19, 2026, Google announced it was bringing its work together under a single product, Antigravity. The terminal part is Antigravity CLI (command: agy). Google writes that this tool keeps Gemini CLI's most important features (agent skills, hooks, subagents, extensions) but is not identical. Consumer use of Gemini CLI ended on June 18, 2026; access with an enterprise license or a paid API key continues.
› agy plugin import gemini| Topic | Gemini CLI | Antigravity CLI |
|---|---|---|
| Global skills folder | ~/.gemini/skills/ | ~/.gemini/antigravity-cli/skills/ |
| Project skills folder | .gemini/skills/ | .agents/skills/ |
| MCP settings | inside settings.json | separate mcp_config.json file |
| Remote MCP address key | url or httpUrl | serverUrl |
| GEMINI.md and AGENTS.md | Read | Read as is |
The project skills folder is not moved automatically: you have to move the contents of .gemini/skills/ to .agents/skills/ yourself. The Antigravity docs also say some custom themes and experimental visual settings may not carry over.
!Antigravity CLI's price and free quota are not covered in this guide; for current information, see the antigravity.google/pricing page.
Source: Google Developers Blog (May 19, 2026), Antigravity docs "Migrating from Gemini CLI" and "Antigravity CLI overview"; October 15, 2026.
How do you use Gemini CLI? First commands, adding MCP and extensions
- Go to the project folder. In the terminal, use cd to switch to the folder you want to work in.
- Start it. Type gemini. On first launch it asks for a sign-in method.
- Write your request. For example, have it summarize what the project does or write tests for a file.
- Approve or decline. For steps that will change a file or run a command, the tool shows the diff or the command; you decide with options such as Allow once.
| Command | What it does |
|---|---|
| gemini | Starts an interactive session |
| gemini -p "question" | One-off query; prints the answer and exits |
| gemini -r latest | Resumes the last session |
| gemini --sandbox | Runs tool commands in a sandbox |
| gemini update | Updates to the latest version |
| /stats model | Shows the session's token usage and quota information |
| /mcp list | Lists connected MCP servers |
| /help | Shows all slash commands |
MCP is the standard that lets AI tools connect to outside services (such as GitHub or a database). In Gemini CLI, you add a server with a command or with the settings file.
› gemini mcp add github npx -y @modelcontextprotocol/server-github› gemini mcp listThe official MCP guide also shows another way: writing an mcpServers block in ~/.gemini/settings.json. In the example, GitHub's MCP server runs with Docker and the access token is read from an environment variable.
› {
"mcpServers": {
"github": {
"command": "docker",
"args": ["run", "-i", "--rm", "-e", "GITHUB_PERSONAL_ACCESS_TOKEN", "ghcr.io/github/github-mcp-server:latest"],
"env": { "GITHUB_PERSONAL_ACCESS_TOKEN": "${GITHUB_PERSONAL_ACCESS_TOKEN}" }
}
}
}Restart Gemini CLI; check with /mcp list that it says "Connected". If the tools don't appear, try the /mcp reload command. Don't write the token in plain text in the settings file; have it read from an environment variable, and give the GitHub token only the permissions it needs.
An extension bundles prompts, MCP servers, custom commands, themes, hooks, subagents and agent skills into one package. It installs from a GitHub address; the official gallery is at geminicli.com/extensions/browse.
› gemini extensions install https://github.com/gemini-cli-extensions/workspace› gemini extensions list!Because an extension can carry commands and MCP servers, don't install one whose source you don't know.
- Tools that change files and run commands ask for approval by default.
- Automatic snapshots before file changes (checkpointing) can be turned on in settings; if it's on, /restore takes you back to the earlier state.
- With --sandbox, commands run in an isolated container; the trusted folders setting limits which folders the tool runs in.
With --approval-mode you choose default, auto_edit, yolo or plan. yolo runs every action without approval; don't use it unless you know what you're doing.
Source: Gemini CLI quickstart, command reference, MCP setup guide, extensions, tools and checkpointing docs; README; October 15, 2026.
