Gemini CLI
Google's open-source terminal AI agent with Gemini and built-in Google Search grounding
About
Gemini CLI is Google's open-source AI agent that brings Gemini directly into the terminal: it queries and edits large codebases (within a 1M-token context window), generates apps from PDFs or sketches, automates operational tasks like PR queries and complex rebases, and connects to MCP servers for extra capabilities — including media generation with Imagen, Veo, and Lyria.
Sign in with a personal Google account for a generous free tier (60 requests/min, 1,000/day with Gemini 3 models) — no API key needed. API-key and Vertex AI auth are available for specific model control and enterprise use. GEMINI.md files give persistent per-project context, and it runs non-interactively (gemini -p) for scripts and CI. Its standout differentiator among terminal agents is built-in Google Search grounding, letting the agent pull real-time API docs and library info mid-task.
What it can do
- • Query and edit large codebases within a 1M-token context window (Gemini 3 models)
- • Built-in Google Search grounding for real-time information mid-task
- • File operations, shell command execution, web fetching, multimodal input (PDFs, images, sketches)
- • Non-interactive mode 'gemini -p' with --output-format json / stream-json for scripts and CI
- • MCP servers and an extension ecosystem (incl. Google's Security and Conductor extensions; media generation with Imagen/Veo/Lyria)
- • GEMINI.md project context files; conversation checkpointing to save and resume sessions
- • Gemini CLI GitHub Action: automated PR reviews, issue triage, and @gemini-cli mentions
Setup
- Install Node.js 18 or higher.
- Install the CLI: 'npm install -g @google/gemini-cli' (or run one-off with 'npx @google/gemini-cli'; also available via brew and conda).
- Run 'gemini', pick a color theme, then choose 'Sign in with Google' and complete the browser OAuth flow. Free tier: 60 requests/min and 1,000 requests/day.
- For a specific model or higher limits, get a key from Google AI Studio and set 'export GEMINI_API_KEY="YOUR_API_KEY"' (free tier: 1,000 requests/day with Gemini 3, mix of flash and pro).
- For enterprise use: 'export GOOGLE_API_KEY="YOUR_API_KEY"' and 'export GOOGLE_GENAI_USE_VERTEXAI=true'.
- Add a GEMINI.md file to your project root for persistent project context; use 'gemini -p "..."' with '--output-format json' for non-interactive scripting.
Prerequisites
Node.js 18+; a personal Google account (free tier) or a Gemini API key / Vertex AI project (paid); an internet connection at install and runtime.
Use cases
Works with
Trust score — 9/10
Scored 9/10 (verified tier, metrics-v1). Evidence: 107,298 GitHub stars (log-scaled vs catalog peers); blended with ~1,729,660 monthly package downloads; last push 0d ago; 100 releases/12mo; 760 open issues. Manual review: free tier 60 req/min + 1,000/day confirmed in README; no red flags. License: Apache-2.0.
★ 107,298 GitHub stars
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 9.2/10
- Maintenance · recency, releases, issue load
- 9.7/10
- Openness · license + docs
- 10.0/10
- Manual review · sanity read + red-flag check
- 9.0/10
- · 107,298 GitHub stars (log-scaled vs catalog peers)
- · blended with ~1,729,660 monthly package downloads
- · last push 0d ago; 100 releases/12mo; 760 open issues
- · manual review: free tier 60 req/min + 1,000/day confirmed in README; no red flags
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.