There is nothing to authenticate and nothing to set up, which makes this the extraction half of a scraping loop rather than a scraper — xpathwithurl retrieves and applies the expression in the same call, and xpath does the same for content already in hand. mimeType is the one real control, deciding how the content is parsed before the expression runs against it. What is absent is everything around the extraction: no pagination, no session, no rendering step. The input is a document and the output is what the XPath selects, which is the right size of tool when something else in the chain is already doing the fetching.
An XPath query server of 2 tools: one that queries XML content passed inline, one that fetches a URL and queries what comes back, both taking the expression and a mimeType.
- Querying a document you already hold: xpath takes the xml content, the query and a mimeType, and returns what the expression selects.
- Querying one you do not: xpathwithurl fetches the url and applies the same query and mimeType to the response, so retrieval and extraction happen in a single call.
No credentials at all. For xpath, the XML content itself; for xpathwithurl, a url the machine running the server can reach.
One command — npx @thirdstrandstudio/mcp-xpath