Gemini CLI — Google's open-source AI agent that brings Gemini models into the terminal for code-aware chat, edits and scripts.
What it is
Gemini CLI is Google's open-source (Apache-2.0) command-line agent for working with Gemini models directly in your terminal, with local project context, tool use, MCP integrations and a scriptable headless mode.
Editor's review
Long-form introduction by the BetterPicker editors · checked against the official site · Oct 23, 2026
Gemini CLI is Google's open-source bet on terminal-based AI coding: an Apache-2.0 agent that reads your project, runs tools and edits files from a prompt, installed with npx or a global npm package. Everything interesting about its cost lives in the authentication choice. Official documentation lists 1,000 model requests per user per day when you sign in with a Google account, 250 per day on the free Gemini API key tier (Flash models only) and higher fixed quotas through paid subscriptions such as Google AI Pro. The project README also notes that as of June 2026 unpaid-tier and Google One users are being transitioned to Google's Antigravity CLI, and that Gemini CLI is not optimized for the newest models - so it is worth re-checking quota and model terms before building habits on it. Stable releases land weekly, preview builds arrive on Tuesdays ahead of them, and the agent runs anywhere Node.js does.
What it does well
Open source and scriptable: the codebase is Apache-2.0, sessions run interactively or headlessly with a prompt flag, and JSON or streamed JSON output makes it usable inside scripts and pipelines rather than only as a chat surface.
Project-aware agent behavior: it works from your working directory, reads GEMINI.md instruction files, offers a read-only plan mode, session checkpointing and model routing, and connects to MCP servers to gain new tools.
A real free tier for individuals: signing in with a Google account covers 1,000 model requests per day in the current documentation, which is generous headroom for experimentation compared with most paid-only assistants.
Who it's for
Developers who already live in the terminal and want a model sitting next to their shell, git workflow and editors; teams standardizing on MCP tooling; and anyone comfortable wiring their own quota path through a Google account, an API key or Vertex AI. If your mental model of AI coding is a chat panel with visual diffs and an approval button, reviewers note the CLI asks you to lean on git discipline instead. Reviewers who compare it with rival terminal agents tend to frame it as the open, hackable option whose ceiling depends on how much quota you are willing to arrange.
Where it falls short
Quota management is part of the job: users report burning through free requests faster than expected on agentic tasks, limits differ by authentication method, and the README's transition note toward Antigravity CLI leaves the free tier's long-term shape unsettled.
No built-in revision safety net: changes land directly in your working tree with no visual version tracking of AI edits like IDE chat panels provide, so frequent commits, checkpointing and careful reviews are on you.
Terminal-only is a commitment: long agentic sessions mean living in a scrolling TUI, one-shot app builders can feel more approachable for greenfield work, and the repo itself steers developers chasing the newest Gemini models elsewhere.
Specs at a glance
Facts from the official site · not editorial opinion
| License | Apache-2.0, fully open source |
|---|---|
| Price | Free to install; usage has a free tier by sign-in method, with paid tiers above it |
| Free quota | 1,000 model requests per day with Google account sign-in; 250 per day on the free Gemini API key (Flash models only) |
| Platforms | macOS, Linux and Windows via Node.js; also Homebrew, MacPorts and Anaconda packages |
| Core capability | Terminal AI agent with local project context, tool use and headless scripting |
| Integrations | MCP servers, IDE integration, GEMINI.md context files, hooks and extensions |
| Best for | Terminal-first developers who want model access beside git and will manage quotas deliberately |
Frequently asked questions
▸Is Gemini CLI free to use?
The tool itself is open source under Apache-2.0 and free to install. Usage depends on authentication: current official docs list 1,000 model requests per user per day signing in with a Google account, 250 per day on the free Gemini API key tier with Flash models only, and larger quotas through paid subscriptions or pay-as-you-go billing.
▸How do I install and start it?
Run it without installing using npx, install globally with npm, or use the Homebrew, MacPorts or Anaconda packages. Then run gemini inside a project directory, pick a theme and authenticate with your Google account or an API key.
▸Can I use it in scripts and CI?
Yes. Headless mode takes a prompt from the command line, returns plain text by default, and supports JSON or streamed JSON output for structured results. Preview, stable and nightly release channels are published on a weekly cadence.
▸What is the Antigravity CLI note about?
The project README states that as of June 2026, unpaid-tier and Google One users are being transitioned to Google's Antigravity CLI, and that Gemini CLI is not optimized for the latest Gemini models. Quota and model terms still shift, so verify them in the official documentation before committing workflows.
Video reviews
4 review videos aggregated · praise and criticism included alike · click through to the original video
Channels that covered it
Channels are aggregated as sources only — we don’t rate creators
Related tools
Where to go next
External links open in a new tab; external content is independent of this site.
Link down? Every object page is re-checked monthly.