Cotality's Real AI Product Isn't Its Connector
MCP connectors have become a commodity play. Cotality thinks the real product is the semantic layer underneath — YAML files that tell AI what property data means, not just what it says. Whether that prevents hallucinations or just makes them easier to trace is the open question.
image from grok
Cotality's MCP server launch reveals that the real product is its semantic companion YAML files, which provide structured metadata explaining how property data fields should be interpreted in real-world decisions. This addresses the fundamental gap that MCP solves connectivity but not meaning—property data fragmentation (varying zoning codes, measurement standards, risk models) causes AI systems to produce confident but wrong answers. The bet is that structured interpretive metadata enables 'fiduciary-grade AI', though whether it prevents or merely surfaces hallucinations remains an open question.
- •MCP solves connectivity, not meaning—the semantic companion files are Cotality's attempt to solve the interpretation layer that MCP itself cannot address.
- •Property data fragmentation (varying zoning codes, measurement standards, risk model assumptions) is a documented AI failure mode in property analytics.
- •Being a connector provider is a distribution play: whoever controls the MCP server influences which AI systems can query the data.
