Does your site work with AI agents?
Test WebMCP loads your page in a real browser, finds the WebMCP tools it registers, and calls them to check that they work. You get a score out of 100 and a fix for each problem.
One free scan per day. A scan takes up to a minute.
How it's different
- 01
Checks both API names
Browsers are moving from navigator.modelContext to document.modelContext. We check both and tell you which one your site uses.
- 02
Calls your tools
We call your read-only tools with test inputs to check that they return real results. A tool that registers but returns an error is marked as broken.
- 03
Tests the live page
We load your page in a real browser and record what it registers, including tools added after the page finishes loading. We don't guess from your source code.
What gets checked
100 ptsTools registered on either API name or through forms, tools that return results, and registration within 5 seconds.
Valid schemas with described properties, descriptions that say what comes back, clear names, annotations, and reasonably sized results.
llms.txt, an MCP server card, an agent skills index, JSON-LD, and a robots.txt that allows AI crawlers.
HTTPS, a WebMCP origin trial token, and no errors that stop tools from registering.
See how it works for the full scoring breakdown.