Google just rolled out a new extensions system for its Gemini CLI command-line AI tool, letting external tools like Figma and Stripe integrate directly into the terminal environment—without requiring Google’s endorsement. According to TechCrunch, the new Gemini CLI Extensions feature allows outside companies to publish extensions publicly on GitHub and lets developers install them manually. This move comes hot on the heels of OpenAI’s launch of ChatGPT apps, though Google’s approach is far less curated: unlike the closed app store model from OpenAI, any developer can create and publish a Gemini CLI extension. The extension launch effectively transforms Gemini CLI from just a terminal AI assistant into a more modular, extensible ecosystem. The CLI itself, introduced in June, now supports the Gemini 2.5 Pro model and is available under an open-source license, enabling developers to tap advanced AI features directly from their terminal workflows.
Sources: InfoWorld, TechCrunch
Key Takeaways
– The new Gemini CLI Extensions let third parties publish integrations without needing Google’s approval, favoring open extensibility over a curated app store model.
– Gemini CLI is already open source and supports Gemini 2.5 Pro, and with extensions, it is positioning itself as a modular platform rather than a closed AI tool.
– Google’s move clearly targets competition with OpenAI’s ChatGPT app ecosystem, offering a freer, developer-centric alternative.
In-Depth
When Google first introduced Gemini CLI in June, it marked a clear bet on embedding AI more deeply into developer workflows. The idea: rather than switching back and forth between your code editor, browser, or AI assistant, you could just talk to Gemini from your terminal. That launch included generous free limits, open-source licensing, and support for coding, research, and even media generation tasks. With today’s extensions announcement, Google is upping the ante—turning the CLI into a platform.
The new Gemini CLI Extensions allow third-party services to build integrations (extensions) that users can install themselves. Companies like Figma and Stripe are among the first to build such extensions, giving developers direct terminal access to their tools. That means workflows like “generate design assets,” “trigger payments,” or “check usage stats” might be done entirely from the terminal, via AI prompts. Importantly, unlike some AI app ecosystems (like OpenAI’s, which are curated), Google is enabling open publishing of extensions, hosted on public repositories like GitHub, with developers free to explore, audit, or fork them. This approach increases innovation potential but also raises trust and security considerations.
The backdrop here is competition: just days earlier, OpenAI launched its own app system within ChatGPT, tightly curates which apps are offered, and controls the ecosystem. Google’s approach is more decentralized—letting the community experiment and iterate. That said, the underlying Gemini CLI remains the same: it leverages the Gemini 2.5 Pro model, supports large context windows, and can ground prompts using web search and local context. Google also supports extension protocols like MCP (Model Context Protocol) to let developers connect external services.
From a strategic perspective, Google is positioning Gemini CLI as a serious alternative to AI-first platforms, especially for developers. By building a more open, modular system, it aims to accelerate adoption and reduce friction. But openness comes with tradeoffs: extension security, maintaining compatibility, and ensuring quality can be hard at scale. Still, for developers who prefer flexibility and control, this extension model could be compelling.