You can run and use a dedicated MCP server to perform parallel Google searches with multiple keywords, manage browser-based searches, and receive structured results in JSON. This guide walks you through using the G-Search MCP to run searches, install and start the server, and configure it for local or client deployments.

How to use

Start the server using a simple one-shot command and your MCP client can send search requests with multiple keywords. The server leverages a Playwright-controlled browser to perform parallel searches, handling verification prompts when needed, and returning results in a structured JSON format that you can easily consume in your workflows.

How to install

# Prerequisites
- Node.js 18 or higher
- npm or yarn

# Install from source (recommended steps)
git clone https://github.com/jae-jae/g-search-mcp.git
cd g-search-mcp
npm install

# Install Playwright browser (Chromium)
npm run install-browser

# Build the server
npm run build

Configuration and usage notes

Configure the MCP server in your Claude Desktop environment using the following example. This registers the G-Search MCP so your client can start it and issue search requests.

{
  "mcpServers": {
    "g_search": {
      "command": "npx",
      "args": ["-y", "g-search-mcp"]
    }
  }
}

Usage tips and examples

Run the server from the command line and use your MCP client to submit queries. You can enable a debug browser window to observe the search process during development.

npx -y g-search-mcp

Debug mode to show the browser window during operations

npx -y g-search-mcp --debug

Notes on parameters you can influence via the client

The server supports multiple configurable parameters for each search, such as the list of queries, the maximum number of results per query, and the page load timeout. You can adjust locale and debugging settings as needed to tailor results to your region and workflow.

Security and reliability considerations

The server is designed to simulate realistic browsing patterns to reduce detection by search engines. It includes automatic verification handling to show a visible browser mode when CAPTCHA verification is required, allowing you to complete verification manually.

Examples of typical workflows

• Execute searches for multiple keywords in a single request and receive a structured JSON response containing title, link, and snippet for each result.

Available tools

search

Executes Google searches with multiple keywords, using a Playwright-controlled browser to perform parallel queries and returning results in a structured JSON format with titles, links, and snippets.