MakerKit → MakerKit Pro
10x Hardware & A Feasible Path to Robotics
Ten times faster hardware development — making robotics possible
Three C's build a machine that moves · the fourth makes it hold up in the real world hold up
Brick robot arm · MK
Quadruped dog · Pro
Dexterous hand · Pro
Heavy-lift drone · Pro
The 3C of MK
No CAD · no coding
still get a mechanism that actually moves
Completeness → one brick robot arm
3C plus one C → MK Pro
Parts go from bricks to real metal
C₄ computes how the machine meets the physical world
Robot dog · dexterous hand · drone — the same path
What we can supply
Taken apart, every piece can be sold on its own
One printer makes the parts
Buildable anywhere · shipped worldwide
The 3C of MK
First: why building hardware is still the privilege of a very few.
The Problem: the n⁹ hurdle of hardware development
Process barrier (3 stages × 3 skill sets)
Idea
Prototype
Product
Every stage needs 3D design, electronics and software → n³ per stage; three stages give (n³)³ = n⁹
The fatal flaw of traditional CAD
- Takes years to master (SolidWorks, Fusion360, etc.)
- Design and reality are disconnected (moves on screen, fails in the real build)
- Beginners and non-engineers have almost no way in
n⁹ harder than software
Demand for smart hardware is soaring globally, yet very few people can actually build it.
Where n⁹ comes from
Software fails along one dimension: code. Hardware has nine.
| 3D design | Electronics | Software | |
|---|---|---|---|
| Idea | If you cannot draw it, you cannot discuss it | No idea how many motors, or what rating | No idea whether the logic is feasible |
| Prototype | Model, print, assemble, rework | Select, wire, debug, fry a board | Firmware, comms, tuning |
| Product | Tolerances, tooling, batch consistency | PCB, certification, supply chain | OTA, stability, production flashing |
Nine cells — any one blocked and the whole thing stops
The nine cells are coupled
Change one dimension → the motor changes → the calibration changes
In software one changed line affects one line; in hardware one moved hole restarts all three tracks
No one person fills all nine cells
In practice: you assemble a team
for a hardware startup, the cost of getting started is an order of magnitude higher
3C collapses the nine cells into one act
Key solution: breaking the "process" and "skill set" barriers
Drastically lowering the cost of building the prototype — the intermediate language
Idea | ||
| ↓ |
→ Process barrier
(PoC validation) | |
Non-professional |
→
Prototype
→ |
Professional engineer |
|
→ Skill barrier
(shared language for specs)
↓ | ||
Product / Production |
The language of hardware is a prototype anyone can reach
MK Cube: the intermediate language between an idea and a professional spec
The solution: the 3C framework
Three things, each owning one segment; together they make a machine that moves.
CONSTRUCT
Building structure
Snap a mechanism together like bricks
Connections carry constraints — assemble it wrong and it will not fit
Assemble it right and it actually moves
CUBE
Smart modules
Blocks that turn electricity into motion — motors · servos
Blocks that turn the world into numbers — distance · attitude · sound
Drag one in and it works, no wiring, no part selection
CODE
Visual logic
Say a sentence → AI generates the blocks
Drag blocks → compile to Python / C++
No programming background needed
Nine cells collapse into one act: Assemble → attach → describe the motion, and then it moves.
Live demo — 3C Platform v2
A browser-based dev environment integrating Construct × Cube × Code (MakerKit.com live demo)
Drag and drop IoT capability
Real-time simulation engine
Generates and compiles Python / C++ on the spot
C₁ Construct — simulatable brick assembly
The constraint exists the moment you snap it on
Axle through hole = revolute joint
Gears meshed = ratio
Pin in slot = prismatic joint
Not annotated afterwards — produced by the act of assembly
Which is why it goes straight into the physics engine
No export · no redefining joints · one click to solve
It moves on screen because it moves under real constraints
One-to-one with the physical bricks
Every part in the digital model has a real brick in your hand. Once the simulation passes, you just build it.
C₁ has accumulated 300+ mechanisms
Not drawings — every one has run in the physics engine and genuinely moves.
