Summary
A read-the-web set: fetch, extract, discover — no browser, no key.
fetch_webpage paired with extract_structured_data does most of the work: get the page as text, then pull the table or the prices out of it rather than asking the model to read raw HTML. generate_sitemap_urls is what scales that up — point it at a domain and you get back the URL list to work through.
What it is
The web server of a three-server toolkit. It fetches a URL as clean text, extracts structured values from HTML, discovers sitemaps and RSS feeds, converts markdown tables to JSON and summarises text.
What you get
- Clean text pulled out of any URL, with the page furniture stripped
- Tables, prices, emails and dates lifted out of HTML as structured data
- A site's sitemaps and RSS feeds, discovered from the domain
- Markdown tables converted to JSON
- Key points and entities extracted from a block of text
Requirements
Node.js. Run npm install in the web-tools-mcp directory and start it with node index.js, then point your client at it. Apache licensed.
Setup effort
One command — cd web-tools-mcp && npm install && node index.js