INDUSTRIAL ROBOT · TELEMETRY NORMALIZATION
16,908 confirmed mappings·18 OEM families·14 protocols·99% field coverage
Forge turns raw KUKA tags into one clean, governed schema your AI agent already understands. One MCP call, no per-robot integration project.
KUKA KRC controllers surface state as $-prefixed system variables ($AXIS_ACT, $TORQUE_AXIS) over KUKAVARPROXY or EKI. Expressive but utterly vendor-specific.
Typical KUKA equipment communicates over OPC UA, ROS 2 (DDS), PROFINET / S7, EtherNet/IP. Each firmware revision, model line and integrator names the same physical quantity differently, so every AI or analytics project pays the integration tax again, the tax that cancels 40% of industrial AI deployments before they ship.
Forge normalizes KUKA telemetry into Forge's canonical field set, the same 366 canonical fields used across all 18 manufacturers. Point Forge at the robot, pass oem: "kuka", and spindle speed is spindle_speed_rpm whether it came from a KUKA controller or anything else on the floor.
Confirmed canonical fields your agent gets from KUKA robots, drawn from Forge's 16,908-mapping corpus:
| Raw tag in | Canonical field out | Meaning | |
|---|---|---|---|
| $AXIS_ACT.A1: 88.2 | → | robot.joint.position: 88.2 | axis 1 actual angle |
| $TORQUE_AXIS_1: 12.4 | → | robot.joint.effort: 12.4 | axis 1 torque |
| $PAYLOAD.M: 6.0 | → | robot.payload: 6.0 | payload mass |
| $PRO_MODE: "AUTO" | → | robot.mode: "AUTO" | program mode |
No SDK, no per-tag mapping, no historian project. Connect the MCP server and normalize live KUKA data:
Unknown tags never get force-mapped. Fields below the confidence floor abstain rather than guess, so your agent never acts on a bad reading.