It measures just over a metre tall, rolls on four wheels, manoeuvres two articulated arms and folds your T-shirts. So far, just another domestic robot. Except that the American company Vulcan Robotics, founded in October 2025 by Nick Maselli, has done something rare in this sector. It has published the entirety of its machine’s plans under an open-access licence: mechanical plans, electronic schematics, code, and even the 3D printing files, all hosted on its GitHub repositories, reports goodtech.info. The robot is called Sourccey. The first assembled units are due to be shipped in September 2026. And the price? Nowhere.

What does “plans published” actually mean?

It means the complete dossier is online, freely accessible, on GitHub repositories — the platform where developers upload and share their code. There you will find the mechanical plans (the geometry of the parts), the electronic schematics (the wiring) and the code that runs the whole system. Including the STL files, the format read by a 3D printer: you download, you start printing, you get the part. Nothing is kept secret.

What it can fold

Sourccey comes with four pre-trained artificial intelligence models, one per clothing family: T-shirts, shorts, jeans, long-sleeved shirts. These are the four provided by the manufacturer; what they achieve in a real living room, on a real laundry basket, is not documented anywhere. The scope is narrow, but it is clear.

And for everything else? You pilot it

To teach Sourccey a gesture not included in the four provided models, you do not “explain” anything to it: you teleoperate it yourself. In other words, a human takes control and guides the robot’s movements. Two ways to do this: paired control arms, which you manoeuvre by hand and which the robot reproduces, or a Meta Quest virtual reality headset.

During this phase, the machine decides nothing at all. It executes. It is your recorded gestures that then serve as the raw material for training a custom model. The chosen format is LeRobot, Hugging Face’s robotics toolkit — the major platform for sharing artificial intelligence models. The model you build in your living room therefore speaks the same language as those of others.

What is it made of?

The structure and casing are printed at home in PLA — the most common plastic in 3D printing — on a desktop 3D printer: at the manufacturer, a Bambu P1S. Everything else is purchased: the board, the sensors, the cameras, the wheels. And no one has published the total bill.

  • Size and weight: 1,030 mm tall and 15.88 kg, with a floor footprint of 414 mm in diameter — 41.4 cm, the size of a small side table.
  • Brain: a Raspberry Pi 5 board, the consumer mini-computer.
  • Movement: four Mecanum omnidirectional wheels, allowing movement in all directions.
  • Arms: two, each with five degrees of freedom — five axes of movement — plus a gripper.
  • Perception: a 360-degree LiDAR — a sensor that measures distances around it using lasers — and four cameras.

How much does it cost?

No one knows, and this is the gaping hole in this announcement. No selling price is published for the kit, i.e. the version as spare parts to be assembled yourself. Nor is there a bill of materials cost — the total cost of parts, what you would spend to buy and print everything yourself. Neither on the Vulcan Robotics website, nor on its specifications page, nor in the articles covering the publication. The announced schedule, however, is dated: first shipments of assembled robots in September 2026.

Does this concern me?

If you are looking for a device to plug in and forget about, not yet: without an announced price and with only four gestures in the catalogue, this is not household appliances. If you have a 3D printer and the desire to get your hands dirty, however, yes — and that is precisely the point of the matter. Everything is readable and downloadable: you can inspect the code, see how each part is designed, print your own.

There is another reason to take an interest, regardless of your level: when the plans for a domestic robot are laid out on the table, you can verify what it actually does. How many cameras, which sensor, which learned gesture and by whom. On Sourccey, the answer is written in black and white: four provided models for four types of clothing, and for everything else, a human in control.