# Getting Started with Gemini CLI

> Install Gemini CLI with npm, npx or Homebrew, sign in with Google for 60 requests a minute free, and run your first prompt inside a project.

- Canonical: https://guides-ai.pages.dev/guides/gemini-cli-first-steps/
- Plate 02.04 · Topic: First steps (https://guides-ai.pages.dev/topics/first-steps/)
- Published: 02 Sept 2026 · Updated: 06 Sept 2026 · 3 min read
- Source site: guides-ai — https://guides-ai.pages.dev/

Gemini CLI is Google's open-source terminal agent — the same shape as Claude Code. It reads
your files, runs commands and edits code from the shell, and the free tier is genuinely usable.

## 1. Install it

```bash
npm install -g @google/gemini-cli
```

What it does: installs the `gemini` command globally.

```bash
brew install gemini-cli
```

What it does: the Homebrew route on macOS and Linux. To try it without installing anything, run
`npx @google/gemini-cli`.

## 2. Sign in

```bash
gemini
```

What it does: on first run, opens a browser to sign in with your Google account. That path gives
**60 requests per minute and 1,000 requests per day** on a personal account, with no billing
setup.

```bash
export GEMINI_API_KEY="your-key-here"
gemini
```

What it does: authenticates with an AI Studio key instead — the free tier there is 1,000
requests per day with Gemini 3. Set it in your shell profile to make it stick.

## 3. Run your first prompt

```bash
cd /path/to/your/project
gemini
```

What it does: starts the agent scoped to that folder, which is the code it can read. Then ask
for something concrete:

```text
Summarize what this project does and list its main entry points.
```

What it does: has it read the folder and report back. Ask for changes the same way — it proposes
edits before applying them. `/help` lists the slash commands.

## Verify it worked

Ask it to make one small edit and check the file on disk afterwards. If it described a change
but nothing moved, you either declined the confirmation or started it outside the project.

## How it compares

Both Gemini CLI and Claude Code are terminal [agents](/glossary/#agent) with MCP support, so
the practical difference is the account you already have and the free allowance. Claude Code
needs a paid Claude plan; Gemini CLI's Google sign-in does not. For a small repo either is
fine — try the one whose limits you won't hit.

Next: [MCP servers and extensions in Gemini CLI](/guides/gemini-cli-mcp-and-extensions/), or
call the model directly with [your first Gemini API request](/guides/gemini-api-first-request/).

Source: [google-gemini/gemini-cli](https://github.com/google-gemini/gemini-cli).