Gemini CLI — Google's AI in Your Terminal

Gemini CLI — Google AI in Your Terminal

Gemini CLI is Google’s open-source command-line AI agent, powered by Google’s current Gemini models (Gemini 3.x as of mid-2026, previously 2.5 Pro) with a 1 million token context window. It brings the same powerful AI that drives Google’s products to your terminal for coding, analysis, and automation tasks. Check the Google Models guide for the exact current default.

Gemini CLI is free and open-source — Google offers a generous free tier via Google AI Studio API keys.

Installation

# Install via npm
npm install -g @google/gemini-cli

# Or run without installing (npx)
npx @google/gemini-cli

# Authenticate
gemini auth login
# Follow the browser flow to connect your Google account / API key

Quick Start

# Start interactive session in your project
cd my-project
gemini

# Ask a one-shot question
gemini "Explain what this codebase does"

# Give it a task
gemini "Add input validation to all API endpoints and write tests"

Key Features

1M Token Context Window

With a 1 million token context window, Gemini CLI can:

• Load your entire project — not just open files
• Process large log files or exported data
• Read multiple repositories simultaneously
• Understand complex dependency graphs

# Load an entire large project
gemini --include "**/*.ts,**/*.json" "Find all security vulnerabilities"

Google Search Grounding

Gemini CLI can search Google in real time during tasks:

gemini "Update the AWS SDK v2 calls to v3 syntax"
# → Searches for AWS SDK v3 migration guide
# → Applies up-to-date patterns to your code

Agentic Mode

Gemini CLI can take autonomous actions:

• Read and write files
• Run shell commands
• Execute tests and fix failures
• Navigate directory structures

gemini "Refactor the authentication module to use JWT and add comprehensive tests"
# Reads auth files → Plans changes → Edits files → Runs tests → Fixes failures

MCP Support

Gemini CLI supports the Model Context Protocol (MCP), allowing it to connect to:

• Databases (read schema, query data)
• APIs (fetch live data)
• Documentation servers
• Internal tools

# Start with MCP server for database access
gemini --mcp-server postgres://localhost/mydb "Optimize these slow queries"

Gemini CLI vs Claude Code

Feature	Gemini CLI	Claude Code
Context window	1M tokens	200K tokens
Search grounding	Yes (Google Search)	Yes (Web search)
Model	Gemini 3.x Flash/Pro	Claude Sonnet 5
MCP support	Yes	Yes
Open source	Yes	Yes
Free tier	Yes (generous via AI Studio)	Requires Claude Pro
File operations	Yes	Yes
Shell commands	Yes	Yes
Best for	Large repos, GCP projects	General coding, AWS

Free Tier

Gemini CLI is free to use with Google AI Studio API keys. Exact request limits vary by current model generation and change over time — check Google AI Studio for current free-tier limits on the Flash and Pro tiers.

For heavier usage, the Gemini API is billed per token (significantly cheaper than OpenAI).

Configuration

# Set default model
gemini config set model gemini-2.5-flash

# Set your project context file (like CLAUDE.md)
cat > GEMINI.md << 'EOF'
# Project Context
- Stack: Node.js + TypeScript + PostgreSQL
- Test runner: Jest
- Lint: ESLint + Prettier
- Always run `npm test` after changes
EOF

# Enable Google Search grounding
gemini config set search true

Common Use Cases

# Understand a new codebase
gemini "Give me a tour of this codebase — what does it do, how is it structured?"

# Fix a failing test
gemini "The test in auth.test.ts line 42 is failing. Fix the underlying issue."

# Security audit
gemini "Review this codebase for OWASP Top 10 vulnerabilities"

# Generate documentation
gemini "Generate OpenAPI 3.0 spec for all API routes in src/routes/"

# Upgrade dependencies
gemini "Upgrade all dependencies to latest versions and fix any breaking changes"

Useful Links

• Gemini CLI GitHub
• Google AI Studio (API Key)
• Gemini API Docs
• Google Anti-Gravity — Google’s IDE-based coding tool