Finding #4 of 5
Graphiti MCP Server is the most under-acknowledged connector in the catalog
0 inbound edges · 0.71 normalised betweenness — highest non-trivial in the graph
Betweenness centrality measures how often a node sits on the shortest path between two other nodes. A node with high betweenness but low inbound-degree is, by definition, an under-acknowledged-but-load-bearing connector — nobody points at it, but the graph routes through it. The catalog's top such node is Graphiti MCP Server: 0 inbound edges, 0.71 normalised betweenness — the highest non-trivial value in the entire graph.
Why it matters structurally: Graphiti MCP Server sits on the shortest path between the MCP-spec substrate cluster (Claude Desktop, Cursor, Windsurf, the MCP spec itself) and the Zep / Graphiti citation-anchor pair (Zep is the most-cited entry in the dedicated-memory-layer category; Graphiti is its open-source graph backend). If you removed Graphiti MCP Server from the graph, the MCP-tooling cluster and the dedicated-memory-graph cluster would disconnect from each other.
The top-five bridge-surprise list is Graphiti MCP Server, MAGMA, Memformers, MemEvolve, RGMem. These are systems with high structural importance but very low explicit citation — the catalog finds them because the typed-edge graph surfaces structural roles that citation hubs miss.
Go deeper
See the bridge-surprises callout →
analysis.md §26.2
Other findings
- #1. Agent stacks leave semantic caching to the data layer 1 of 100 priority-cohort agent products
- #2. 91.7% of catalogued products publish no peer-reviewed benchmark 919 of 1,002 products · only 4 scores independently verified
- #3. The MCP spec is the catalog's #3 inbound substrate Inbound runtime-deps: Claude 62 · GPT 52 · MCP spec 34 · Gemini 22 · Qwen 16
- #5. 77% of FM-dependent products lock onto three vendors 108 of 140 FM-dependent rows depend on OpenAI / Anthropic / Google