$1,688Maker site, October 2026. Full price, no deposit, ships fall 2026. Includes the open-source SDK, Discord community and priority shipping. Developers and researchers can also apply for a beta

AvailablePython from a laptopTwo 7-DOF arms$1,688Assembled in San FranciscoLeRobot datasetsPython SDK

Weeks+Shipping: Fall 2026. Ordered direct from Nori at full price, shipping fall 2026. Nori's docs say the A3 ships assembled and has gone to customers since July 2026. New Atlas reports the first batch of over 100 sold out and a second batch of 100 is being assembled. Nori says it is assembled in San Francisco, so the July 2026 FCC rule on foreign-made robots does not apply. Nori ↗

Remote control: Install the Nori Lab desktop app (currently run from source on macOS with Python 3.12+ and lelab), or pip install "nori-sdk[all]" and drive the robot from Python with RemoteTeleop over WebRTC. A mock_session() allows development with no robot.

Official ↗Order ↗A3 docs ↗Python SDK ↗URDF ↗Printable parts ↗Humanoids Daily ↗

Overview

A very low-cost wheeled two-arm robot with a telescoping lift, Feetech STS servos and a Raspberry Pi 5, sold more as a developer platform than a finished home helper. Skills are taught by teleoperated demonstration in the Nori Lab app, trained in the cloud and can be shared on a marketplace, and the Python SDK, protocol and printable parts are open. The catch is that nothing autonomous comes pre-loaded, the 1.5 kg servo arms are position-controlled, the differential base cannot strafe, and full A3 docs are still marked "coming soon". It is closest in spirit to XLeRobot but ships assembled.

Specifications

Category	Wheeled rovers · Laptop is the brain
Released	2026
Country	USA
Availability	Available
Sales	First batch of over 100 units sold out (2026, estimate) New Atlas ↗
Programmability	Full SDK
Built-in autonomy	None / scripted · Moves only when controlled, or plays preset motions
LLM support	Official LLM tools · Nori Lab is a fork of Hugging Face's LeLab, and recordings become LeRobot-format datasets, so LeRobot's language-conditioned policies apply. The Nori-Lab repo also holds cloud-inference adapters for VLA models (pi0.5, GR00T, MolmoAct2), which look like work in progress. A speaker and mic are fitted for spoken commands. Nori docs, Training ↗
MuJoCo simulation	Maker URDF only · Official URDF/xacro (nori_description), which Nori says converts to MuJoCo by the standard URDF path. Mass is about 10% light and inertia approximate. nori_description ↗
NVIDIA Isaac simulation	Maker URDF only · Official URDF/xacro, with Isaac Sim import tips in the docs. No packaged Isaac asset. Nori A3 docs ↗
Gazebo simulation	Maker URDF only · Official URDF/xacro ROS 2 package with ros2_control and RViz launch files, no Gazebo world. A kinematic browser sim exists at lab.norirobotics.com. nori_description ↗
Best for	Ages 14+ (estimated)
Vision sensors	Camera, Lidar
Actuators	Smart servos (Dynamixel, Feetech…)
Audio	Speaker, Microphone
Compute type	Raspberry Pi, Laptop / PC as the brain
Remote control	Python from a laptop
Size	45 × 45 cm base, 69–145 cm tall (telescoping lift), 20.4 kg
Drive	Differential drive with passive casters, three-stage telescoping lift (30 mm/s)
Compute	Raspberry Pi 5 (4 GB) for the control loop and safety layers. Policy inference runs off-board in the desktop app, and training runs in Nori's cloud
Vision	4× 720p RGB cameras up to 30 fps (both wrists, head and neck). The live feed is a tiled 320×240 15 fps stream
Lidar	Lidar with 12 m range, 8–12 Hz scan, 0.72° resolution at 10 Hz
Sensors	Servo position and current feedback, gripper grip-force estimate from motor current (no force sensor), speaker and microphone
Runtime	6–8 h (24 V, 18 Ah)
Speed	Not published
SDK	nori-sdk (Python, Apache-2.0, on PyPI) and @nori/sdk (TypeScript) over the open nori-protocol, a ROS 2 gateway on the robot, Nori Lab app (LeRobot-based recording and cloud training), skills marketplace