Gemini DeepSearch MCP 是一款自动化研究代理，利用 Google Gemini 模型和 Google 搜索进行深入、多步的网络研究。它生成复杂的查询，综合搜索结果信息，识别知识空白，并生成高质量且引用丰富的答案。

特色

• 利用双子座模型和谷歌搜索的自动化多步研究
• FastMCP 集成支持 HTTP API 和 stdio 部署
• 可配置的工作量等级（低、中、高）用于研究深度
• 引用丰富的回答带来源追踪
• LangGraph 驱动的工作流程，带状态管理

用途

开发服务器（HTTP + Studio UI）

用 Studio UI 启动 LangGraph 开发服务器：

本地MCP服务器（stdio）

启动MCP服务器时，使用stdio传输以便与MCP客户端集成：

测试

运行测试套件：

测试MCP标准服务器：

使用MCP检查器

兰史密斯追踪

API

deep_search 工具接受：

• query（字符串）：研究问题或待调查主题
• 努力（字符串）：研究努力水平——“低”、“中等”或“高”
• 低：1查询，1循环，闪存模型

Gemini DeepSearch MCP

Gemini DeepSearch MCP is an automated research agent that leverages Google Gemini models and Google Search to perform deep, multi-step web research. It generates sophisticated queries, synthesizes information from search results, identifies knowledge gaps, and produces high-quality, citation-rich answers.

Features

• Automated multi-step research using Gemini models and Google Search
• FastMCP integration for both HTTP API and stdio deployment
• Configurable effort levels (low, medium, high) for research depth
• Citation-rich responses with source tracking
• LangGraph-powered workflow with state management

Usage

Development Server (HTTP + Studio UI)

Start the LangGraph development server with Studio UI:

make dev

Local MCP Server (stdio)

Start the MCP server with stdio transport for integration with MCP clients:

make local

Testing

Run the test suite:

make test

Test the MCP stdio server:

make test_mcp

Use MCP inspector

make inspect

With Langsmith tracing

GEMINI_API_KEY=AI******* LANGSMITH_API_KEY=ls******* LANGSMITH_TRACING=true make inspect

API

The deep_search tool accepts:

• query (string): The research question or topic to investigate
• effort (string): Research effort level - “low”, “medium”, or “high”
  • Low: 1 query, 1 loop, Flash model
  • Medium: 3 queries, 2 loops, Flash model
  • High: 5 queries, 3 loops, Pro model

Return Format

HTTP MCP Server (Development mode):

• answer: Comprehensive research response with citations
• sources: List of source URLs used in research

Stdio MCP Server (Claude Desktop integration):

• file_path: Path to a JSON file containing the research results

The stdio MCP server writes results to a JSON file in the system temp directory to optimize token usage. The JSON file contains the same answer and sources data as the HTTP version, but is accessed via file path rather than returned directly.

Requirements

• Python 3.12+
• GEMINI_API_KEY environment variable

Installation

Install directly using uvx:

uvx install gemini-deepsearch-mcp

Claude Desktop Integration

To use the MCP server with Claude Desktop, add this configuration to your Claude Desktop config file:

macOS

Edit ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "gemini-deepsearch": {
      "command": "uvx",
      "args": ["gemini-deepsearch-mcp"],
      "env": {
        "GEMINI_API_KEY": "your-gemini-api-key-here"
      },
      "timeout": 180000
    }
  }
}

Windows

Edit %APPDATA%/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "gemini-deepsearch": {
      "command": "uvx",
      "args": ["gemini-deepsearch-mcp"],
      "env": {
        "GEMINI_API_KEY": "your-gemini-api-key-here"
      },
      "timeout": 180000
    }
  }
}

Linux