Web Search MCP

Web Search MCP is an MCP server that provides search engine querying and webpage scraping tools directly to LLM clients. It connects to Google, DuckDuckGo, and Bing via a headless browser setup, using an automated fallback hierarchy to retrieve live search results even when an individual search engine enforces bot blocks. Developers, researchers, and AI assistant users employ this server to gather current online data without requiring paid search API subscriptions. The server can render pages using standard Selenium with Google Chrome or through a standalone binary called Obscura, processing JavaScript-heavy sites and extracting clean HTML parsed through BeautifulSoup. In addition to returning ranked organic results with resolved destination URLs, titles, and snippets, Web Search MCP allows language models to fetch and inspect full text content directly from target web pages, monitor engine availability statuses, and reset blocking trackers during active agent sessions.

Category: Browser & Web Automation

Tags: google, headless chrome, search, web scraping

Visit Web Search MCP

How to install and configure Web Search MCP

1. Ensure Python 3.10+ is installed alongside Google Chrome for the default Selenium backend, or install the standalone Obscura binary. 2. Install the package using uv or pip: bash uv sync # Or using pip pip install -e . 3. Add the server configuration to your Claude Desktop configuration file: json { "mcpServers": { "web-search-mcp": { "command": "uvx", "args": ["git+https://github.com/pranavms13/web-search-mcp"] } } } 4. Alternatively, start the server directly via command line by executing python main.py or web-search-mcp.

What you can do with Web Search MCP

• Querying multiple search engines sequentially with automatic failover to DuckDuckGo or Bing whenever Google blocks automation requests. - Fetching up to 100 search result URLs while unwrapping redirect links to reveal the true destination addresses. - Extracting clean, readable plain text from target webpages up to 20,000 characters for LLM document summarization. - Checking search engine health and engine block statuses using the included status and reset reporting tools.

Key facts

• Open Source
• https://github.com/pranavms13/web-search-mcp
• Browser & Web Automation, Web Search & Research
• google, headless chrome, search, web scraping

Part of MCP Servers

What can Web Search MCP do?

Web Search MCP enables AI assistants to search Google, DuckDuckGo, and Bing with automatic fallback if an engine is blocked. It unwraps redirect URLs, retrieves titles and snippets, and extracts full text content from target webpages using headless rendering.

Which MCP clients work with Web Search MCP?

Web Search MCP works with Claude Desktop, Cursor, and any other Model Context Protocol client capable of spawning local process servers via uvx or direct Python command execution.

How do I install Web Search MCP?

You can install dependencies using uv sync or pip install -e . in your local environment. To run it inside Claude Desktop, configure the uvx command pointing to git+https://github.com/pranavms13/web-search-mcp under your mcpServers settings.

Is Web Search MCP open source?

Yes, Web Search MCP is open-source software licensed under the MIT License. The complete source code is available in the pranavms13/web-search-mcp GitHub repository.

Does Web Search MCP require Google Chrome?

The default backend uses Selenium and webdriver-manager, which requires Google Chrome to be installed. However, you can alternatively use the lightweight Obscura backend by configuring the BROWSER_BACKEND environment variable to obscura without installing Chrome.