MCP Web Research Server

MCP Web Research Server is an MCP server that provides AI assistants with live web search, site navigation, and page extraction tools. It connects Model Context Protocol clients like Claude Desktop to the internet via automated browser automation backed by Playwright. Software developers, researchers, and general users employ this server to overcome model training cutoffs and retrieve verified real-time information directly inside conversations. The server exposes capabilities to run Google queries, extract structured page content, and capture visual screenshots of visited websites. In addition to tools, it exposes an agentic research prompt designed to guide multi-step investigations, maintain search history, and track visited URLs within a structured research session. Captured screenshots and session logs are exposed back to the client as MCP resources, allowing models to review source documents and cite referenced websites with accurate source links during active discussions.

Category: Browser & Web Automation

Tags: information-retrieval, real-time, web scraping, web search

Visit MCP Web Research Server

How to install and configure MCP Web Research Server

1. Ensure you have Node.js version 18 or higher installed along with npm and the Claude Desktop application. 2. Open your Claude Desktop configuration file located at ~/Library/Application Support/Claude/claude_desktop_config.json on macOS. 3. Add the server definition under the mcpServers object: json { "mcpServers": { "webresearch": { "command": "npx", "args": ["-y", "@mzxrai/mcp-webresearch@latest"] } } } 4. Save the file and restart Claude Desktop to load the new tools and prompts.

What you can do with MCP Web Research Server

• Querying Google to gather breaking news and recent topic developments directly within an ongoing Claude Desktop chat session. - Navigating to specific URLs to extract full text content and documentation for context-grounded analysis and summarization. - Taking webpage screenshots using automated Playwright capture to inspect visual layouts and review referenced pages as MCP resources. - Executing the built-in agentic-research prompt to run iterative, multi-step online investigations with documented source citations and session tracking.

Key facts

• https://github.com/mzxrai/mcp-webresearch
• Browser & Web Automation, Web Search & Research
• information-retrieval, real-time, web scraping, web search

Part of MCP Servers

How do I install MCP Web Research Server?

You install the server by configuring Claude Desktop to run its package via npx. Add the webresearch entry to your claude_desktop_config.json pointing the command to npx with the arguments -y and @mzxrai/mcp-webresearch@latest, then restart Claude Desktop.

What tools are included in MCP Web Research Server?

The server includes three primary tools: search_google for executing Google search queries, visit_page for loading web addresses and extracting content with an optional screenshot parameter, and take_screenshot for capturing the current browser view.

What is the agentic-research prompt?

It is a preconfigured prompt included with the server that guides Claude to conduct structured research. It directs the model to start with broad inquiries, prioritize authoritative domains, iterate on findings, and supply URL citations for all retrieved facts.

What license does MCP Web Research Server use?

MCP Web Research Server is open-source software released under the MIT license by author mzxrai. You can view the code, report issues, and inspect development scripts directly on its official GitHub repository.

Which platforms are supported by MCP Web Research Server?

The server requires Node.js version 18 or later and Playwright. The project repository lists macOS as a verified platform, while Linux support has not yet been formally verified.