Skip to main content

Gemini CLI

Gemini CLI is Google's open-source terminal agent for coding tasks. It connects to Google's Gemini models and offers a generous free tier, making it accessible for individual developers and small teams.

Installation​

npm install -g @google/gemini-cli

Or run directly without installing:

npx @google/gemini-cli

Starting Gemini CLI​

Run it from inside your project directory:

gemini

This launches an interactive session where you can have ongoing conversations with the AI. You can also pass a prompt directly for one-off queries:

gemini -p "explain the authentication flow in this project"

Authentication​

Three options:

  • Google OAuth (default): sign in with your Google account for free-tier access
  • API key: set GEMINI_API_KEY for direct API access
  • Vertex AI: use Google Cloud credentials for enterprise deployments

Key features​

Large context window​

Gemini CLI supports up to 1 million tokens of context, allowing it to process large codebases in a single session without losing track of earlier files.

Free tier​

With Google OAuth login, you get:

  • 60 requests per minute
  • 1,000 requests per day

This is enough for most individual development workflows without needing a paid plan.

Google Search grounding​

Gemini can search the web during a session to find documentation, API references, or solutions. This keeps its responses current without relying solely on training data.

MCP support​

Gemini CLI supports the Model Context Protocol (MCP), so you can connect it to external tools and data sources using the same server ecosystem as Claude Code.

Project context files​

Gemini reads GEMINI.md files (similar to CLAUDE.md) for project-specific instructions. It also respects AGENTS.md as a cross-tool standard.

Tips​

  • Use GEMINI.md at your project root to set conventions and constraints
  • For large refactoring tasks, the 1M context window helps Gemini keep track of dependencies across files
  • Combine with the free tier for low-cost experimentation before committing to a paid plan

Skills and reusable prompts​

Gemini CLI does not have a built-in slash command or skills system. To replicate skills:

  • Use shell functions or wrapper scripts. Place your prompt templates in a prompts/ directory in your repository and reference them explicitly:
gemini "$(cat prompts/review.md)"

Or define a shell helper:

gemini-skill() {
  local skill="$1"; shift
  gemini "$(cat ~/gemini-skills/$skill.md) $*"
}

gemini-skill review

You can use GEMINI.md to document the available skills and how to invoke them, so the model is aware of the convention even if it cannot auto-invoke them.

For background on skills as a concept and how Claude Code handles them natively, see Skills.

Limitations​

  • Execution speed can be slower than other CLI tools, especially for complex multi-step tasks
  • Can be unreliable on tasks that require precise, multi-file coordination
  • Limited to Google's Gemini models (no support for Claude, GPT, or open-source models)