Public data index › MCP Ecosystem Continuity Record › Methodology

Monthly MCP Ecosystem Continuity Record methodology

The reader accepted 5,346 sealed Official MCP Registry observations across 45 dated captures. An unsuccessful newest source receipt makes every page unavailable; it never becomes a plausible empty catalogue.

This methodology keeps each source claim bounded. This public page measures the Official Registry lane. The paid record is a request-only, remote/manual hybrid: it adds an independently validated owned mcp.so sitemap-aggregate pack and up to 25 requested Official Registry identities. No checkout or automatic fulfillment is active, and a request schedules no recurring delivery.

Two reuse regimes

The Official MCP Registry terms dedicate Registry Data to the public domain under CC0. Named Official identity details therefore reproduce bounded CC0 fields. The mcp.so lane is different: our owned sealed observations emit only sitemap URL, slug, and lastmod facts. No listing-page prose is republished. A sitemap lastmod field change is not proof that listing content changed.

Public Official-lane privacy boundary

The public-page SQL reader selects only exact registry identity, snapshot date, a bounded detail projection, content hash, and collection time. It never selects raw JSON, wallet or client addresses, endpoint headers, descriptions, repository handles, price observations, on-chain reputation events, or mcp.so records. The separate history-pack builder validates exact source_id=mcp_so rows and strips them to the three sitemap facts before any buyer artifact is accepted.

Separate comparison rules

Versions are compared only when the same exact registry name appears on two captured dates. We do not join by a similar name, repository, endpoint, or operator. A publication event uses the Registry's own published date. We publish volume and version history, never a score or a claim that one server is safer than another. The mcp.so aggregate lane compares exact slugs across consecutive held sitemap observations. There is no fuzzy cross-source join; every missing comparison stays unavailable.

Capture boundary

This is not a complete registry census. The Official Registry API returns 100 records per page. Newest collections can seal a cursor walk only after reaching a terminal page; historical captures predate that terminal-pagination proof. Absence is therefore not evidence that a server is unregistered. Names, versions, statuses, and dates reproduce Registry Data; they are not security reviews or endorsements. The captured window is 2026-07-22 through 2026-09-07. The Registry remains authoritative.

Source receipt

The newest successful response was 60,340 bytes, collected 2026-09-07T19:52:41+00:00. Source hash: 646e8687f8c903c70180ed246b8c4af98e87eb5344ec1db1c12bf78c6c4ca3d2. Collection batch seal: 06f030f8d710e4128dc7acc8ddea90b42a7e6a51b9f68e3ed28b4c31cbacd589.

Official MCP Registry API query · Registry terms and CC0 dedication · Return to the continuity-report offer.