gemini-cli-mcp
MCP server for Google's Gemini CLI with OAuth, enabling use of Gemini through Claude Desktop/Code without API keys – just log in with a Google account.
README
Gemini CLI MCP Server
A Model Context Protocol (MCP) server and CLI wrapper for Google's Gemini CLI with OAuth authentication support. This allows you to use Gemini through Claude Desktop/Code without needing API keys - just login with your Google account.
Features
- OAuth Authentication: Login with your Google account, no API keys required
- MCP Server: Full integration with Claude Desktop and Claude Code
- CLI Tool: Command-line interface for direct usage
- Multi-Model Support: Use different Gemini models (gemini-2.5-pro, gemini-2.5-flash, gemini-2.5-flash-lite)
- File Analysis: Analyze code and documents with context
- Session Persistence: Maintains authentication across sessions
- Headless Mode: Support for environments without browser access
Installation
npm install -g @jacob/gemini-cli-mcp
Prerequisites
You need to have Google's Gemini CLI installed:
# Install globally
npm install -g @google/gemini-cli
# Or use npx
npx @google/gemini-cli
The Gemini CLI provides 60 requests/minute and 1,000 requests/day with a personal Google account.
Quick Start
1. Initial Authentication
# The Gemini CLI handles authentication automatically on first use
# Just run any command and it will prompt for auth if needed
gemini-mcp query "Hello"
# Check authentication status
gemini-mcp auth-status
2. CLI Usage
# Send a query
gemini-mcp query "Explain quantum computing in simple terms"
# Query with specific model (defaults to 2.5-pro)
gemini-mcp query "Write a Python function for binary search" --model gemini-2.5-flash
# Analyze files
gemini-mcp analyze --files src/main.ts src/utils.ts --instruction "Find potential bugs"
# Query with context files
gemini-mcp query "Explain this code" --context main.py utils.py
3. MCP Server Configuration
Add to your Claude Desktop or Claude Code configuration:
{
"mcpServers": {
"gemini": {
"command": "npx",
"args": ["-y", "@jacob/gemini-cli-mcp"],
"env": {
"GEMINI_MODEL": "gemini-2.5-pro",
"GEMINI_TIMEOUT": "30000",
"GEMINI_AUTO_LOGIN": "true",
"GEMINI_THINKING_ENABLED": "true"
}
}
}
}
Available MCP Tools
gemini_query
Query Gemini with a prompt and optional context files.
{
prompt: string; // Your question or prompt
model?: string; // Optional: "gemini-2.5-pro" (default), "gemini-2.5-flash", or "gemini-2.5-flash-lite"
context?: string[]; // Optional: Array of file paths for context
}
gemini_analyze
Analyze files with specific instructions.
{
files: string[]; // Files to analyze
instruction: string; // Analysis instruction
model?: string; // Optional: Model to use
}
gemini_auth_status
Check current authentication status.
gemini_login
Initiate authentication flow.
{
headless?: boolean; // Show URL instead of opening browser
account?: string; // Specific Google account to use
}
gemini_list_models
List available Gemini models.
CLI Commands
Authentication Commands
# Login
gemini-mcp login [options]
--headless # Display URL instead of opening browser
--account <email> # Specify Google account
# Logout
gemini-mcp logout
# Check auth status
gemini-mcp auth-status
Query Commands
# Send query
gemini-mcp query <prompt> [options]
-m, --model <model> # Model to use
-c, --context <files> # Context files
-t, --timeout <ms> # Timeout in milliseconds
# Analyze files
gemini-mcp analyze [options]
-f, --files <files> # Files to analyze (required)
-i, --instruction <text> # Analysis instruction (required)
-m, --model <model> # Model to use
Configuration
# Set configuration
gemini-mcp config --set key=value
# Get configuration
gemini-mcp config --get key
# List all configuration
gemini-mcp config --list
Configuration options:
model: Default model (gemini-2.5-pro, gemini-2.5-flash, gemini-2.5-flash-lite)timeout: Request timeout in millisecondsheadlessDefault: Default to headless authenticationautoLogin: Automatically login if not authenticatedthinkingEnabled: Enable thinking mode by default (shows model's reasoning)thinkingBudget: Default token budget for thinking process
Environment Variables
GEMINI_MODEL: Default model to use (gemini-2.5-pro, gemini-2.5-flash, gemini-2.5-flash-lite)GEMINI_TIMEOUT: Request timeout (ms)GEMINI_AUTH_CACHE: Authentication cache directoryGEMINI_HEADLESS: Default to headless modeGEMINI_AUTO_LOGIN: Auto-login on first useGEMINI_THINKING_ENABLED: Enable thinking mode by defaultGEMINI_THINKING_BUDGET: Default thinking token budget
Examples
Example 1: Code Review
# Analyze a TypeScript project for issues
gemini-mcp analyze \
--files "src/**/*.ts" \
--instruction "Review this code for potential bugs, security issues, and suggest improvements"
Example 2: Documentation Generation
# Generate documentation for a module with thinking mode
gemini-mcp query "Generate comprehensive JSDoc documentation for this module" \
--context src/auth-manager.ts \
--model gemini-2.5-flash \
--thinking-enabled
Example 3: Using in Scripts
#!/bin/bash
# Check if authenticated
if ! gemini-mcp auth-status > /dev/null 2>&1; then
echo "Please login first: gemini-mcp login"
exit 1
fi
# Run analysis
gemini-mcp analyze --files "$@" --instruction "Find security vulnerabilities"
Troubleshooting
Authentication Issues
- Browser doesn't open: Use
--headlessflag and manually visit the URL - Session expired: Run
gemini-mcp loginagain - Wrong account: Use
--account email@gmail.comto specify account
Connection Issues
- Ensure Gemini CLI is installed and in PATH
- Check internet connectivity
- Verify Google account has access to Gemini
MCP Server Issues
- Restart Claude Desktop/Code after configuration changes
- Check logs for error messages
- Verify the MCP server is running:
npx @jacob/gemini-cli-mcp
Security
- Authentication tokens are stored locally in
~/.gemini-cli-mcp/auth/ - Tokens are only accessible by your user account
- Use
gemini-mcp logoutto clear stored credentials - Never share your authentication cache directory
Development
# Clone the repository
git clone https://github.com/yourusername/gemini-cli-mcp.git
cd gemini-cli-mcp
# Install dependencies
npm install
# Build
npm run build
# Run locally
npm start
# Development mode
npm run dev
License
MIT
Contributing
Contributions are welcome! Please read our contributing guidelines and submit pull requests to our repository.
Support
For issues and feature requests, please use the GitHub issue tracker.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。