What is Gemini CLI?
Gemini CLI is an open-source AI agent by Google that makes the Gemini models available directly in the terminal. The agent is published on GitHub under the Apache 2.0 license and uses a reason-and-act (ReAct) loop: it plans a step, runs a tool, evaluates the result, and plans the next step. Built-in tools cover Google Search and grounding, web fetch, reading and writing files, shell commands, and grep. Through the Model Context Protocol (MCP), you can extend its capabilities with custom tools and data sources.
Gemini CLI solves a concrete problem in day-to-day development: many tasks happen on the command line, while AI assistance has so far been tied mostly to the IDE or the browser. With Gemini CLI, developers bring the AI agent to where they already work. The agent can read and change code, run commands, and research on the web without requiring a context switch. In Google Cloud Shell, Gemini CLI is available with no additional setup.
Core Features
- ReAct loop with built-in tools: The agent combines reasoning and tool execution in a loop, using built-in tools such as Google Search, web fetch, file operations, shell commands, and grep.
- MCP extensibility: Local and remote Model Context Protocol servers add tools, APIs, and internal data sources, so the agent can be adapted to your own environment.
- Slash commands for control: Built-in commands such as /memory, /stats, /tools, and /mcp control context, configuration, and available tools directly within the session.
- Cross-platform and scriptable: Gemini CLI runs on macOS, Windows, and Linux in any terminal, is usable without setup in Cloud Shell, and supports a non-interactive mode for script-based automation.
Typical Use Cases
Analyze and fix bugs: Developers let the agent read the relevant code, narrow down the root cause of an error, and propose or directly apply a fix without leaving the terminal.
Generate features and tests: The agent creates new functionality, improves test coverage, and adjusts existing files by combining file operations and shell commands within the ReAct loop.
Connect internal systems via MCP: Through custom MCP servers, the agent accesses internal APIs, databases, or tools and becomes part of your own development and operations processes.
Benefits
- Open source under Apache 2.0: transparent implementation, custom extensions, and no proprietary lock-in
- AI assistance directly in the terminal instead of context switching between IDE, browser, and command line
- Extensible via MCP: connect your own tools and data sources to internal processes
- Flexible pricing from a free tier through a Code Assist subscription to pay-as-you-go via API key
Integration with innFactory
As a certified Google Cloud Partner, innFactory supports you with the adoption and operation of this service.
Typical Use Cases
Frequently Asked Questions
What is Gemini CLI?
Gemini CLI is an open-source AI agent (Apache 2.0) by Google that makes the Gemini models available directly in the terminal. The agent uses a reason-and-act (ReAct) loop and built-in tools such as Google Search, web fetch, file operations, shell commands, and grep. Through the Model Context Protocol (MCP), it can be extended with custom tools and data sources. Google announced Gemini CLI on June 25, 2025.
When should I use Gemini CLI?
Gemini CLI is a good fit when you want to work directly in the command line without switching to an IDE: analyzing and fixing bugs, generating new features and tests, doing research and content generation, and automating recurring tasks via scripts in non-interactive mode. It is also useful when you want to connect the agent to your own internal systems through MCP.
How much does Gemini CLI cost?
Gemini CLI itself is free and open source. A free tier with a daily request quota is available when you sign in with a personal Google account. For higher quotas and production use, you can use a Gemini Code Assist subscription (Standard or Enterprise) or a Gemini API key with pay-as-you-go pricing. Quotas are shared between Gemini CLI and Gemini Code Assist agent mode. Current limits are listed in the official pricing documentation.
How do I extend Gemini CLI with MCP?
Gemini CLI supports local and remote Model Context Protocol (MCP) servers. With them you connect additional tools, APIs, and internal data sources that the agent can use within its ReAct loop. Configured servers and available tools can be inspected and managed through built-in slash commands such as /mcp and /tools.
