schemastats
Validate
WebMCP

WebMCP for Schema.org

This site exposes its Schema.org data to AI agents as callable tools, over the Model Context Protocol. No key, no sign-up, no scraping.

What WebMCP is

WebMCP is an emerging browser standard that lets a website hand an AI agent a set of tools instead of a page built for human eyes. Rather than guessing which button to click or parsing HTML, an agent visiting schemastats.com finds a declared list of operations it can call directly, with typed arguments and structured answers.

Here the tools are served by an MCP server on the same domain, and Cloudflare's edge bridges them into the page so that browsers implementing WebMCP register them automatically. The same server also answers any MCP client that connects to it directly.

Tools exposed

Four read-only tools cover the whole dataset: the Schema.org vocabulary and, for every type and property, how many domains use it month by month.

  • search_schema_terms

    Search types and properties by partial name and get each one's adoption bucket. The usual starting point when only part of a term name is known.

    Arguments: query (required), kind, limit

  • get_schema_term

    The official definition of one term — description, hierarchy, expected value types, deprecation — together with its adoption history and, for types, the properties it declares.

    Arguments: name (required), kind

  • get_adoption_overview

    Aggregate figures for one month: how many types and properties were observed and how they spread across domain-count buckets.

    Arguments: month

  • get_adoption_trends

    Terms that moved to a higher or lower domain-count bucket compared with the previous month.

    Arguments: month, direction, limit

How to use it

From a browser agent

Nothing to install. Browsers that implement WebMCP (Chrome 146+, Edge 147+) pick up the tools from the bridge script Cloudflare injects into every page, and register them on document.modelContext for the agent running in the tab.

<script type="module"
  src="/.webmcp/bridge.js"
  data-packs="c2pa,mcp-server-client"></script>

From any MCP client

The server speaks Streamable HTTP and holds no session, so it works with any MCP client that accepts a URL. Point it at this endpoint — POST only, no authentication:

https://www.schemastats.com/mcp

Example call

curl -sS https://www.schemastats.com/mcp \
  -H 'Content-Type: application/json' \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "tools/call",
    "params": {
      "name": "get_schema_term",
      "arguments": { "name": "Product" }
    }
  }'

What it does and does not do

  • Read-only. There is nothing to create, change or delete: every tool answers a question about public data.
  • No personal data. The tools reach the Schema.org vocabulary and Google's published usage statistics — no accounts, no logs, no visitor data.
  • No crawling on your behalf. The validator's crawler is deliberately not exposed as a tool, so an agent cannot use this endpoint to fetch third-party pages.
  • Rate limited to 60 requests per minute per IP address. Beyond that the endpoint answers 429.
  • Open to any origin and free to use. Figures are buckets, not exact domain counts, because that is how the source dataset is published.

See also