Skip to main content

PhysicalAI LeRobot Plugin

Bridges LeRobot robot and teleoperator configs into PhysicalAI, the Python library and runtime for robot control, transport, and CLI workflows. It registers with Physical AI Studio, the application that discovers catalog plugins and provides robot setup, teleoperation, and workflow experiences.

PyPI version Python versions

Overview

This plugin lets Studio users select supported LeRobot hardware from a schema-driven UI, auto-resolve serial devices, and run them through a PhysicalAI-compatible adapter without writing robot-specific glue code.

Each installed LeRobot follower robot and teleoperator is registered as a Studio catalog entry:

  • follower type: LeRobot_<follower_type>_Follower
  • leader type: LeRobot_<teleoperator_type>_Leader

Installation

Requires Physical AI Runtime 0.4.0 or newer for structured calibration errors.

uv add physicalai-lerobot-plugin

Run with the PhysicalAI CLI

The PhysicalAI CLI run subcommand executes a RobotRuntime from a YAML config. The bundled config teleoperates a LeRobot follower with its matching leader teleoperator:

uv run physicalai run --config packages/physicalai-lerobot-plugin/examples/runtime/teleop.yaml

The example uses the so101_follower / so101_leader pair; set config_type and config_kwargs to any of the bundled follower/leader types below. Press Ctrl+C to stop.

Calibration

Adapter connections are non-interactive, including CLI connections. Calibrate the robot using LeRobot's terminal workflow before connecting from Studio or the PhysicalAI runtime. The adapters do not automatically recalibrate or write saved calibration values to motors when they differ.

LeRobot selects <id>.json beneath calibration_dir. With a null directory, it uses its environment-dependent default and the device's internal name; that name can differ from the registered config type (for example, so101_leader uses so_leader in LeRobot 0.6). To reuse an existing file, set calibration_dir to its containing directory and id to its filename without .json.

Missing, invalid, empty, or mismatched calibration raises a structured RobotCalibrationError. Missing and mismatched calibration include the device's resolved file path when available. Studio reports these as robot_calibration_required with setup guidance instead of a generic connection failure. Calibration-free devices remain supported.

Third-party LeRobot extensions

Install this package first, then install a LeRobot extension into the same Python environment. Extensions are unverified and are not imported by default. To trust all installed LeRobot extensions, set TRUST_UNVERIFIED_PLUGIN=1 before starting the CLI or Studio. This executes package code from extensions with names that start with lerobot_robot_ or lerobot_teleoperator_.

For example, after installing a compatible LeSlider package, its registered follower and leader types appear automatically in the Studio catalog. An extension must import its config registration code from its package root and use LeRobot's register_subclass(...) mechanism. Hardware SDK dependencies remain the responsibility of the extension package.

Bundled LeRobot followers

  • bi_so_follower
  • bi_rebot_b601_follower
  • bi_openarm_follower
  • so100_follower
  • so101_follower
  • koch_follower
  • omx_follower
  • hope_jr_hand
  • hope_jr_arm
  • openarm_follower
  • rebot_b601_follower
  • reachy2
  • earthrover_mini_plus
  • unitree_g1
  • lekiwi
  • lekiwi_client

Bundled LeRobot leader teleoperators

  • so100_leader (for so100_follower)
  • so101_leader (for so101_follower)
  • koch_leader (for koch_follower)
  • omx_leader (for omx_follower)
  • openarm_leader (for openarm_follower)
  • rebot_102_leader (for rebot_b601_follower)
  • reachy2_teleoperator (for reachy2)

Payload models

How payload fields are built

Payload classes are generated dynamically from each LeRobot config dataclass.

  1. Fields are derived directly from the selected LeRobot config dataclass.
  2. Complex fields are supported recursively:
    • nested dataclasses, dict[...], list[...], tuples, optional/union types
  3. Required/default behavior:
    • fields without dataclass defaults are required
    • fields with defaults/factories are optional in payload
  4. Runtime endpoint resolution:
    • port fields are resolved at build time through the Studio factory when possible (including nested bimanual arm configs)
    • for configs with a serial port field, the schema is annotated with a Studio "Select robot" connection section bound to that port; the payload model itself adds no extra Studio fields

Common payload fields

There is no fixed cross-robot payload contract anymore. Fields come from the selected LeRobot config type.

Payload examples

1) LeRobot_so100_follower_Follower

{
  "port": "/dev/ttyACM0",
  "disable_torque_on_disconnect": true,
  "use_degrees": true,
  "id": "so100-main"
}

2) LeRobot_hope_jr_hand_Follower

{
  "port": "/dev/ttyACM1",
  "side": "left",
  "disable_torque_on_disconnect": true,
  "id": "hope-left-hand"
}

3) LeRobot_bi_so_follower_Follower (nested bimanual config)

{
  "left_arm_config": {
    "port": "/dev/ttyACM0",
    "disable_torque_on_disconnect": true
  },
  "right_arm_config": {
    "port": "/dev/ttyACM1",
    "disable_torque_on_disconnect": true
  },
  "id": "bi-so-main"
}

Notes

  • Test-only robots are intentionally not registered by this plugin.
  • Payload models are generated from LeRobot config dataclasses, so fields can differ by robot type.

Development

uv sync
uv run pytest

See the repository's contribution guide for development information.

Metadata

Release files for physicalai-lerobot-plugin 0.4.1

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for physicalai-lerobot-plugin 0.4.1
File Size Uploaded Provenance Actions
physicalai_lerobot_plugin-0.4.1.tar.gz 24.2 kB With provenance Details for physicalai_lerobot_plugin-0.4.1.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for physicalai-lerobot-plugin 0.4.1
File Interpreter ABI Platform Actions
physicalai_lerobot_plugin-0.4.1-py3-none-any.whl Python 3 none any Details for physicalai_lerobot_plugin-0.4.1-py3-none-any.whl

Total release size: 51.5 kB

Release files / physicalai_lerobot_plugin-0.4.1.tar.gz

Download URL physicalai_lerobot_plugin-0.4.1.tar.gz
Size 24.2 kB
Tags Source
SHA-256 checksum
How to use checksums
02a51a5cbf2913cef9118207a672675d30a3dd036f5edf4d94c882eb228f5767
BLAKE2b-256 checksum
How to use checksums
f417dda1aaa43a8e3c266a6d85d57cf6e15d355d16f8d55f32b454c904870b9b
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 6, 2026.

Transparency log

Release files / physicalai_lerobot_plugin-0.4.1-py3-none-any.whl

Download URL physicalai_lerobot_plugin-0.4.1-py3-none-any.whl
Size 27.3 kB
Tags Python 3
SHA-256 checksum
How to use checksums
763d0086b2446dba9dcef7b711842395c83a371428a65a3d5c913e1b58bc3c30
BLAKE2b-256 checksum
How to use checksums
8b41cf9a1e06e94c90e516ca94f7836975dbe63e0e80f962c669ae47e2febd1e
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 6, 2026.

Transparency log

Security for physicalai-lerobot-plugin 0.4.1

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Full provenance

Both files in this release have been uploaded with provenance metadata.

Select from 2 files

Attestation for physicalai_lerobot_plugin-0.4.1.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 6, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page