On June 25, 2025, Google launched Gemini CLI: an open-source AI agent that runs in the terminal, released under the Apache 2.0 license on GitHub (google-gemini/gemini-cli), letting developers read codebases, edit files, and run commands with natural language — with self-correction along the way.
The pricing is the real headline. Signed in with a personal Google account, the free preview allows 60 model requests per minute and 1,000 per day — effectively an all-you-can-use tier for individual developers. TechCrunch, Ars Technica, and VentureBeat all covered the launch the same day, and the tool instantly became the talk of developer communities. Installation stays low-friction: a global npm install, an instant npx run, or Homebrew. A non-interactive mode with JSON output plugs into CI pipelines, and conversation checkpointing plus GEMINI.md give long tasks state and context.
An AI agent in the terminal
Gemini CLI is positioned as a terminal-first agent: once it reads your local codebase, you can ask it in plain English to explain code, implement features, debug, install dependencies, or run migration-style command sequences. It runs on Gemini 2.5 Pro and works as a self-correcting agent — acting, checking results, and adjusting its next step.
The built-in capabilities are substantial: Google Search grounding for real-time information, file operations and shell commands, and support for MCP (Model Context Protocol), so you can plug in custom servers and tools such as GitHub, Slack, or databases. Project-level configuration lives in a GEMINI.md file, with personal and team settings managed separately.
The free tier is the real offensive
The allowance — 60 requests a minute, 1,000 a day — was the most discussed part of the launch. TechCrunch, citing Google, noted that this is roughly double what internal testers averaged; teams needing more volume or specific models can pay per use with AI Studio or Vertex API keys.
Against the competition, the move lands hard. Anthropic’s Claude Code and OpenAI’s Codex CLI are tied to subscriptions or API billing; Gemini CLI makes a usable agent free, effectively trading Google compute subsidies for share of the terminal entry point. The terms say preview, and Google can change them — but even if pricing arrives later, the user base and the habit will already be established. For a tool that only asks for a personal Google account, that is a remarkably low barrier. For hobby projects, learners, and students, there is almost no reason to say no.
Ecosystem position and the competition
Gemini CLI does not stand alone. Google also offers Gemini Code Assist in the editor and the asynchronous agent Jules, forming a trio that covers in-editor, in-terminal, and background execution. TechCrunch’s assessment was measured: Claude Code and Codex CLI still integrate more smoothly and run more efficiently, and Google is trading openness and price for the chance to catch up.
Openness under Apache 2.0 has a second meaning: the CLI can be self-hosted and modified. Enterprises can wire it into internal workflows and toolchains without being locked to a single vendor, and within days of launch the community had already started building extensions and custom integrations. Open source answers the lock-in question institutionally, and MCP hands extensibility to the community — together they outsource ecosystem development to developers worldwide. For Google, this is a classic platform play — open the entry point, keep the models and cloud services inside the ecosystem.
What it means for the dev-tools market
By mid-2025, the AI coding race had shifted from “which model is strongest” to “which entry point sticks.” The terminal is where engineers spend most of their time; whoever owns it is best positioned to become the default surface for agentic workflows, and then to carry the rest of the development flow.
Gemini CLI pushed the competition straight into price and openness. For developers, the short-term effect is purely positive: a free allowance, a permissive license, and an extensible tool interface. For rivals, it is pricing pressure that demands a response. How quickly the tool took off shows the market had been waiting for a free, open-source terminal agent for a long time. The launch invites comparison with how VS Code rose: free, open, low-friction — gather the developers first, monetize later. For incumbents in terminal tooling, it starts a re-sorting of the field.
Sources
- TechCrunch: Google unveils Gemini CLI, an open source AI tool for terminals
- SiliconANGLE: Google introduces Gemini CLI
- GitHub: google-gemini/gemini-cli
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.