Getting Started with Gemini CLI
1. Installation
Official installation methods for Gemini CLI:
Method 1: Run directly from GitHub (no installation required)
npx @google/gemini-cli
Method 2: Global installation (recommended)
npm install -g @google/gemini-cli
After installation, you can use it directly:
gemini
Note:
- Make sure your Node.js version is 18 or higher
- You might need to use
sudo on Linux/Mac if you encounter permission issues
- If you have problems running directly from GitHub, try the second installation method
2. Getting Your API Key
To use Gemini CLI, you'll need an API key from Google AI Studio:
- Go to Google AI Studio
- Sign in with your Google account
- Navigate to the API keys section
- Create a new API key
Once you have your key, set it in your environment:
export GEMINI_API_KEY=your_api_key_here
For Windows users:
set GEMINI_API_KEY=your_api_key_here
For persistent configuration, add it to your shell profile or use the CLI config:
gemini config set apiKey your_api_key_here
3. Verify Installation
Verify that Gemini CLI is installed correctly:
gemini --version
When running for the first time, you will see the following steps:
- Choose a color theme (you can select from multiple options that suit your terminal)
- Authenticate (you can use either your personal Google account or an API key)
Simple test:
gemini
Then at the prompt, enter:
> Hello, what can you do for me?
4. Real-world Examples
Exploring a New Codebase
cd your-project-directory
gemini
> Describe the main pieces of this system's architecture.
Working with Existing Code
cd your-project-directory
gemini
> Help me migrate this codebase to the latest version of Java. Start with a plan.
Automating Workflows
gemini
> Make me a slide deck showing the git history from the last 7 days, grouped by feature and team member.
Basic Commands
Chat Mode
Start an interactive chat session:
gemini chat
This opens an interactive session where you can have a conversation with Gemini.
One-off Questions
For quick questions without starting a chat:
gemini ask "How do I check disk space on Linux?"
Code Generation
Generate code directly:
gemini code "Create a Python function to fetch JSON data from an API"
Summarizing Text
Summarize text content:
gemini summarize --file document.txt
Or pipe content:
cat document.txt | gemini summarize
Advanced Usage
Working with Files
Process file content:
gemini ask --file code.js "Explain what this code does"
Generate content to a file:
gemini code "Create a React component for a login form" > LoginForm.jsx
Custom Model Selection
Choose specific Gemini models:
gemini ask --model gemini-pro-vision "Describe this image" --file image.jpg
Streaming Responses
Enable streaming for real-time responses:
gemini ask --stream "Write a short story about a robot"
Shell Integration
Use Gemini to help with shell commands:
gemini shell "How to find all files modified in the last 7 days"
Power Features
MCP Servers (Model Context Protocol)
Connect to MCP servers to extend Gemini's capabilities:
gemini chat
Inside the chat session:
/mcp connect http://localhost:3000
You can also specify MCP servers at startup:
gemini chat --mcp http://localhost:3000
MCP servers help Gemini integrate with services like:
- Image generation with Imagen
- Video generation with Veo
- Music generation with Lyria
- Custom enterprise services
Built-in Tools
Gemini CLI includes powerful built-in tools:
- grep: Search for patterns in files
- terminal: Execute shell commands
- file read/write: Manipulate file contents
- web search: Search the internet for information
- web fetch: Download content from URLs
To see available tools in chat:
/tools list
Special Commands
Gemini CLI supports special commands in chat mode:
- /memory: Work with Gemini's long-term memory
- /stats: Display statistics about your current session
- /tools: Manage and use built-in tools
- /mcp: Connect and manage MCP servers
For example, to save something to memory:
/memory save "My project uses React 18 and TypeScript 5"
Context Window
Gemini CLI can handle large codebases with its 1M token context window:
gemini chat
Then ask about your codebase:
What's the architecture of this project?
Gemini will analyze your files and provide insights, all while keeping the context in memory.
Customization
Configuration Settings
View current configuration:
gemini config list
Set a configuration value:
gemini config set temperature 0.7
Custom Aliases
Create command aliases for frequent operations:
gemini alias add explain="ask 'Explain the following code:'"
Then use your alias:
gemini explain --file script.py
Creating Custom Instructions
Set persistent system instructions:
gemini config set systemInstruction "You are a helpful assistant specializing in JavaScript development."
Tips & Tricks
Saving Chat History
Export your chat history to a file:
gemini chat --save chat_history.json
Continue from a previous chat:
gemini chat --load chat_history.json
Combining with Other Tools
Pipe from other commands:
git diff | gemini ask "Summarize these code changes"
Improving Response Quality
Be specific in your prompts for better results:
gemini ask "Write a Python function to parse CSV files that handles quoted fields and escaped quotes"
Instead of just:
gemini ask "Write a CSV parser"