scrapegraphlisted

LLM-powered web scraping — extract structured data from a URL/page using a plain-language prompt, or answer a question by searching the web. Use when the user wants to "scrape", "extract data from", "pull info from a website", "get structured JSON from a page", "crawl and extract", or search the web and extract a structured answer. Backed by ScrapeGraphAI + OpenRouter (cheap models).
iamnolanhu/claude-dotfiles · ★ 1 · AI & Automation · score 62

Install: claude install-skill iamnolanhu/claude-dotfiles

# scrapegraph LLM-driven scraping (ScrapeGraphAI + Playwright, OpenRouter-backed). Give it a URL and a natural-language prompt; it returns structured JSON. Prefer this over raw HTML fetching when the user wants _specific fields_ extracted ("titles and prices", "founders and socials"). > **Setup required.** This skill assumes a [ScrapeGraphAI](https://github.com/ScrapeGraphAI/Scrapegraph-ai) > MCP server and/or its `sgai` CLI wrapper are available. Either run a hosted MCP instance and > point your MCP config at its URL, or install the CLI locally. Replace the placeholders below > (`<scrapegraph-mcp-host>`, `<scrapegraph-mcp-dir>`) with your own. Without one of these, use > `crawl4ai` or `firecrawl` instead. **Default = a hosted ScrapeGraphAI MCP instance** (the `scrapegraph` MCP, URL set in your MCP config). Use its MCP tools (`smart_scraper`, `search_scraper`, `crawl`, `scrape_many`, `omni_scraper`, `markdownify`) when available — no local browser/LLM needed. **Fall back to the local CLI** (below) only when the MCP isn't loaded or you're offline. ## When to use - "Scrape X and extract Y" / "pull the prices from this page" / "get structured data from <url>" - "What does this company do? (from their site)" — single-page extraction - "Search the web for X and give me a structured answer" — uses web search + extraction For a plain page dump (no field extraction), prefer crawl4ai. For known library docs, use context7. ## How to run **Primary — hosted MCP:** if the `scrapeg