Add mcpbeat to Resources - #4610
Conversation
|
This is valuable because it measures observed state rather than merely listed state. The 15-minute initialize checks and tool snapshots expose an important distinction for the MCP ecosystem: registry discovery ≠ current operational state. A registry can establish that a server has been published and where it is expected to exist. Runtime observation can establish that something responded at a particular time. Neither, by itself, establishes the full authority/evidence chain a relying party may need: who operates it, which deployment/version is authoritative, what changed, and whether that state has subsequently been withdrawn or superseded. I think keeping those layers explicit will become increasingly useful as MCP moves into enterprise use: Discovery → Observed state → Operator/authority → Evidence → Freshness. That is very close to the resolver-first model we are building with ECZ-ID: not replacing registries or monitoring, but making independently resolvable identity, authority and current evidence available alongside them. — Adam Al-Zayed |
2 participants
mcpbeat monitors the MCP ecosystem rather than listing it: every remote address in the registry receives a real JSON-RPC
initializehandshake every 15 minutes, and the tool list of each responding server is recorded together with the date it was read.Current numbers, measured today:
Adoption comes from actual npm and PyPI download counts rather than GitHub stars, and Agent Skills are deduplicated by SHA-256 of their content, so a skill repackaged into forty repositories counts once.
Everything is published as CSV under CC BY 4.0 at https://mcpbeat.com/data/ — no key, no signup, so any number on the site can be verified against the raw rows.