FoundryNet / Solutions / Cross-vendor equipment AI

CROSS-VENDOR · FACTORY FLOOR INTELLIGENCE

One Schema. Any Manufacturer. Every Machine on Your Floor.

18 OEM families·55 manufacturer pages·14 protocols·16,908 mappings

Haas, Fanuc, Siemens, KUKA, DMG Mori, Mazak, ABB, Yaskawa. Every OEM speaks a different language. Forge translates all of them into one canonical schema any AI agent can reason about.

Get Forge on AWS Marketplace →See pricing →

Pick your manufacturer below, or paste a raw tag dump and let Forge identify it.


The problem

Every vendor solved naming independently

No manufacturer set out to be incompatible. Each one designed a controller, named its registers sensibly for its own engineers, and shipped. Fanuc exposes FOCAS. Siemens exposes S7 data blocks. KUKA exposes KRL variables. Haas buries production state in Q-variables and macro registers. Every one of those decisions is locally reasonable and collectively unusable.

Your agent does not need a better model to fix that. It needs someone to have already done the translation — with the units checked, the scales checked, and the wrong answers withheld rather than guessed.

18OEM families
366Canonical fields
16,908Confirmed mappings
14Protocols

Supported manufacturers

The vendors already in the corpus

A selection of the 55 manufacturers with dedicated coverage pages. Each links to its own tags, protocols and sample normalizations:

CNC & machine tools

Haas
MTConnect, Modbus TCP
99% coverage →
Fanuc
MTConnect, OPC UA
99% coverage →
Siemens Sinumerik
OPC UA, PROFINET / S7
99% coverage →
Mazak
MTConnect, OPC UA
99% coverage →
DMG Mori
MTConnect, OPC UA
99% coverage →
Okuma
MTConnect, OPC UA
99% coverage →
Makino
MTConnect, OPC UA
95% coverage →
Heidenhain
OPC UA, MTConnect
99% coverage →

Industrial robots

ABB Robotics
OPC UA, EtherNet/IP
99% coverage →
KUKA
OPC UA, ROS 2 (DDS)
99% coverage →
Yaskawa Motoman
EtherNet/IP, ROS 2 (DDS)
95% coverage →
Fanuc Robotics
EtherNet/IP, PROFINET / S7
95% coverage →
Universal Robots
ROS 2 (DDS), Modbus TCP
99% coverage →
Kawasaki Robotics
EtherNet/IP, ROS 2 (DDS)
91% coverage →
Stäubli
OPC UA, ROS 2 (DDS)
91% coverage →
Omron Robotics
EtherNet/IP, ROS 2 (DDS)
91% coverage →

Heavy vehicles & AMRs

Caterpillar
J1939 (CAN), HTTP / REST
99% coverage →
Cummins
J1939 (CAN), HTTP / REST
99% coverage →
John Deere
J1939 (CAN), HTTP / REST
91% coverage →
Volvo Trucks
J1939 (CAN), HTTP / REST
91% coverage →
MiR
ROS 2 (DDS), HTTP / REST
99% coverage →
OTTO Motors
ROS 2 (DDS), HTTP / REST
95% coverage →
Locus Robotics
HTTP / REST, MQTT Sparkplug B
91% coverage →
Geek+ (Geekplus)
HTTP / REST, MQTT Sparkplug B
91% coverage →

See all 55 manufacturers →


What you get

Cross-vendor questions your agent can finally answer

"Which five machines are degrading fastest?"

Requires spindle load to mean the same thing on a Haas and a Mazak. It does, once both are spindle_load_pct with the unit and scale resolved.

"Compare this cell against that cell."

Different cells rarely share a vendor. Normalized fields make a cross-cell comparison arithmetic instead of an integration project.

"What changed before the failure?"

Cross-correlation across mixed vendors only works when the timestamps and field names line up. Forge emits both.

"Roll the whole fleet into one OEE number."

Fleet-level OEE across vendors is a normalization problem long before it is an analytics problem.

Get Forge on AWS Marketplace →See pricing →

Related

Keep reading

Mixed fleet normalization

The mechanics: 5 layers in, 6 validators out, one canonical schema.

Industrial MCP server

32 tools that give any agent access to every vendor above.

Predictive maintenance

Once the fleet is normalized, prediction stops needing an integration project.


From the newsletter

The Machine Agent

Longer-form writing on industrial AI infrastructure by Foundry Labs.

The Tower of Babel

The most expensive unsolved problem in industrial AI

Read on Substack →