mcp-websearch

This MCP web search server provides a hybrid, multi-engine web search capability with intelligent content extraction for AI-ready results. It exposes a set of MCP tools that let you perform fast goquery-based searches, fetch and extract content from result pages using chromedp, perform cross-engine searches, and generate AI-friendly summaries. The available tools include: websearch_basic for fast single-engine results, websearch_with_content which extracts page content, websearch_multi_engine to search across Bing, Brave, and DuckDuckGo with content extraction, and websearch_ai_summary to return AI-ready aggregated content. To use these, configure your MCP client (such as Claude Desktop) to point at the websearch MCP server entry (e.g., in your claude_desktop_config.json) and call the desired tool with its required parameters like query and max_results. The server prioritizes DuckDuckGo, then Bing, then Brave, and gracefully falls back if any engine fails, returning structured results suitable for AI processing.

Prerequisites:

• Go 1.21 or higher
• A supported environment with network access and, for content extraction, a Chrome/Chromium browser installed

Install via go install (recommended):

go install github.com/liliang-cn/mcp-websearch-server@latest

Build from source:

git clone https://github.com/liliang-cn/mcp-websearch-server
cd mcp-websearch-server
go build -o mcp-websearch-server

Run the server directly (after installation or built binary is available):

# Run the server (stdio mode by default)
mcp-websearch-server

# Optional: show help
mcp-websearch-server --help

Examples:

# Run tests (if you have tests in the repo)
go test ./...

Notes and tips:

• Ensure Chrome/Chromium is installed for chromedp-based content extraction. The server uses intelligent content extraction with article detection and content cleaning.
• If you install via go install, ensure your GOPATH/bin or ~/go/bin is in your PATH so the mcp-websearch-server binary can be invoked from any shell.
• The server supports concurrent extraction with rate limiting; if you deploy at scale, monitor and adjust concurrency as needed.
• For Claude integration, configure the MCP tools in Claude Desktop and, if applicable, copy the provided SKILL.md for Claude Code compatibility.
• If you need to customize engines or selectors, you can modify the search and extraction logic under the search/ and extraction/ directories and re-build the binary.