jamubc/gemini-mcp-tool

What it solves

This project provides a bridge between AI assistants (like Claude) and the Gemini CLI (or its successor, the Antigravity CLI), allowing assistants to leverage Gemini's massive token window for analyzing large files and entire codebases without consuming excessive tokens in the primary assistant's context window.

How it works

It implements a Model Context Protocol (MCP) server that exposes Gemini's capabilities as tools. The AI assistant can call these tools to send prompts to the Gemini CLI, using a special @ syntax to reference specific files or directories. It supports both standard analysis and a "sandbox mode" for safe code execution and testing.

Who it’s for

Developers using MCP-compatible AI assistants (such as Claude Code or Claude Desktop) who want to integrate Google Gemini's large-context analysis and sandboxed code execution into their workflow.

Highlights

• Large Context Analysis: Uses Gemini's token window to handle massive projects and files via the @ syntax.
• Sandbox Execution: Provides a dedicated tool for safely testing code and running scripts in an isolated environment.
• Dual Backend Support: Automatically handles the transition from the retired Gemini CLI to the Antigravity CLI (agy).
• Easy Integration: Simple one-line setup for Claude Code and standard configuration for Claude Desktop.