Part of the library in physical form — gears, linkages, cams, differentials, tracks and more
18 families settled; structural families expanding
The classification was computed by the program
Gears → centre distance and tooth ratio
Worm → perpendicular, non-intersecting axes
Rack → linear travel proportional to rotation
Classified from geometry, not hand-labelled
So the library knows what is inside it
Ask for a mechanism that turns rotation into linear motion and it returns candidates, instead of making you browse a catalogue.
Full list and animations in the appendix.
How far 3C goes — one complete brick robot arm
The structure is snapped together · the motor is dragged in · the motion is spoken
This is the proof that 3C is complete
Serial joints · an end-effector with orientation · repeatable motion
If one joint constraint is fake, the end-effector drifts
It runs
C₁ constraints right · C₂ drive works · C₃ logic runs
All three C's close on the same object
At this point, someone who cannot use CAD and cannot code is holding a machine that moves.
C₂ Cube — distributed IoT smart modules
"Electricity into motion" and "world into numbers", packaged as blocks you drag in and use.
Blocks that push
DC motors · servos · steppers
Drag one in and you get speed / angle / torque interfaces
No datasheets · no wiring
Blocks that sense
Distance · attitude · sound · vision
The output is a number, not a voltage
Feed it into C₃ and it decides
They network themselves
Wi-Fi mesh · no central controller · no harness
Adding a module = adding a capability
Same set of Cubes; change the structure and it is a different machine
C₃ Code — AI Vibe Code
"back up and turn right when an obstacle is detected"
Blockly visual programming
Build logic by dragging. No programming knowledge required.
LLM code generation
Natural language → Blockly blocks
the glue between Construct and Cube
No programming background needed
OTA deployment
Deploy code to every Cube at once over Wi-Fi. No cables.
Code glues Construct + Cube → anyone can build a working machine
3C plus one C → MK Pro
Parts go from bricks to real metal. The extra C computes how the machine meets the physical world.
From MK to MK Pro: the three C's stay, a fourth is added
The method is unchanged; the parts change — plastic → metal, toy motor → torque-controlled actuator
| MakerKit | MakerKit Pro | What changes | |
|---|---|---|---|
| C₁ Construct | Plastic bricks on a standard hole grid | Metal structural parts, 3D-printed parts | The constraints are identical; the parts can now take the load |
| C₂ Cube | Toy motors + sensor modules | Torque-controlled actuators + IMU / encoders / lidar | Same interface, higher range and precision |
| C₃ Code | Blockly logic | Blockly logic + policies trained by reinforcement learning | Still "give it a motion" — only now the motion can be trained |
| C₄ Calibrate | — not needed on screen | Computing how the machine meets the physical world | The extra layer the real world adds |
MK: passes in simulation, built by hand
MK Pro: make that machine hold up on the ground
Delivered with paid assistance from MK engineers
You run the first three C's yourself
C₄ needs a calibration rig · real-machine measurement · policy retraining
Comes with Pro; our engineers walk you through it
Building a machine is like building a house
Same principles, different dimensions
Same framework, the quantities that go in differ.
| ① Geometry | ② Kinematics | ③ Actuation | ④ Physics · only known once lived in | |
|---|---|---|---|---|
Robot dog | Leg length · joint placement | Joint angle ↔ foot | Motor choice · gear ratio | Friction · impact · latency |
 Robot arm | Link length · joint placement | Joint angle ↔ end-effector pose | Torque · gear ratio | Friction · backlash · compliance |
Dexterous hand | Phalanx length · joint placement | Fingertip pose | servos + tendon coupling | Tendon hysteresis · fingertip friction |
Drone |
Wheelbase · rotor pitch | Rotor layout and moment arm | Thrust and RPM | Rotor efficiency fade · heating · load swing |
Serial mechanisms, solved link by link
Underactuated: 4 rotors for 6 degrees of freedom
C₄ — bring back the numbers of the fourth layer
To do work, a machine has to touch the real world somewhere. That contact is what C₄ computes.
Robot dog
Foot ↔ ground
How hard the ground pushes back, how slippery it is, how stiff the joint turns, how much torque is left after a run.
Dexterous hand
Fingertip ↔ object
How several servos combine into one gesture, whether the fingertip grips or slips, and the difference between pulling and releasing.
Drone
Rotor ↔ air
How much lift a given RPM gives, how four rotors combine into an attitude, how much lift is lost as the battery drains.
Different contact, same method — Measure on the machine, feed it back into simulation, and train again.
