580 sites implement WebMCP
of 647 tested · measured 2026-08-31
full-index static pass 30 Aug 2026 · 14,351 listed · 11,538 hosts asked · 1,224 could not be asked · 647 carrying the code · runtime-tested 31 Aug 2026
Checks passed across the 580 probed hosts: Markdown negotiation 375 · Markdown twin at a predictable URL 102 · llms.txt 377 · API catalog 542 · MCP server card answers 92.
Check a site for WebMCP — free, anonymous, and it has no effect on any ranking or listing.
What these sites let an agent do
The tools most often registered, by how many of the 580 register each. Real verbs, read from the page at runtime — enumerated, never called.
997 distinct tool names across 580 sites.
How much of this is one plugin
Two instruments read this set and they do not agree, which is the finding rather than a problem. The tool signature — the exact set of names a site registers, treated as a fingerprint — puts 39.5% of the 580 in a cluster with at least one other site, and the largest single cluster is 18.1%: one WordPress plugin, registering the same two tools on every site that installs it. The platform column — which fingerprints hosts — attributes just 0.9%, because it does not fingerprint plugins at all. Where the two disagree the signature is the better instrument, and the size of that gap is why we built it.
580 register tools when the page is executed. 67 carry the registration code and register nothing
Any directory can list sites that claim a standard — the claim is in the source and costs nothing. Separating the two requires testing the behaviour, and dating the result. That is the only thing on this page you cannot get elsewhere.
What WebMCP is
WebMCP lets a page register tools an agent can call, in the page itself, rather than at a separate endpoint. The site declares what can be done and the agent calls it directly — the page becomes the API. Read the WebMCP explainer — it is the authority on the design and we are not trying to be.
How we test it
We execute the page in a headless browser with a modelContext supplied, exactly as an agent browser would, and count what it registers. Tools are ENUMERATED and never called: invoking a stranger's tool could place an order.
A static scan of the source proves a registration EXISTS. It does not prove anything registers. A script that throws, is gated behind consent, or never runs registers nothing — and the first version of our own harness got this wrong in the opposite direction, scoring pages that shipped their own modelContext polyfill while missing conformant implementations that wait for the browser to supply one.
The rate replicates: 89.6% on 31 August against 89.9% on the 30th, five days apart and on an independently re-derived host list.
Sites that implement it
0.7% of these are on teachify, 0.2% of these are on mintlify. The platform supplies the machinery; the domain signals the commitment, and a report says which applies.
All 580 of these sites, through the API: every one, in a fixed order, 25 per call, with the total stated.
WebMCP is one signal among several. The rubric says exactly what we look for and how each tier is reached, and a probe can be requested — the probe decides, never the request.
Method and date
Measured 2026-08-31. full-index static pass 30 Aug 2026 · 14,351 listed · 11,538 hosts asked · 1,224 could not be asked · 647 carrying the code · runtime-tested 31 Aug 2026 Every figure on this page carries the population it was measured against; a count without a denominator is not a measurement. How the index is built · All standards.