knowngood.sh
menu

MCP adoption: 48,822 sites, November 2026

48,822 of the 49,613 sites we tested implement Model Context Protocol — the index as of November 2026.

49,613 hosts strict-re-graded, of 57,273 listed as of 4 Nov 2026: 57,266 in the full pass of 2 Oct 2026, and 7 probed since

Checks passed across the 48,822 probed hosts: Markdown negotiation 4,751 · Markdown twin at a predictable URL 4,359 · llms.txt 47,896 · API catalog 1,438 · MCP endpoint answers tools/list 48,822.

What these sites let an agent do

The tools most often registered, by how many of the 48,822 register each. Real verbs, read from the page at runtime — enumerated, never called.

get_product41,414
create_checkout41,385
search_catalog41,367
lookup_catalog41,356
get_checkout41,346
complete_checkout41,342
cancel_checkout41,341
update_checkout41,341
get_cart31,580
create_cart31,579
update_cart31,577
cancel_cart31,576

997 distinct tool names across 48,822 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 87.5% of the 48,822 in a cluster with at least one other site, and the largest single cluster is 64.6%: one WordPress plugin, registering the same two tools on every site that installs it. The platform column — which fingerprints hosts — attributes just 93.8%, 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.

What the number separates

48,822 pass when we test the behaviour. 791 do not.

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 Model Context Protocol is

MCP is the protocol an agent uses to discover and call a server's tools. On the open web a site advertises one through a server card, and an agent connects to the endpoint the card names. Read the MCP specification — it is the authority on the design and we are not trying to be.

How we test it

We fetch the server card and then CONNECT to the endpoint it advertises, and count it only where the server answered tools/list — or challenged us for auth, which is an answer.

A server card is a file. It costs nothing to publish and proves nothing about whether anything is listening. The gap between cards published and servers that answer is the single largest gap we measure.

Sites that implement it

64.6% of these are on shopify, 20% of these are on wix, 9.2% of these are on mintlify. The platform supplies the machinery; the domain signals the commitment, and a report says which applies.

✓Business.Digital business.digital
External digital department for medium-sized companies in Germany, offering integrated web development, e-commerce, ERP, automation, AI, and software solutions with consulting and ongoing support since 2006.
shopware
✓David Bueno Vallejo davidbuenov.com
Spanish professor and researcher David Bueno Vallejo offers technology consulting, AI training diagnostics, and publishes projects, blog posts, conferences, and educational resources on his personal website.
✓Delx delx.ai
Delx is an independent AI agent lab researching continuity, recovery, verifiable work, and governed exchange for autonomous systems, providing protocols, tools, and research artifacts for AI agent lifecycle and operational resilience.
✓EFROS efros.com
US cybersecurity-first MSP and MSSP offering managed IT, 24/7 SOC, system integration, and AI Governance for regulated industries including healthcare, financial services, manufacturing, logistics, and legal, serving nationwide from Chicago area.
✓Every every.to
Subscription service offering AI-focused ideas, apps, training, and editorial columns from AI practitioners, including AI-powered software products and livestreamed camps, founded in 2020 and cited by major publications.
✓Faheem faheem-ai.app
Arabic AI-powered learning assistant aligned with the Egyptian curriculum, offering adaptive learning conversations, interactive tests with analytics, study booklets, and voice-first Q&A for Arabic-speaking students, parents, and educational institutions in Egypt.
✓JobsPipe jobspipe.dev
API platform providing live, deduplicated, normalized job postings from 30+ ATS and job boards, plus company technology stack scanning and labor market insights, serving global job data consumers and AI agents.
✓NVDC nvdc.ai
Decentralized, verifiable, OpenAI-compatible GPU inference network with NVTrust attestation, offering paid GPU inference, attestation verification, USD payments via Stripe, and an experimental relay board for deferred jobs.
✓Cameron Rye Portfolio rye.dev
Cameron Rye's personal portfolio and technical blog featuring software engineering projects, AI integration tools, MCP servers, and developer resources, with blog posts, project details, resume, and contact options accessible online.
✓VertoDigital vertodigital.com
B2B digital marketing agency specializing in pipeline-driven marketing for technology companies with $10M–$500M ARR, offering inbound and outbound pipeline growth, pipeline intelligence, and a free Pipeline Readiness Assessment. Based in Sofia and Boston, serving US, UK, and EU clients.
✓a2b Logistics www.a2b.sa
B2B logistics company in Riyadh, Saudi Arabia offering trucking, warehousing support, customs-clearance coordination, supply-chain solutions, and fleet services with kingdom-wide coverage and 24/7 support.
✓Growth Academy Global www.growthacademy.global
Growth Academy Global teaches established service business owners AI-era client acquisition via LinkedIn category ownership, LinkedIn Live, AI search visibility, and Codex agent workflows. Founded by Shanee Moret, it targets B2B owners selling $20K+ services.

All 48,822 of these sites, through the API: every one, in a fixed order, 25 per call, with the total stated.

Own a site?

Model Context Protocol 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 2 Oct 2026. 49,613 hosts strict-re-graded, of 57,273 listed as of 4 Nov 2026: 57,266 in the full pass of 2 Oct 2026, and 7 probed since 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.