Different physical environment, different things to calibrate
Snap the structure together
Attach motors and sensors
Give it motion
Up to here, all three lines take the same path
↓ here it forks ↓
| Machine | The physical environment it meets | What to calibrate |
|---|---|---|
| Robot dog | Foot ↔ ground | Joint friction · touchdown impact · torque fade under load · command-to-motion latency |
| Dexterous hand | Fingertip ↔ object | Tendon force and hysteresis · fingertip friction · How several servos combine into one pose |
| Drone | Rotor ↔ air | Lift and attitude torque · rotor efficiency fading with charge · motor heating · slung-load swing |
All three are Physical AI — a closed loop of sensing · learning · actuation
Different contact → C₄ computes different quantities
Robot dog — same house; the contact is foot and ground
Real trot gait — the policy was trained in simulation and runs directly on the machine
Leg length · joint placement → joint angle ↔ foot → motor choice · gear ratio → friction · impact · latency
How far it reaches How far the leg can step · How much force Whether it holds · How much room it takes Whether the motor fits inside the leg
Joint friction (measured several times the simulator default) · touchdown impact · torque fade under sustained load · command-to-motion latency
Dexterous hand — same house; the contact is fingertip and object
Tendon-driven multi-finger hand — one finger's pose is the sum of several servos
Phalanx length · joint placement → fingertip pose → servos + tendon coupling → tendon hysteresis · fingertip friction
How far it reaches How wide the fingers open · How much force Whether it grips · How much room it takes Whether the servos fit inside the palm
Tendon hysteresis between pull and release · whether fingertip friction holds · how several servos combine into one pose
Drone — same house; the contact is rotor and air
Heavy-lift configuration · load slung under the fuselage
Wheelbase · rotor pitch → Rotor layout and moment arm → Thrust and RPM → rotor efficiency fade · heating · load swing
Whether it can lift Thrust margin · Whether an attitude can be composed Control authority · Whether it can fly Takeoff weight (underactuated — a different set of three)
Lift per RPM · lift lost as the battery drains · efficiency lost as motors heat · how a slung load swings
Two tiers, two price points
The difference is C₄ — the real-world layer costs something.
3C · for education and makers
Brick kit + online platform
Once the simulation passes, build it as shown and it moves
Cost: plastic parts + generic modules + SaaS
3C + C₄ · for R&D and products
Metal and printed structural parts + torque-controlled actuators + calibration and training pipeline
The premium is C₄: calibration rig · real-machine measurement · simulation correction · policy training
is a people-and-compute cost, not a parts cost
Pro sells more than tougher parts — is a calibrated parameter set that already runs on the machine
Offshore engineering and hardware supply
The full engineering capability can be deployed offshore · machines and components can be customised
Offshore engineering · Custom robotics · Component supply
Supplied by module — no whole-machine minimum
Whole machines cost a lot to ship and take long to deliver. Broken down to modules, every layer can be delivered on its own.
Shipped with calibration parameters
Bolt it on and it works — no trial of your own
Works with someone else's machine too
Protocol and SDK are open
Sold individually
No whole-machine minimum order
Why these can ship worldwide
Because They do not depend on any one country's supply chain.
Everything is off the shelf
Motors · bearings · screws — orderable online
Nothing custom · no long-lead stocking
A desktop printer makes the parts
No machine shop, no tooling. Setting up production drops from millions to thousands.
It travels as files, not containers
Drawings and calibration parameters are files —Send them and it is made locally. Freight, tariffs and export bans are all bypassed.
So we are not producing in one place and then shipping it out —
We can produce anywhere — the United States included.
Same parts, different arrangement, different machine
So we do not stock a separate line per machine — One supply chain carries the whole product line.
One actuator module → quadruped / biped / wheeled; only link length and joint order change
Reprint it on the spot
About US$1.50 in filament · printed locally in hours
No waiting on freight · no supply risk
Costs are fully visible
Every screw has a unit price and a purchase link, Local cost can be computed in any country.
One kit, one printer, one drawing set — none of them is tied to a place.
What we can do
More than off-the-shelf — The technology can be deployed to your site; machines can be built to your requirements.
Move this capability to your side
3C platform · C₄ calibration pipeline · actuators and drives —
Delivered as a whole, with our engineers alongside.
