New Release: Meet myCobot 280 X Edition

Ten Years on a Desk

Ten years is a long time in robotics. Over that decade the myCobot 280 has become the arm most people actually got their hands on — small enough to live on a desk, affordable enough to be bought by a person rather than a purchasing department, and open enough that you can drive it from six lines of Python. It has ended up in teaching labs, robotics clubs, research groups and small production cells, and the community has built things with it that we did not plan for.

To mark the anniversary we made a special edition of the 280: myCobot 280 X Edition.

Free Anniversary Gifts with the X Edition Robotic Arm

Each X Edition ships with an anniversary badge & a patch of commemorative sticker in the box, and carries its own unique serial number — so every arm in the run is individually identified rather than just being one of a batch.

The badge is not only a keepsake. It is the object the arm goes looking for once the vision work starts.

Everything in the myCobot ecosystem carries straight over: the full accessory range, the AI Kit, pymycobot, ROS / ROS 2 and MoveIt, and myBlockly. Anything you have already written for a myCobot 280 runs on the X Edition unchanged.

Two Version for Your Choice

The X Edition comes in two versions, and which one you want depends on whether you would rather the arm be a peripheral or a computer.

The mechanical arm is the same in both: same reach, same payload, same repeatability, same accessories. The difference is entirely in what is doing the thinking.

Programming It

Both variants are programmed the same two ways, and you can move between them freely.

myBlockly is visual, block-based programming — drag blocks together and the arm moves. It is where most classrooms start, because a student can have the arm doing something deliberate within a few minutes and without a syntax error in sight.

Python, through the pymycobot library, is where you end up once the blocks stop being enough. It is the same set of commands underneath, so the step up is small: read joint angles, send coordinates, open and close the gripper. The vision work, the AI Kit and the badge pick all run on it.

For anyone who wants to go further, the arm also supports ROS / ROS 2 with MoveIt, and drag-and-teach, where you physically move the arm through a sequence and it records the poses.

Specifications

Wide Range of Compatible Accessories

The X Edition works with the full set of myCobot end effectors, so we mounted all of them on it. Between them they cover most of what a desktop arm gets asked to do.

• Camera Flange 2.0

Mounts the camera at the wrist so the arm sees from where it works — the basis of eye-in-hand calibration, and of everything the arm does with a camera.

• Adaptive Gripper

The general-purpose hand. Parallel jaws that self-adjust to rigid objects of varying width.

• Flexible Gripper

Compliant fingers for soft, irregular or awkwardly shaped objects that a rigid jaw would damage or fail to hold.

• Suction Pump 2.0

Flat, smooth and light items — approach from above and lift, rather than getting fingers around an edge.

• Pen Holder

Makes repeatability legible: an arm that draws shows you its accuracy on paper instead of in a spec table.

• G-shape Base

Clamps the arm to a desk edge. Not an end effector, but the accessory that makes the rest of them usable.

“Pick up the badge and give it to me”

For the anniversary demo, the arm finds the commemorative badge from its own box, picks it up and places it in a person's hand. The instruction is a plain sentence, and there is no badge-specific training anywhere in the pipeline.

Three stages: an LLM turns the instruction into a target description and a plan; an OWL open-vocabulary detector locates that description in the camera image; pymycobot converts the detection to arm coordinates through the hand-eye transform and runs the approach, grasp and handover.

The middle stage is the interesting one. A conventional pipeline needs a trained class for every object it can pick, which is why demos usually stop at colored blocks. An open-vocabulary detector is prompted in text, so what the arm can be asked to pick is bounded by what you can describe rather than by what someone remembered to label.

The whole thing runs locally on a MacBook Air. No cloud API, no GPU workstation, no inference bill — which matters more than it sounds, because it means the demo works on a classroom laptop with the wifi switched off.

Four Ways to See: The AI Kit

The AI Kit is a complete work cell — arm, camera, work surface and software — and it is the better place to start, because it exposes each stage of a vision pipeline separately instead of hiding all of them behind one model. Four recognition modes:

Running all four on one arm in a single sitting makes the trade-off legible. Color detection is instant to set up and collapses when someone changes the lighting. ArUco is unglamorous and almost perfectly reliable. YOLO handles objects the others cannot describe and asks for a labelled dataset in return. Open-vocabulary detection, as in the badge pick, skips the dataset entirely and gives up some of the precision a trained model would have given you.

Getting Started Yourself

Before any vision code, confirm the connection. If this moves the arm home and prints two lists, everything else here will work.

from pymycobot import MyCobot280
import time
# Windows: "COM3" macOS: "/dev/tty.usbserial-XXXX"Linux: "/dev/ttyUSB0"
mc = MyCobot280("/dev/ttyUSB0", 115200)
mc.power_on()
time.sleep(1)
# Six joint angles in degrees, speed 0-100
mc.send_angles([0, 0, 0, 0, 0, 0], 40)
time.sleep(3)
print("Angles:", mc.get_angles())
print("Coords:", mc.get_coords())

Where to Take It Next

● Swap the badge for anything you can describe in a sentence — the open-vocabulary pipeline does not care what the object is.

● Chain the LLM into a multi-step plan: sort, stack or arrange, rather than a single pick.

● Drive the same routine through ROS 2 and MoveIt for collision-aware planning.

● Use ArUco to pin a coordinate frame to the work surface and see how much absolute accuracy improves.

● Run two arms and hand parts between them.

Thank You for Being Part of Our Journey

Ten years is not a milestone a company reaches alone.

Every arm we have shipped ended up somewhere we could not see — a lab bench, a classroom, a garage, a workshop on the other side of the world. People taught themselves inverse kinematics on one over a weekend. Teachers built whole courses around them. Engineers filed the awkward bug reports, the ones that are tedious to write and impossible to fix without. Others published what they made, and in doing so taught the next person where to start.

We may not know every name behind every creation, but we know every contribution matters.

As we celebrate our 10th Anniversary, we’re grateful to be part of this amazing robotics community. Join us on our website for our anniversary celebration and discover more ways to celebrate this milestone together.

Here’s to the next ten years of innovation and possibilities.