Web Search

You can run a high-performance web search MCP server that converts search results from multiple engines into structured JSON or Markdown outputs. It supports multi-layer depth crawling, anti-detection browsing, and automatic engine fallback, making it easy to build reliable search workflows with customizable output formats.

How to use

You interact with the MCP server through a client that communicates with the MCP API over HTTP or through a local process. To perform a search, you issue a request with your query, choose a search engine, set the crawl depth, and specify the output format. The server handles SERP parsing, page content extraction, and, if requested, external link crawling, returning results in JSON or Markdown along with metadata about the search.”

How to install

Prerequisites: Python is required on your system. You can also run the server via Docker if you prefer containerized deployment.

• Clone the project and enter the folder
• Install development dependencies using Python’s package installer
• Install the Camoufox browser helper through its fetch command
• Optionally prepare a containerized deployment using Docker

Configuration and usage notes

The MCP server exposes an HTTP endpoint for remote control and can be started in development or production mode. You can customize browser behavior through environment variables to adjust concurrency, proxy settings, operating system fingerprints, fonts, and other evasive browsing features. These settings influence how the embedded browser instance mimics real users and can improve resilience against anti-bot measures.

Troubleshooting and tips

If you encounter verification challenges from a search engine, rely on the built-in engine fallback mechanism. The system automatically tries the primary engine and falls back to secondary engines as needed. Check health endpoints to confirm the MCP server is ready and the browser pool is initialized.

Tools and endpoints you can use

The MCP server provides several tools to perform searches and fetch content. You can search with different engines, fetch page content, and list available engines. Use these tools to build flexible search workflows.

Available tools

web_search

Execute a search using a chosen engine with a specified depth and return results in JSON or Markdown.

get_page_content

Fetch and extract the main content from a given URL after a search result.

list_search_engines

List the supported search engines available for MCP queries.

initialize

Initialize an MCP session with client metadata and capabilities.