Machines built for your scenario
Quadruped · robot arm · dexterous hand · drone —
Same kit, different topology — no starting from zero.
Parts built to your spec
Actuators · driver boards · structural parts —
Sold individually; wire it into your own machine.
Buildable → stands up → manufacturable anywhere
10x Hardware
Hardware development, ten times faster.
Passes in simulation, built by hand
Make it hold up in the real world
MK → MK Pro makes building robots possible.
Everyone can build.
All on one page
3C: buildable
Hardware is hard across nine cells — three stages × three skill sets.
3C collapses nine into one: Assemble → attach → describe the motion.
300+ mechanisms, 20+ mechanism families, one complete brick robot arm.
C₄: stands up on the machine
Building a machine is like building a house — Decision sequence cannot be reversed, Size determines force.
The first three layers are computable on screen, The fourth layer can only be measured on the machine.
Dog · arm · hand · drone share one framework; the quantities differ.
Supply: manufacturable worldwide
Modules that stand on their own — actuators · driver boards · structural parts.
Ordinary parts · printable structure · transmittable drawings.
Technology can go offshore; machines can be customised.
Everyone can build.
20+ mechanism families
300+ models · classified from geometry by program · 20+ families
Every animation is real motion out of simulation · isometric view so the transmission reads clearly
Mechanism families · 1 – 3
Spur gears
Big wheel drives small: faster. Small drives big: stronger
In real machines: behind almost every motor — in every joint of the robot dog
appears 253 times in the library
Belt / pulley drive
A belt carries force to a distance; pull too hard and it slips
In real machines: printer paper feed, treadmills, distal joints of robot arms
appears 189 times in the library
Worm drive
Turn it and it advances a little; let go and it will not back-drive
In real machines: powered blinds, servos, lift tables — anywhere it must hold with power off
appears 113 times in the library
Mechanism families · 4 – 6
Cam
One rotation and the follower rises and falls on the rhythm you chose
In real machines: engine valves, timed motions on assembly lines
appears 78 times in the library
90° change of direction (bevel / crown)
Sends the force around a corner
In real machines: car driveshaft to wheel, robot arm wrists
appears 74 times in the library
Rack and pinion
Turns rotation into travel in a straight line
In real machines: car steering racks, 3D printer beds, electric linear actuators
appears 55 times in the library
Mechanism families · 7 – 8
Gearbox / shifter
Flick it into another gear and the output speed changes
In real machines: car gearboxes, machine-tool headstocks — one power source, many speeds
appears 38 times in the library
Linkage (four-bar and friends)
A few bars in a loop: move one and they all move
In real machines: the robot dog's leg is one — so is an excavator boom
appears 25 times in the library
Mechanism families · 9 – 11
Turntable / slewing ring
Rotates large loads, and stays steady
In real machines: excavator upper body, robot arm bases, radar mounts
appears 24 times in the library
Ball joint
One point that can rotate in every direction
In real machines: the human shoulder, parallel robots, car suspension
appears 13 times in the library
Differential
Both wheels can turn at different speeds while both still drive
In real machines: every car has one — inner and outer wheels differ through a turn
appears 11 times in the library
Mechanism families · 12 – 14
Same car — here, look at the universal joint on the driveshaft
Universal joint
Two shafts that are not in line can still pass force between them
In real machines: car driveshafts, wherever a robot arm turns a corner
appears 10 times in the library
Same tracked vehicle — here, look at the drive sprocket
Chain and sprocket
Like a bicycle chain: spans a distance without slipping
In real machines: bicycles, conveyors, some robot arms
appears 5 times in the library
Damper / spring
Absorbs a hit, then pushes back
In real machines: car suspension, landing cushioning on robots
appears 4 times in the library
Mechanism families · 15 – 17
Same car — here, look at the steering linkage on the front wheels
Steering mechanism
The two front wheels turn through different angles
In real machines: every car — and wheeled robot chassis
appears 3 times in the library
Rope / pulley / winch
It can only pull, never push, so they come in pairs
In real machines: the tendons of a dexterous hand — cranes and lifts
appears 3 times in the library
Same tracked vehicle — here, look at the track itself
Track
The whole length touches the ground, so rough terrain is no problem
In real machines: tanks, excavators, off-road robots
appears 3 times in the library