On June 25, Google unintentionally published a blog post about Gemini CLI, an open-source project, only to remove it shortly after. The post, which introduced a free command-line tool that would make Google’s Gemini AI models accessible within developer terminals, was discovered by developer Meet Patel on X (formerly Twitter) and captured via a screenshot and archive.org link. The tool, based on the Gemini 2.5 Pro model, offers a vast 1 million token context. The Gemini CLI offers 60 model requests per minute, up to 1,000 per day, for free with a Google login, and supports Gemini Code Assist for VS Code. The tool uses Google Search for real-time web grounding and allows for extensions, scripting, and automation. A notable aspect of the tool is the provision of a free Gemini Code Assist license with a personal Google account login, offering significant value to individual coders. The tool aims to enhance developer efficiency by reducing the need to switch between different tools and environments. It is designed for coding, debugging, and experimentation, and could even be used for generating videos and research tasks. The open-source nature of Gemini CLI, licensed under Apache 2.0, promotes code auditing, extension development, and workflow customization. The reason for the sudden removal of the post, authored by Google engineers Taylor Mullen and Ryan Salva, remains unknown, but the cached version reveals the likely future of AI agents for developers.
Gemini CLI: Google’s Command-Line AI Tool Accidentally Leaked, Quickly Pulled From Public View
