Gemini API CLI is in beta for agents and scripts

The Gemini API changelog for 27 October 2026 says Google released the Gemini API CLI, gemini-api, in beta. The changelog calls it a stateless, zero-dependency command-line client for AI coding agents, scripts, and CI pipelines. It runs on the Interactions API to generate text and media, analyze files, and run Gemini managed agents.

The CLI guide (opens in a new tab) says the same binary works from Antigravity, Claude Code, Codex, Cursor, OpenClaw, Hermes, or any shell. It is an API client, not an interactive coding agent like @google/gemini-cli. Documented commands include generate, image, tts, music, video, analyze, transcribe, and agent. Media commands write a file and print its path. The guide says this is an open-source beta, breaking changes can land between minor versions, and automated environments should pin a release. Inline file inputs cap at 14 MB. Larger files go through files upload, and the guide says the Files API deletes uploaded files after 48 hours. Authentication is a Gemini API key.

Who it's for: a coding agent, script, or CI job that needs to call Gemini from a shell without adding an SDK to the project.

Fits: Best for Vibe Coding stays Cursor, then Claude Code, then Codex, then Copilot. gemini-api is a client those agents can call. It does not replace the edit loop. Gemini stays the writing and research product it already was. The verdict is unchanged.