Spot 5.2 lets AI agents dispatch robots via MCP layer

Spot 5.2 builds on Boston Dynamics' prior work on AIVI, the visual inspection system it previously said would use Google Gemini models trained on industrial data. That system was designed for specialized inspections of 5S boards, sight glasses, pallets and gauges; the new release extends it to video through AIVI-Learning, which the company says can catch conditions harder to see in still images, such as water dripping from pumps, slipping conveyor lines and equipment warning lights.

The Orbit redesign around assets rather than individual inspections is meant to mirror how equipment is already organized in CMMS, MES and EAM platforms. By importing asset hierarchies and metadata, Orbit allows more targeted alerts—pump-specific anomalies for the engineers responsible for pumps—and lets plant managers review trends across a facility, an equipment class or a production area. The expansion of Orbit's APIs to accept inputs from programmable logic controllers, fixed sensors and security cameras, rather than only from Spot, is what the company describes as the foundation for the MCP layer.

The stakes appear to hinge on how many of those external systems industrial operators are willing to wire into Orbit, since the autonomous dispatch and security workflows the company describes depend on those connections. For plant engineers and security teams already running such systems, the update points toward a model where an anomaly detected by one device triggers a physical investigation by a robot—though whether that reduces downtime or simply adds a layer to existing processes is not something the body resolves.