one-search-mcp and google-ai-mode-mcp
These two tools are competitors, as both offer an MCP server designed to provide web search capabilities, with tool A supporting various search engines like SearXNG and Tavily, and tool B specifically focusing on free Google AI Mode search with citations.
About one-search-mcp
yokingma/one-search-mcp
🚀 OneSearch MCP Server: Web Search & Scraper & Extract, Support agent-browser, SearXNG, Tavily, DuckDuckGo, Bing, etc.
Implements Model Context Protocol (MCP) server with pluggable search backends (SearXNG, Tavily, DuckDuckGo, Bing, Google, etc.) and exposes four tools—`one_search`, `one_scrape`, `one_map`, `one_extract`—for web search, scraping, and structured data extraction. Uses local browser automation via `agent-browser` for privacy-preserving searches and scraping without external API dependencies, with automatic Chromium detection across Chrome, Edge, and Canary installations. Integrates with Claude Desktop, Cursor, and Windsurf through standard MCP configuration files.
About google-ai-mode-mcp
PleasePrompto/google-ai-mode-mcp
MCP server for free Google AI Mode search with citations. Query optimization, CAPTCHA handling, multi-agent support. Works with Claude Code, Cursor, Cline, Windsurf.
Leverages Puppeteer with stealth techniques and persistent browser profiles to automate Google AI Mode searches, using a 4-stage completion detection system (SVG thumbs-up → aria-label → text → timeout) that achieves 87% faster results across multiple languages. Extracts AI-synthesized answers and citations via 17 language-agnostic selectors, then converts to Markdown with inline references—eliminating the token cost of manual multi-page research while preserving source attribution. Exposes a single MCP tool callable from any agent (Claude, Cursor, Cline, Windsurf, Zed) with optional file persistence and automatic CAPTCHA handling via visible browser fallback.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work