Gemini CLI Cheatsheet

The most basic command. Sends a prompt to the Gemini model and prints the response.

Example: gemini "What are the 5 largest cities in the world?"

Run Gemini CLI without a permanent installation. Perfect for quick tests or infrequent use.

Example: npx @google/generative-ai "Suggest a good name for a tech blog."

Includes the content of a file as part of the prompt context. The file content is prefixed to your prompt.

Example: gemini -f readme.md "Summarize this README file."

Pipes the content of a file (or any standard output) directly to Gemini CLI as its context.

Example: cat package.json | gemini 'Explain the purpose of these dependencies.'

Send multiple files to Gemini for analysis.

Example: gemini --files="v1.js,v2.js" "What changed between versions?"

Request output in JSON format for easy parsing in scripts.

Example: gemini --json "Extract name, email, and phone from: John Doe, john@example.com, 555-1234"

Request output in Markdown format.

Example: gemini --markdown "Write documentation for a REST API"

Output only code blocks without explanatory text.

Example: gemini --code-only "Write a Python function to calculate Fibonacci numbers"

Specify which Gemini model to use.

Example: gemini --model=gemini-1.5-flash "Summarize this paragraph"

Control randomness in responses (0.0-1.0). Lower values are more deterministic.

Example: gemini --temperature=0.9 "Write a sci-fi short story"

Limit the length of the response.

Example: gemini --max-tokens=100 "Summarize this article"

Set a system prompt to guide the model's behavior.

Example: gemini --system="You are a Python expert" "Help me with this code"

Save a conversation to continue later.

Example: gemini --save=project-planning "Let's plan a new web app"

Continue a previously saved conversation.

Example: gemini --load=project-planning "What were the key features we discussed?"

Executes the `ls` command and understands its output. You can ask for specific formats or analysis.

Example: gemini 'Show me the contents of the src folder, recursively, and identify any .css files'

Performs a `grep` search for content within files. Useful for code analysis and discovery.

Example: gemini 'Search for the text `TODO:` in all `.ts` files.'

Uses glob patterns to find files by path. The agent will then typically read or analyze these files.

Example: gemini 'Read all the `.md` files from the blog and give me a summary of each.'

Performs a web search using Google Search to answer questions about current events or up-to-date information.

Example: gemini 'What were the main announcements from the last Google I/O event?'

Allows the agent to remember specific facts or preferences for future interactions.

Example: gemini 'From now on, when you suggest installing a package, use pnpm.' 'Okay, I will remember to use pnpm.'

Executes an arbitrary shell command and analyzes its output. This is a very powerful but potentially dangerous tool.

Example: gemini 'Install the `is-odd` package using npm and then write a node script to use it.'

Set your API key as an environment variable so you don't have to provide it with every command.

Example: Add this to your .bashrc or .zshrc file for persistence.

Use prompting to control the output format. Asking for 'raw code' helps when you want to pipe the output directly into a new file.

Example: gemini -f old.py "..." > new.py