makermods/shop/so-101
Two white SO-101 robot arms with grippers and serial-bus wiring
STANDARD OPEN ARM · LEADER / FOLLOWER
5 JOINTS + GRIPPER12 × SERIAL-BUS SERVOS / PAIR
[ OPEN-SOURCE ROBOT ARM · HUGGING FACE SO-101 ]

SO-101.

Move the leader by hand. The follower mirrors you. Record demonstrations, train a policy, and turn a desktop robot arm into a practical physical-AI lab.

$300 one leader + one follower ships in August
[ CHOOSE YOUR SO-101 KIT → ]
pair · bimanual · bimanual OpenBooth bundle
open design · Apache-2.0 source repository
MakerMods Lab/Feetech STS3215/3D printable/USB
[ LEADER + FOLLOWER · 02 ]

Built to teach by showing.

SO-101 is a paired teleoperation system. Guide the low-resistance leader arm through a task; the follower reproduces the motion while MakerMods Lab captures the camera views, observations, and actions that become training data.

[ OPEN SOURCE · 03 · MAKERMODS LAB ]

Train your own SO-101 skills.

MakerMods Lab gives your robot one place to collect demonstrations, train policies, and grow a reusable skill library. Install it with one command and start building your own SO-101 behaviors.

[ VIEW ON GITHUB → ] Open source · build, inspect, and contribute
terminal01 · install + launch
uv tool install "git+https://github.com/makermods-robotics/makermodslab" && makermodslab
github.com/makermods-robotics/makermodslab ↗
MakerMods Lab interface showing a searchable library of SO-101 robot skills MakerMods Lab · SO-101 skill workspace
OpenBooth configured with SO-101 robot arms OpenBooth · SO-101 skill lab
[ ADD OPENBOOTH · 04 ]

Give every task the same stage.

OpenBooth adds a controlled workspace, two cameras, and a public catalog of recorded SO-101 skills. Start from community data, then record the demonstrations your next behavior still needs.

  • 01Consistent camera views and background
  • 02Bimanual SO-101 bundle available
  • 03Public skills hosted on Hugging Face
[ BUY THE BUNDLE → ]Browse OpenBooth skills
[ READ THE SOURCE · 05 ]

Documented all the way down.

Hugging Face documents the SO-101 setup, while the open hardware repository publishes the parts and build files behind the arm.

Hugging Face

SO-101 setup guide

Motor setup, assembly, ports, and calibration.

OPEN DOCS ↗ GitHub

Hardware repository

BOM, print files, CAD, optional hardware, and simulation.

VIEW SOURCE ↗
[ HARDWARE · 06 ]

The open desktop learning arm.

A repairable structure, off-the-shelf servos, and published build files keep the system understandable from the gripper to the training workflow.

so-101 · standard open armspec_sheet.txt
configurationleader + follower teleoperation pair
actuation5 arm joints + gripper 6 servos per arm
follower motors6 × Feetech STS3215 1/345 gearing
leader motors3 × 1/147 · 2 × 1/191 · 1 × 1/345
control linkserial-bus controller · USB to computer
structure3D-printed parts · desk-clamp mounting
softwareMakerMods Lab
workflowscalibrate · teleoperate · record · train · evaluate
datasetsMakerMods Lab recordings · Hugging Face Hub
source filesBOM · STL · STEP · assembly guide · simulation
repository licenseApache-2.0
[ FAQ · 07 ]

Before your first episode.

You guide the leader arm by hand and the powered follower mirrors it. MakerMods Lab records the follower's motion and connected camera views as demonstrations for robot-learning datasets.
Use MakerMods Lab to connect and calibrate the arms, teleoperate the follower, record tasks such as pick-and-place, train a policy, and evaluate the learned behavior.
Not for MakerMods Lab calibration, teleoperation, or recording. Training time and hardware depend on the policy and compute target you choose.
OpenBooth gives the arms repeatable camera views and a consistent visual workspace. It also connects the setup to a public catalog of SO-101 recordings on Hugging Face, so experiments can start from shared data.
Yes. Hugging Face publishes the official SO-101 setup documentation, and The Robot Studio's SO-ARM100 repository contains the bill of materials, printable files, optional hardware, and simulation assets under the Apache-2.0 license.