Web Metadata, OpenGraph & Contact Extractor

What Web Metadata, OpenGraph & Contact Extractor MCP server does

Web Metadata, OpenGraph & Contact Extractor MCP server turns a public URL into structured website information. The documented output can include page title and description, OpenGraph fields, canonical and language data, authorship, robots directives, hreflang tags, headings, image counts, contact details, social profiles, detected technologies, structured data, product attributes, links, feeds, security headers, and an on-page SEO score.

It also produces cleaned Markdown from page content, with word count and reading-time information. That output is intended for use in AI agents, retrieval workflows, and applications that need page text without handling the original HTML directly.

How it works

The API fetches a public webpage, parses its HTML, and extracts information from metadata, visible page elements, links, and structured encodings. Product details can be read from JSON-LD, OpenGraph product fields, and Schema.org Microdata. The response reports per-field product confidence and can identify disagreements between those sources. A quality object summarizes the result with a score, sources used, rendering status, and warnings.

The extractor detects more than 40 technologies across CMS, ecommerce, frontend, analytics, payments, and hosting categories. It also identifies public email addresses and phone numbers after DOM cleanup. Phone values may include normalized details such as E.164 formatting, ISO country code, and possible or valid flags when a country code is present; numbers without one are not guessed.

The documented production endpoint is /api/v1/extract, called with a URL and RapidAPI headers. Separate documented endpoints provide Markdown output and a security-focused response. A no-key demo route exists for the live demo, but the README describes it as rate-limited per IP rather than a stable integration endpoint.

Tools and capabilities

The documented capabilities include:

• Extracting SEO and OpenGraph metadata, canonical URLs, favicons, language, authors, robots directives, and hreflang values.
• Finding public email addresses, phone numbers, and profiles across listed social platforms.
• Detecting technologies such as WordPress, Shopify, WooCommerce, React, Next.js, Vue, Angular, Svelte, TailwindCSS, Stripe, and GA4.
• Parsing Schema.org data and product name, price, currency, availability, brand, rating, and review count.
• Running a 14-point SEO audit with a percentage score, checks, severities, evidence, and warnings.
• Classifying up to 100 page links as internal or external.
• Discovering RSS and Atom feeds.
• Auditing HSTS, CSP, X-Frame-Options, X-Content-Type-Options, Referrer-Policy, and Permissions-Policy with graded results.
• Returning clean Markdown for AI and RAG workflows.

Setup and configuration

The source material documents access through RapidAPI. Production requests use an X-RapidAPI-Key header, an X-RapidAPI-Host header, and a URL query parameter. The README provides Python and cURL examples for the extraction endpoint and identifies the RapidAPI API as the integration route.

No MCP-specific configuration file, command, package installation procedure, transport, or environment variable is documented in the supplied material. The live demo can be tried without an API key, but its demo endpoint should not be treated as the documented stable API for application integration.

Limitations and notes

Results describe what can be found in the fetched public page and its structured data. Product-source agreement indicates internal consistency among the encodings found on that site; it is not independent verification because several encodings may originate from the same product record. Phone normalization does not infer a country for numbers lacking one.

The extractor applies an IP-pinned anti-SSRF check, validates private, loopback, and cloud-metadata ranges, and pins connections to the resolved IP with TLS SNI. Adaptive download limits are used for detected SPA frameworks, with larger limits for those pages. The README also describes an in-memory cache and rate limiting for the demo route. The supplied documentation does not establish a specific MCP client compatibility matrix.