home / mcp / web search mcp server

Web Search MCP Server

Provides multi-engine web search with full-page extraction for local LLM workflows and quick summaries.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "mrkrsl-web-search-mcp": {
      "command": "node",
      "args": [
        "/path/to/web-search-mcp/dist/index.js"
      ],
      "env": {
        "MAX_BROWSERS": "3",
        "BROWSER_TYPES": "chromium,firefox",
        "DEFAULT_TIMEOUT": "6000",
        "BROWSER_HEADLESS": "true",
        "MAX_CONTENT_LENGTH": "10000",
        "BROWSER_FALLBACK_THRESHOLD": "3"
      }
    }
  }
}

This MCP server provides multi-engine web search and content extraction capabilities that you can use with local LLMs. It combines browser-based search with HTTP fallbacks to return rich results and full page content from top search results, all orchestrated as a reusable MCP endpoint.

How to use

You connect to this MCP server through an MCP client and invoke the three available tools to perform web search tasks. Use full-web-search for comprehensive results with extracted page content, get-web-search-summaries for quick snippets, and get-single-web-page-content for extracting content from a single URL. The server handles multiple engines, concurrent content extraction, and automatic fallbacks to keep results responsive.

Practical usage patterns include research workflows where you want a thorough collection of sources with full-page content, or rapid skimming where you only need search result summaries. For known pages, you can fetch main content from a single URL without traversing search results.

How to install

Prerequisites: Node.js 18.0.0 or higher and npm 8.0.0 or higher.

1. Download the latest release ZIP file and extract it to a chosen directory on your system, for example ~/mcp-servers/web-search-mcp/.

2. Open a terminal in the extracted folder and install dependencies, including Playwright browsers.

npm install
npx playwright install
npm run build

3. Create the MCP configuration to point to the built file. The final runtime command runs the local server binary with the generated dist index.

{
  "mcpServers": {
    "web-search": {
      "command": "node",
      "args": ["/path/to/extracted/web-search-mcp/dist/index.js"]
    }
  }
}

Environment and configuration notes

Environment variables help tailor performance and behavior. The following options are shown for customizing the server runtime. You can set them in your mcp.json under the corresponding server’s env block or in your runtime environment.

{
  "mcpServers": {
    "web-search": {
      "command": "node",
      "args": ["/path/to/web-search-mcp/dist/index.js"],
      "env": {
        "MAX_CONTENT_LENGTH": "10000",
        "BROWSER_HEADLESS": "true",
        "MAX_BROWSERS": "3",
        "BROWSER_FALLBACK_THRESHOLD": "3"
      }
    }
  }
}

LibreChat and Docker notes

If you are using LibreChat, include the MCP server in the librechat configuration and mount your local MCP directory into the container as needed.

Troubleshooting

Slow responses can occur if browser automation or network conditions are constrained. Try reducing timeouts, limiting the number of concurrent browsers, or increasing system resources. If you encounter browser-related errors, ensure Playwright browsers are installed with npx playwright install.

If you see content length issues, adjust MAX_CONTENT_LENGTH to larger values and re-run the server with the updated environment.

Notes on compatibility and tools

This server exposes three MCP tools for web search tasks: full-web-search, get-web-search-summaries, and get-single-web-page-content. Each tool has specific inputs and behaviors as described in their respective sections.

Available tools

full-web-search

Comprehensive web search tool that uses multiple engines, fetches full content from results, and returns structured data with extracted content.

get-web-search-summaries

Lightweight search tool that returns only result snippets/descriptions without following links for full content extraction.

get-single-web-page-content

Utility tool to extract the main content from a single URL, removing navigation and ads.