Crawly

Browser-backed web search and content retrieval using Playwright, supporting DuckDuckGo, Google, and Yandex with SSRF protection.

This MCP server enables local LLMs to search the web and fetch live page content without relying on remote models. It provides three tools: search (browser-backed queries on DuckDuckGo, Google, or Yandex returning up to 5 organic results), fetch (browser-rendered content from 1–5 URLs in HTML or extracted text), and page_search (content search on a single page). Uses Playwright/Patchright with persistent profiles and fingerprint patches for bot-detection evasion. Available as a CLI tool, stdio MCP server, or HTTP endpoint.