The MCP spec is the catalog's #3 inbound substrate

The T2-1 runtime-dependency graph captures 212 edges of the form 'X depends on Y at runtime' (not 'X cites Y'). Counted by inbound edges — i.e. how many catalog entries treat the target as a runtime substrate — the top five are: Claude (62), GPT (52), the Model Context Protocol specification (34), Gemini (22), Qwen (16).

A protocol specification ranking third — ahead of every foundation model except Claude and GPT — is the v6 finding that wasn't visible in the citation-graph hubs. Citations measure scholarly influence; runtime dependencies measure adoption-as-substrate. The MCP spec is being adopted as substrate, not just as a wire format.

This matters because it changes the competitive frame. If MCP is a substrate, then 'MCP server X' systems are not protocol-of-the-week experiments — they're load-bearing components of the agent stack with the same dependency surface as 'agent depends on GPT'. The implication is that MCP-server entries deserve substrate-grade ranking treatment (they currently rank lower than they should on adoption signal), and the spec itself is closer to a foundation-model-tier dependency than to a side feature.