Machine Protection System (MPS) requirements for the currently operating Linac Coherent Light Source (LCLS- I) demand that fault detection and mitigation occur within one machine pulse (1/120th of a second at full beam rate). The MPS must handle inputs from a variety of sources including loss monitors as well as standard state-type inputs. These sensors exist at various places across the full 2.5Km length of the accelerator and beam lines. A new MPS has been developed based on a distributed star network where custom-designed local hardware nodes handle sensor inputs and mitigation outputs for localized regions of the LCLS accelerator complex. These Link- Nodes report status information and receive action commands from a centralized processor running the MPS algorithm over a private network. The individual Link- Node is a 3u chassis with configurable hardware components that can be setup with digital and analog inputs and outputs, depending upon the sensor and actuator requirements. Features include a custom MPS digital input/output subsystem, a private Ethernet interface, an embedded processor, a custom MPS engine implemented in an FPGA and an Industry Pack (IP) bus interface, allowing COTS and custom analog/digital I/O modules to be utilized for MPS functions. These features, while capable of handling standard MPS state-type inputs and outputs, allow other systems like beam loss monitors to be completely integrated within it. To date, four different types of Link-Nodes are in use in LCLS-I. This paper describes the design, construction and implementation of t Show Partial Abstract