Gemini API CLI: a command line for agents and CI
Google released a beta of the gemini-api client: stateless and dependency-free, for coding agents, scripts and CI pipelines.
On October 27, Google released a beta of the Gemini API CLI. It’s a command-line client called gemini-api, built mainly for coding agents, scripts and CI pipelines rather than for hands-on terminal work. Don’t confuse it with Gemini CLI, which is a separate interactive agent.
What’s new
The client is stateless and has no dependencies. Each feature is a regular shell command that calls the Interactions API. You can use it to:
- generate text and media;
- analyze files;
- run Gemini managed agents.
There are useful flags for agents: --usage shows the full command tree, --schema shows request JSON schemas, and --dry-run -o json lets you check a call offline. The docs also include a snippet for AGENTS.md or CLAUDE.md so a coding agent knows about the tool.
How to try it
You can install it with a script on macOS, Linux or Windows, through go install, or by downloading a prebuilt binary from GitHub Releases. For Docker and CI, the docs suggest binaries, since they let you pin a version. The key goes in the GEMINI_API_KEY variable.
Limitations
This is a beta, and the GitHub repository calls the client experimental, so commands may still change. Pin a version in any script that needs to keep working.
Prepared with AI from public sources.