Free, no signup
WebMCP checker
Every Shopify store now gives AI agents in the browser a set of WebMCP tools: search the catalog, read a product, change the cart, go to checkout. Paste your store URL to see the tools your page registers, their safety hints, and what fails. What is WebMCP on Shopify? It works for any web page with WebMCP, not only Shopify. It never calls a tool. To grade the tool definitions, use the WebMCP validator.
Public pages only. The page opens in a headless browser that identifies itself as MCPCanary/1.0. A product page and the home page can register different tools; test the page your shoppers start on.
Questions
Does my Shopify store have WebMCP tools?
Yes, if it is a Liquid storefront or uses the Hydrogen developer preview. Shopify turned them on for all of these stores; there is nothing to install. Paste the store URL above to see them.
What does the checker do?
It opens the page in a headless browser, waits until the page has finished registering tools, and lists each tool with its description, input schema, and safety hints (readOnlyHint, consequentialHint, untrustedContentHint). It never calls a tool. Private and local addresses are blocked.
What do the failure codes mean?
WEBMCP_NO_TOOLS: the page loaded but registered no WebMCP tools. TARGET_BLOCKED_BY_WAF: the site answered 401, 403, or 429 (bot protection). TARGET_UNAVAILABLE: the page did not load, answered an error status, or has a certificate that browsers do not trust. TARGET_TIMEOUT: no answer in time.
Can it test the checkout tools?
No. Shopify registers the checkout tools (get_checkout, update_checkout, complete_checkout, navigate_to_storefront) only after proceed_to_checkout with a full cart. The checker reads one page in one state.
Is WebMCP the same as MCP?
No. MCP connects an AI app to a server over HTTP. WebMCP tools live in the open page and run in the shopper's own session. Both describe tools the same way: a name, a description, and a JSON Schema. To test an MCP server, use the MCP server tester.
Are my tool definitions good for agents?
The WebMCP validator grades them from A to F, with Chrome's WebMCP budgets (500 characters per description, 150 per parameter description, 30 per name) and a fix for each finding.
How do I know when the tools break?
Set up a monitor: copy the agent prompt under a result, or see the docs. A monitor emails you when a tool fails, and tells you when the tool list changes.