Affordable
A smaller, more practical entry point for embodied-AI experiments.
MechaSeed One — in development
Prototype direction / 01A modular small-biped robot for physical AI learning, experimentation, and developer workflows.
Built on an Open Duck Mini v2-style hardware foundation, with a compatibility layer targeting MicroDuck-style developer workflows.

What MechaSeed is
Most physical-AI robots are expensive, closed, or difficult to access. MechaSeed exists to make small-biped robotics more modular, understandable, and practical to experiment with.
A smaller, more practical entry point for embodied-AI experiments.
Built around inspectable joints, replaceable modules, and real fasteners.
Developer workflows are designed to stay legible, adaptable, and learnable.
A compact platform for taking ideas from simulation into motion.
Meet the robot
MechaSeed One is being designed around a practical small-biped architecture: modular joints, developer-friendly electronics, and a hardware stack suitable for learning, experimentation, and iteration.
Current status: engineering direction under validation. The visual below is not a final production specification.

What it is designed to do
Develop and test compact-biped locomotion.
Explore expressive pitch, yaw, and roll control.
Work through a concise Python-oriented control surface.
Shape behaviors in simulation before moving hardware.
Inspect, repair, replace, and iterate on real modules.
Use a platform designed for learning, not a sealed appliance.
Capabilities shown are engineering targets, not shipping specifications.
Compatibility, clearly explained
MechaSeed pairs a practical Open Duck Mini v2-style small-biped hardware foundation with a compatibility layer targeting MicroDuck-style developer workflows.
Compatibility has a specific meaning here. It refers to developer workflows and API concepts, not identical hardware or guaranteed policy interchangeability.
Architecture / target stack
Click a layer to see the role it plays. The compatibility layer is an adapter between MechaSeed’s runtime and familiar control concepts—it is not a claim of official equivalence.
A practical small-biped layout informs the mechanical and electronics foundation.
Compatibility matrix
Current targets are shown plainly so expectations can evolve with evidence.
| Layer | Status |
|---|---|
| Small-biped hardware foundation | Planned |
| Modular joint architecture | Planned |
| MicroDuck-style control concepts | Targeted |
| Developer-friendly API surface | Planned |
| Simulation-oriented workflow | Planned |
| Direct official policy compatibility | Not promised |
Try the API
This illustrative interface shows the intended command model. The API is not yet a released contract.
# Walk forward and turn
robot.move(
vx=0.2,
vy=0.0,
yaw=0.3,
)# Aim the articulated head
robot.head(
pitch=0.15,
yaw=-0.2,
roll=0.05,
)# Read joints and sensing
state = robot.state()
print(state.imu)
print(state.joints)
Built for developers, students, and labs
A desktop robot should be close enough to inspect, practical enough to iterate on, and open enough to teach with.

Developers + makers

Students + educators

Labs + teams
MechaSeed One is a planned modular small-biped robot for physical-AI learning, experimentation, and developer workflows. It is currently in prototype validation; the images on this page show the intended engineering direction, not final production hardware.
No. MechaSeed is an independently developed platform. It aims to support familiar MicroDuck-style developer workflows through a compatibility layer, rather than reproduce official MicroDuck hardware.
Compatibility refers to developer workflows, API concepts, and simulation-oriented control patterns. It does not mean identical hardware or guaranteed policy interchangeability.
No. MechaSeed is not affiliated with or endorsed by MicroDuck or Pollen Robotics.
That is not promised. Differences in geometry, actuators, sensing, runtime behavior, and calibration may require adaptation and validation on MechaSeed hardware.
Makers, robot developers, students, educators, university labs, and small robotics teams that want a practical route into small-biped physical-AI experiments.
No. Pricing and final specifications are still under evaluation. The founding waitlist helps us understand the configurations and price ranges that matter most.
There is no announced launch date. The next milestone is a working engineering prototype, followed by a small founding batch only after the core behavior is validated.
Help shape the first build
Join for honest prototype progress, compatibility updates, early access, and founding-batch news.