Robotyx
The control system for an industrial site. It sees the site, plans the work ahead against a model of what the site can physically do, and orchestrates the people and machines that carry it out.
Three things, and the order matters.
One namespace
Every asset on the site publishes into a single namespace and is described once. A new asset joins by publishing into it, rather than being wired to everything else as a bespoke project.
A world model
A physics model of the site that people and machines are managed on together, with safety, compliance and reporting carried inside the model rather than assembled afterwards.
Agents under approval
Agents draft the next action, a person approves it, and the system dispatches and logs the result. Every step is attributable, and gates cannot be self-promoted.
Modules, not monoliths.
Robotyx ships as a catalogue of modules including safety and compliance, operations and drilling, and AI assets. Sites turn on what they need and add the rest over time.
Edge first, link optional
The edge agent reads existing industrial protocols, is read-only by default, and buffers locally when the satellite link drops. Remote sites stay operational when connectivity does not.
Built to be built on
An open developer surface with a documented interface, so partners and operators can build against the platform directly. We ship the substrate; other people build the applications that run on it.
Where Robotyx sits.
Partner apps, operator workflows and sub-agents, connected through the open interface.
Namespace, world model and orchestration — the layer that decides what happens next and makes it happen.
Controllers, sensors and equipment from the vendors who already build them well. Robotics command is on the roadmap.
Want to see it on a real site?
We run structured pilots with operators, and we are selective about them.