Google Brings Gemini AI to Developers’ Terminals

Google is bringing its powerful Gemini AI models directly to developers’ command lines with the introduction of a new tool, Gemini CLI. Announced on Wednesday, this agentic AI tool is designed to run locally from the terminal, offering developers a natural and intuitive way to interact with their codebases.

With Gemini CLI, developers can now ask questions like:

• Explain sections of code
• Write new features
• Debug existing code
• Run commands

The tool enables natural language interaction, making it easier for developers to communicate with their code using simple requests.

Gemini CLI joins Google’s growing suite of AI coding tools, including Gemini Code Assist and Jules, the asynchronous AI coding assistant. But Gemini CLI is designed to compete more directly with tools like OpenAI’s Codex CLI and Anthropic’s Claude Code, which are known for their speed and seamless integration into development workflows.

Versatility Beyond Code

While primarily built for coding, Google says Gemini CLI can do much more:

• Generate videos via the Veo 3 model
• Create research reports using the Deep Research agent
• Access real-time information through Google Search
• Connect to MCP servers and external databases

To encourage collaboration and adoption, Google is open-sourcing Gemini CLI under the Apache 2.0 license, allowing the developer community to contribute on GitHub. The company hopes this move will help build a thriving ecosystem around the tool.

Generous Free Usage Limits

Developers can use Gemini CLI with high limits even as free users:

• 60 model requests per minute
• 1,000 requests per day

According to Google, this is “roughly double the average number of requests developers made when using the tool,” aimed at removing barriers for new users.

AI coding tools like Gemini CLI are surging in popularity, but challenges remain. A 2024 Stack Overflow survey reveals that only 43% of developers trust the accuracy of such tools. Independent studies also indicate that AI-generated code can sometimes introduce bugs or miss security vulnerabilities.