Comment by bhewes

Good to hear. We have our MCP Running inside Larkspur, an Oil and Gas Operater in OKC. It's been six months with c-suite use. Everything runs local models, the DBs (all three) Postgres w/ Tiger and Neo4j. The local models are for processing messy energy data. And the oil bros access the mcp currently from Claude.

Spec wise: Ampere 128C, 256gb, 4xa4000,& 7800x3d, 96gb, Blackwell 24gb, 4070s. Nvidia fiber and mirkotik. Working on RDMA (but isn't really needed yet, but we have it)

OS - Tumbleweed ARM headless, CachyOS x86 hyprland. It's been easier staying on the edge of the kernel with less abstractions.