Gemini CLI

Gemini CLI is Google’s open-source terminal coding agent.

Overview


Type	CLI
Open Source	Yes
Best For	Large context, multimodal work, Google ecosystem

Key Features

Flexible Access — Works with both Google account access and API keys
Large Context — Up to 1M tokens (see caveats below)
Agentic — File editing, terminal commands
Google Integration — Works with Google Cloud
Multimodal — Image understanding

Installation

npm install -g @google/gemini-cli

Access Model

Gemini CLI supports both Google account access and API-key-based access. Limits and account terms change often, so check Google’s current docs before relying on a specific allowance.

Privacy

Setting	Value
Open Source	Yes (client)
Training	GCP terms apply
Jurisdiction	US (Google)

Models Available

Gemini 3 Pro (up to 1M context)
Gemini 3 Flash (fast iteration)
Gemini Deep Think (reasoning)

Getting Started

# Authenticate with Google
gemini auth login

# Or use API key
export GOOGLE_API_KEY=...

# Start in your project
cd your-project
gemini

Key Commands

Command	Action
`/help`	Show commands
`/clear`	Reset context
`/model <name>`	Switch model
`/image <path>`	Add image to context

Use Cases

Large Codebase Analysis

Gemini’s large context window helps with codebase-wide tasks:

gemini "Analyze this entire codebase and explain the architecture"

Note: For very large codebases, results may be inconsistent. Consider chunking or using targeted queries.

Visual/UI Work

Gemini excels at visual tasks:

gemini /image screenshot.png "Improve the UI of this page"

Multimodal Debugging

gemini /image error-screenshot.png "What's causing this error?"

Tips

Use Gemini 3 Pro for larger context needs
Use Gemini Flash for quick iterations
Leverage multimodal for UI/visual tasks
Start with the access path that matches your Google workflow, then verify current limits separately
Don’t rely on full 1M context — quality degrades past ~200k