BetterWebSearch MCP: A Keyless, Open-Source Web Search Tool for AI Agents

Developer PhantomPixelDev has released BetterWebSearch MCP, an open-source Model Context Protocol server designed to improve how AI agents perform web research. Addressing common issues like API key requirements and excessive token consumption, the tool provides a local, zero-config solution that defaults to DuckDuckGo. It features a three-tier extraction pipeline—prioritizing fast HTTP requests before falling back to structured data or Playwright for JavaScript-heavy sites. Benchmarks indicate that the tool's 'web_research' function can reduce context usage by approximately 86% compared to traditional extraction methods while maintaining high factual retention. The server includes built-in security features, such as SSRF protection and prompt-injection detection, and is compatible with major MCP clients like Claude Desktop, Cursor, and VS Code. BetterWebSearch is MIT-licensed and available via npm, offering a privacy-focused, infrastructure-light alternative for developers building autonomous AI agents.