Is Gemini CLI still free? Google account sign-in, API keys and enterprise licenses
The software itself is free and licensed under Apache 2.0. The real question is how you access the model: Gemini CLI doesn't carry any AI of its own; it connects to Google's models, and the cost of that access depends on the sign-in method.
| Personal Google account (free Code Assist, AI Pro, AI Ultra)No | Shut down on June 18, 2026. Sign-in with the "Login with Google" option no longer works; Google is pointing these users to Antigravity CLI. |
|---|---|
| Gemini API key (paid)Yes | Pay as you go. Google writes that Gemini CLI will remain accessible with paid API keys. Price and quota are on the Gemini API page; this guide gives no figures. |
| Gemini API key (free)Unclear | Conflicting. The quota docs say 250 requests a day and only one fast model, the README says 1,000 requests a day, and Google's announcement mentions only paid keys. Try it yourself to see whether it works. |
| Vertex AI (Google Cloud)Yes | Pay as you go, billed through Google Cloud. The quota docs mention 90 days of unbilled use for "Express mode"; whether that still applies today is not clear in the official sources. |
| Gemini Code Assist Standard / Enterprise (enterprise)Partly | Keeps working with an enterprise license. But Google announced that new Standard and Enterprise subscriptions can't be purchased after October 9, 2026, and that renewals will stop in 2027. |
| Sign-in method | Daily request limit (per the docs) |
|---|---|
| Code Assist Standard (enterprise) | 1,500 requests / user / day |
| Code Assist Enterprise (enterprise) | 2,000 requests / user / day |
| Gemini API key (paid) | Varies by plan |
| Vertex AI | Varies by account |
!The Google account rows still in the quota docs (1,000, 1,500, 2,000 requests) and the README's 60 requests per minute no longer apply. This guide gives no free-limit figures.
- If you want to try it on your own: get an API key from Google AI Studio, put it in the GEMINI_API_KEY environment variable, start the gemini command and choose the Use Gemini API key option. Check its quota in AI Studio.
- If you work with a company or school account: you need a Google Cloud project and a Gemini Code Assist license; you set the GOOGLE_CLOUD_PROJECT environment variable. Ask your administrator.
- If you have Cloud infrastructure: Vertex AI; it needs GOOGLE_CLOUD_PROJECT and GOOGLE_CLOUD_LOCATION.
- For scripts or CI: the docs recommend an API key or Vertex AI.
› export GEMINI_API_KEY="YOUR_KEY"› $env:GEMINI_API_KEY="YOUR_KEY"!An API key is like a password: don't share it with anyone or commit it to a Git repository. The docs suggest storing the key in the .gemini/.env file in the project folder or in the operating system's keychain.
To see your remaining quota, use the /stats model command in the tool. Prices are in US dollars and set on Google's billing pages. Payment methods, taxes and cancellation steps are not clear in the official sources; check the relevant Google account or the Cloud console.
Source: Google Developers Blog announcement (May 19, 2026); Gemini Code Assist deprecation pages (updated October 9, 2026); the geminicli.com top banner; quota, authentication and FAQ docs; October 15, 2026.
Gemini CLI license, my data and Turkish
Apache 2.0 open-source license. The terms for using and modifying it are in the license text; Google's service terms also apply. This is not legal advice.
The tool runs on your computer, but your prompts and the files you have it read are sent to Google's model; the docs offer no local model option. Think before having it read sensitive files.
- Free Gemini API key: the Gemini API "Unpaid Services" terms.
- Paid Gemini API key: the Gemini API "Paid Services" terms.
- Vertex AI and Code Assist Standard/Enterprise: Google Cloud terms.
- Google's plans page says data on the free API tier may be used to improve products; it recommends the paid API for privacy.
On by default. It collects tool names, durations and configuration; the docs say prompts, responses, file contents and personal information are not collected. To turn it off, set privacy.usageStatisticsEnabled to false in settings.json.
Using Gemini CLI's sign-in session with other software (such as Claude Code, OpenClaw or OpenCode) counts as a violation of the terms and can lead to your account being suspended. If you want to use Gemini with another tool, the official route is an API key.
