google-gemini/gemini-cli
OfficialAn open-source AI agent that brings the power of Gemini directly into your terminal.
Gemini CLI is Google's open-source AI agent that brings Gemini 3 directly into your terminal, with a free tier of 1,000 requests per day and a 1M token context window. If you want a fast, terminal-first coding assistant without paying per call, this is where most developers start.
Our Review
Gemini CLI is Google's open-source terminal AI agent -- 104,000 GitHub stars and 13,700 forks as of May 2026. Released in mid-2025 by the Google Gemini team and distributed as @google/gemini-cli on npm, it gives you direct terminal access to Gemini 3 with the most generous free tier of any major coding agent.
What Gemini CLI does:
- •Free Gemini 3 access 60 requests/minute and 1,000 requests/day with any personal Google account, no card required, no per-token billing for individual developers.
- •1 million token context window navigate entire monorepos in a single context, more than 5x the context of competing CLI agents in early 2026.
- •Built-in Google Search grounding queries can reach real-time web information, useful for asking about packages, errors, or current API behavior.
- •MCP client and server connect external MCP servers for custom integrations, or expose Gemini CLI capabilities as MCP services.
- •Multimodal input accept PDFs, images, and design sketches as part of your prompt to generate apps or analyze visual context.
- •Custom context via GEMINI.md per-project instructions file that tailors agent behavior, similar to CLAUDE.md.
- •Non-interactive scripting run inside shell scripts or CI pipelines for automated PR triage, code review, and refactoring tasks.
Gemini CLI ecosystem:
- •Gemini CLI GitHub Action automated PR reviews, issue triage, on-demand
@gemini-climentions in issues and pull requests. - •MCP server library works with the broader MCP ecosystem (filesystem, GitHub, Slack, Postgres, custom servers).
- •Google Cloud integration Vertex AI media generation via Imagen, Veo, and Lyria MCP servers.
- •Three release channels nightly (daily), preview (Tuesdays UTC 23:59), stable (Tuesdays UTC 20:00 with QA).
Getting started:
Install with npm install -g @google/gemini-cli or run instantly with npx @google/gemini-cli. Sign in with your personal Google account for the free tier. Homebrew (brew install gemini-cli), MacPorts, and Anaconda installation also supported. Full docs at geminicli.com/docs.
Limitations:
Free tier requires a personal Google account, so Workspace accounts may need a different authentication method or a paid tier. Code execution runs in your local shell by default, not a sandboxed container (Codex CLI's differentiator). On complex multi-file refactors, third-party benchmarks place Gemini CLI's first-pass correctness at 85-88% versus Claude Code's 92%. Nightly builds can ship regressions, so use the stable channel for production work.
Our Verdict
Gemini CLI in 2026 is the strongest "first AI coding agent" for developers who want to learn the pattern without paying anything. 104,000 GitHub stars, weekly stable releases, Apache 2.0 license, and 1,000 free requests per day with any personal Google account.
For Vibe Builders, Gemini CLI removes the friction of card-on-file billing and per-token pricing. The free tier is generous enough to cover small-project development for months. The GEMINI.md per-project context file works the same way as CLAUDE.md, so the patterns transfer if you later switch tools.
For Developers, the 1M token context window is the differentiator. Where Claude Code and Codex top out around 200K tokens, Gemini CLI loads entire monorepos in a single shot. Pair that with built-in Google Search grounding, MCP client and server support, and the official GitHub Action for PR review and you have a complete CLI agent for serious work.
Skip Gemini CLI if you need 92% first-pass correctness on complex multi-file refactors -- Claude Code still leads that benchmark. Skip if you need sandboxed autonomous execution by default -- Codex CLI runs commands inside isolated containers and Gemini CLI does not. Skip if Workspace account auth is your only option -- the generous free tier is for personal Google accounts.
Frequently Asked Questions
What is Gemini CLI?
Gemini CLI is Google's open-source terminal AI agent built by the Google Gemini team. It gives you direct access to Gemini 3 from the command line with built-in Google Search grounding, MCP support, and a 1 million token context window. As of May 2026 it has 104,000 GitHub stars and is distributed as @google/gemini-cli on npm under the Apache 2.0 license.
Is Gemini CLI free?
Yes. Gemini CLI offers 60 requests per minute and 1,000 requests per day with any personal Google account, no credit card required and no per-token billing for individual developers. Workspace accounts may need a different authentication path or paid tier. The CLI itself is open source under Apache 2.0 with no license cost.
Gemini CLI vs Claude Code vs Codex -- which should I choose?
Choose Gemini CLI when you want the largest context window of 1M tokens, a generous free tier, and tight integration with Google Search grounding. Choose Claude Code when you need the highest first-pass correctness on complex multi-file refactors. Choose Codex CLI when you need sandboxed autonomous execution inside isolated containers by default.
How do I install Gemini CLI?
Install Gemini CLI with npm: run npm install -g @google/gemini-cli for a global install, or run npx @google/gemini-cli once without installing. Homebrew users on macOS or Linux can use brew install gemini-cli. MacPorts and Anaconda installs are also supported. Then sign in with a personal Google account to activate the free tier.
Does Gemini CLI support MCP?
Yes. Gemini CLI works as both an MCP client and an MCP server. As a client it connects to external MCP servers like filesystem, GitHub, and Slack for extended capabilities. As a server it can expose its own tools to other MCP-aware agents. Google publishes companion MCP servers for Imagen, Veo, and Lyria media generation via Vertex AI.
What is gemini-cli?
Gemini CLI is Google's open-source AI agent that brings Gemini 3 directly into your terminal, with a free tier of 1,000 requests per day and a 1M token context window. If you want a fast, terminal-first coding assistant without paying per call, this is where most developers start.
What license does gemini-cli use?
gemini-cli uses the Apache-2.0 license.
What are alternatives to gemini-cli?
Explore related tools and alternatives on My AI Guide.
Open source & community-verified
Apache-2.0 licensed: free to use in any project, no strings attached. 104,383 developers have starred this, meaning the community has reviewed and trusted it.
Reviewed by My AI Guide for relevance, quality, and active maintenance before listing.
Topics