We've built a system for operationalizing hardware at scale, starting with building a CSV of identifiers for equipment in a rack, pushing this data to an internal control plane called MetalCP. This data is used by Temporal workers to implement a Host Import workflow that matches devices to their representations in our internal DCIM tool (Railyard), connects to the datacenter's management network, scrapes BMC data, creates static DHCP leases for the BMC and management NICs on the host using discovered MAC addresses from the scrape, updates a DB with server details, generates stable interface and drive names, flags faults in hardware, and configures Ansible. Additionally, we use Pixiecore to automate PXE booting, leveraging Redfish APIs to obtain near-real-time images of server screens (Claude) for state monitoring during OS installation. The dataplane network uses FRR with BGP unnumbered to achieve uniform routing setups on switches and servers, making expansion plug-and-play. Our tooling is proving an entire software vertical in itself, including MetalCP as our in-house RANCID/Oxidized replacement.