The GitHub security advisories list was empty on October 15, 2026; that doesn't mean there are no vulnerabilities. The tool can run commands: read the approval prompts, and use the sandbox in folders and projects you don't trust. To report a vulnerability, Google points to g.co/vulnz.
| Interface languageUnclear | There is no language option in the official docs; all the docs are in English. |
|---|---|
| Turkish responsesUnclear | The official sources don't mention Turkish support. Write in Turkish and check the result yourself. |
| Access from TurkeyUnclear | The installation docs refer to supported locations; the official docs don't make clear whether Turkey is on that list, so check before installing. |
| Local-currency pricingNo | There is no official Turkey pricing page; prices are in US dollars. |
Gemini CLI errors: sign-in, 429, command not found
| Symptom | Fix |
|---|---|
| Google account sign-in doesn't work | Since June 18, 2026, sign-in is closed for personal accounts. Use a paid API key or Vertex AI, or move to Antigravity CLI. |
| named user ... Standard edition error | If GOOGLE_CLOUD_PROJECT or GOOGLE_CLOUD_PROJECT_ID is set, the enterprise subscription check is forced. If you are an individual, remove these variables; if you are on an enterprise plan, ask your administrator for a license. |
| API error: 429 Resource exhausted | The request limit was exceeded. Check usage in AI Studio or the Cloud dashboard, space out your requests or request a quota increase. |
| gemini command not found | Check whether npm's global directory is on your PATH; reinstall with npm install -g @google/gemini-cli@latest. |
| Certificate error on a corporate network | Try NODE_USE_SYSTEM_CA=1; if that doesn't work, give the path to your company's root certificate with NODE_EXTRA_CA_CERTS. |
Source: Gemini CLI troubleshooting and FAQ docs; Google deprecation page; October 15, 2026.
Frequently asked questions
What is Gemini CLI and what is it for?
Google's open-source AI agent that runs in the terminal. In a project folder, it reads files, changes them with your approval and runs commands; it can be extended with MCP, extensions and skills.
Is Gemini CLI still free?
The software is free and open source. But sign-in with a personal Google account, and the free quota that came with it, ended on June 18, 2026. You now need a paid API key, Vertex AI or an enterprise license; the status of free API keys is contradictory in the docs.
How do you install Gemini CLI?
With Node.js 20 or later installed, type npm install -g @google/gemini-cli in the terminal; on macOS and Linux, brew install gemini-cli also works. Then start it by typing gemini.
Does Gemini CLI work on Windows?
Yes. The docs list Windows 11 24H2 and later as a recommended system and PowerShell as a shell. Node.js 20+ is required.
How do you add MCP to Gemini CLI?
With the gemini mcp add command, or by writing an mcpServers block in ~/.gemini/settings.json. You check the connection with gemini mcp list.
Are Gemini CLI and Antigravity CLI the same thing?
No. Antigravity CLI is a separate terminal tool that Google recommends in place of Gemini CLI (command: agy). Google offers a migration tool that carries over extension, skill and MCP settings.
Tasks and setup kits that use Gemini CLI
Related guides
Sources
- Gemini CLI GitHub repository (README, license)
- Installation docs
- Authentication docs
- Quota and pricing docs (describe the old situation)
- License, terms and privacy
- Command reference (cheatsheet)
- MCP server setup
- Google announcement: transitioning Gemini CLI to Antigravity CLI
- Gemini Code Assist consumer accounts (shutdown)
- Gemini Code Assist deprecations
- Antigravity: migrating from Gemini CLI
- Gemini CLI plans page
- Latest releases
Not clear in the official sources: Whether a free Gemini API key works in Gemini CLI today (the docs contradict each other) · Antigravity CLI's price and quota · access from Turkey · a Turkish interface and Turkish responses · local-currency pricing · cancellation and subscription steps · the current status of Vertex AI Express mode · the absence of published security advisories (the list was empty, which is not a guarantee). Each is noted in the relevant section along with what you can do.





