Google positions it for terminal use and shell-capable coding agents, alongside its typed Gen AI SDKs for Python, TypeScript, Go, and Java.
The binary is available for macOS, Linux, and Windows, with installation via shell scripts or Go. The source describes commands for listing models, testing prompts, analyzing local documents, transcribing audio to SRT, generating images and video, and producing speech or music. Media commands can write output to disk and return only the resulting file path on standard output.
Google calls the CLI stateless and self-describing, and says it can invoke Managed Agents including Deep Research. The announcement was published on October 19, 2026. It does not establish that the preview interface is stable, which functions or models are available to every account, or the pricing and quota terms for these API operations.