Construction is the first application — the M200 Construction Mule, in prototype build. Research and exploration programs follow, each labelled for exactly what it is
Software complete and validated in simulation · first prototype in build · no physical robot exists yet
Delaware · Malta · Working worldwide
One program is a product in development. One is research. The rest are concepts we are studying — and every card says which, so range is never mistaken for a product line.
Concept rendering — First product · software complete, prototype build starting
Autonomous material transport for construction sites: ~200 kg payload, waypoint navigation with follow-me mode in the next phase, all-weather duty, and a safety system designed for working near people. The only program with an engineering pack behind it.
Status: First product · software complete, prototype build starting
Concept illustration — Exploration · concept stage
A future arm-equipped machine that places material rather than only transporting it — blocks onto a course, panels onto a frame. Dexterous manipulation on an unstructured site is an open research problem industry-wide, which is why this is a long-horizon exploration and not a roadmap item.
What makes this domain plausible is the platform underneath, not existing hardware — nothing here has been built.
Status: Exploration · concept stage
Concept illustration — Exploration · concept stage
A small solar-assisted electric craft running on-demand routes across bays and along coastal cities, sized in concept for a handful of passengers — four at most. Water replaces the road; the autonomy, safety and fleet layers are the same ones under the M200.
What makes this domain plausible is the platform underneath, not existing hardware — nothing here has been built.
Status: Exploration · concept stage
Concept illustration — Exploration · concept stage
An autonomous platform working along crop rows: transport, irrigation and spraying first. Selective harvesting — judging what is ripe and picking it without damage — is the harder question and would come later as research, never as a promised feature.
What makes this domain plausible is the platform underneath, not existing hardware — nothing here has been built.
Status: Exploration · concept stage
Concept illustration — Exploration · concept stage
Locating people and animals in rubble, floods, avalanche and maritime incidents, then carrying first-aid payloads and breathable air to them. Life-safety work demands a higher reliability and certification bar than material transport — which is exactly why this follows rather than leads.
What makes this domain plausible is the platform underneath, not existing hardware — nothing here has been built.
Status: Exploration · concept stage
Concept rendering — Research · simulation only
A four-legged research platform for the terrain wheels cannot handle — stairs, debris, unfinished floors. Simulation-first by design: no physical build until measured milestones are passed.
Status: Research · simulation only
Underneath every program above sits the same stack: the autonomy, the deterministic safety architecture, and the fleet software that turns a machine into a service. What changes from sector to sector is the body the stack rides in and the payload it carries.
That is the honest reason a small company can name several domains: not because it has built machines for them, but because the expensive, hard-to-get-right part is built once and proven on the first one. Each further domain still has to earn its own evidence — and until it does, it stays labelled exploration.
This is the actual mission logic our platform runs — a task in plain language, split across a fleet, executed safely. Press play or step through.
Site lead — 07:12
01 A task, in plain language
The site lead types the job the way they would say it — in plain language, with 50+ languages planned. No programming, no forms.
These views are rendered directly from the M200's engineering CAD — the same model the prototype is being built from. What you see is what gets welded.
Bare-frame stage shown honestly: enclosure and sensor hardware land during the prototype build.
The interesting part of this company is not a feature list — it is the discipline underneath. Here is how the machine is engineered, without the parts a competitor would want.
The manufacturing pack is produced from the design model by code. Every sheet carries the checksum of the exact model revision it came from, so a drawing can never quietly drift from the design — if the model moves, the mismatch is caught automatically before anything reaches a shop.
Where a critical dimension depends on a supplier we have not chosen yet, the drawing does not carry a plausible-looking number. It carries a blocked marker naming exactly which answer is missing, and the sheet cannot be released until that answer arrives. A wrong hole is scrap steel, not a rounding error.
All of it validated in simulation — no physical robot exists yet
Autonomy, the safety architecture and the fleet layer are built once and carried by every machine. That is why construction is the first application rather than the only one — and why a second domain is an engineering question for us, not a rebuild.
One parametric model is the single source.
Running today
Sheets produced by code from that model.
Running today
Shop-ready packs, once suppliers are chosen.
Ahead
Fenced testing before anyone works nearby.
Ahead
Supervised sites, measured against signed targets.
Ahead
The software platform is complete and validated in thousands of simulation runs. The physical program is next — and we say plainly what exists and what doesn't.
Mission logic, fleet coordination, plain-language tasking, reporting — complete, validated in thousands of simulation runs.
First physical M200 unit: chassis, drivetrain, sensors, safety chain. Procurement and detailed design under way.
Structured physical test program in a controlled yard before any robot works near people.
Supervised trials on partner construction sites, measured against signed targets.
CE path for industrial mobile machinery, with an external safety consultancy planned.
Robots-as-a-service: sites rent outcomes, not machines.
All performance figures to date are from simulation and are labeled as such. No physical robot exists yet — that build is this phase.
A multi-layer safety system with an independent hardware emergency stop. The stop decision never depends on AI — people always have right of way.
Site leads give tasks the way they would say them — available today in multiple languages, with 50+ planned. Every task is checked against site rules before a robot moves.
Tasks split into trips and are assigned across the fleet automatically. Owners see every assignment and its reasoning in the monitoring panel.
Construction sites lose connectivity. Robots keep executing queued work, and the field app keeps working offline and syncs when the network returns.
Every evening: trips completed, material moved, safety events — a plain-language report in the owner's language, automatically.
The M200 starts as a material carrier. Its module interface is designed for future tools — dumper bed, tool carrier, inspection payloads.
Construction has terrain no wheeled robot can serve: stairwells, debris fields, unfinished floors. ACR-1 is our quadruped research program aimed at exactly that ground.
It is deliberately simulation-first: locomotion is trained and tested in physics simulation against measured milestones, and no physical unit is built until those milestones are passed. ACR-1 is a research program — it is not a product and carries no availability claims.
Every capability must pass measured simulation milestones before any hardware is considered.
Stairs, rubble, slopes, and surfaces that change from day to day.
ACR-1 runs the same fleet, tasking, and safety software stack as the M200.
The program on paper — stable links you can cite in applications and share by e-mail.
Problem, products, program status, roadmap, and the ask — for grant committees and investors.
Open PDF · 1.1 MB→One-page product summary: capabilities, safety approach, autonomy ladder, status.
Open PDF · 752 KB→One-page research program summary: simulation-first methodology and target capabilities.
Open PDF · 349 KB→We are talking to grant programs, pilot-site partners and investors who want field robotics built the careful way — construction first, evidence before claims. Tell us which one you are.
Leave an e-mail