Open Source RoboticsRobots, software & learning

LeRobot

Software

Tools for collecting robot demonstrations and training and deploying learned policies. [1]

Installation & first result →Official websiteGetting started

For
Python developers exploring robot learning [1]
Source rights
Some licenses are unresolved. Review licenses
Unknowns
operating system not fully established. Review gaps

Source researched · Not tested by us
Sources checked · Research details

Licenses & related hardware
Robots
AlohaMini2, AM-ARM200, Grabette, Open-source Leader Arm, reBot Arm B601, SO-101 and 1 moreTagged in our catalog; tags do not establish tested compatibility.
License
Apache-2.0 · license not fully assessed for model weights, datasets, documentation · Details by part

Requirements · Research details and dates

On this page

Overview

You can begin by inspecting an existing dataset, recording demonstrations with a robot, or training a policy. Each route needs a different set of dependencies. LeRobot's dataset format combines state/action data and video, giving you a way to examine the inputs and actions before choosing hardware. [1] [5]

Is this for you?

Workflows
  • Record demonstrations
  • Train policies
  • Deploy policies

[6]

What you need

Languages / runtimes
  • Python

[1]

Dependencies
  • Python environment
  • PyTorch
  • Workflow-specific extras

[5]

Compute for the workload
  • Training compute depends on the selected policy and dataset; no universal GPU budget is established.

No training benchmark performed.

[6] · Editorial assessment

Installation & first result

Open the upstream getting-started guide [5] · Project documentation [1]

Installation

Choose the installation guide matching the selected revision and add the extras required for the workflow.

[5]

Hardware, models & data

File formats
  • Parquet
  • MP4
  • images

[1]

Project notes

Install the extras for your task

The installation guide separates the base package from workflow extras: dataset for loading and creating datasets, training for policy training, and core_scripts for recording, replaying and calibration. Install the extras for the task you intend to run. [5]

The reviewed guide uses Python 3.12 and lists platform-specific video-decoding requirements. Follow the installation instructions for your selected revision. [5]

A useful first result does not require buying an arm

The repository documents dataset loading and simulation evaluation. Our suggested starting point is an existing dataset: inspect its observations and actions to understand what a later hardware experiment will need to record. The documented hardware sequence is teleoperate, record, train and deploy. [1] [6]

Budget for the experiment, not just the package

Training compute and storage depend on the policy and dataset. A physical setup adds parts, assembly and calibration. We have not established a cost for a specific experiment.

We have not installed this revision or reproduced these workflows.

What we checked

Drafting and source review used Codex, followed by a separate AI editorial check. No independent human technical review. No affiliate links or paid placement.

Not yet established

  • A limitations summary
  • Requirements: operating system, host computer
  • A step-by-step setup summary
  • Version-specific hardware compatibility (ecosystem tags alone do not establish it)

Updates

Upstream last updated
Sep 29, 2026 [2]
Latest release
Aug 3, 2026 [4]
Profile updated
Oct 4, 2026
Sources last checked
Sep 29, 2026
Next review
Dec 28, 2026
Reviewed revision
e0d5021 · software repository

Source files and licenses

Library source is available under the cited root license. Related model/data licenses and documentation/media rights are not fully assessed; inclusion does not certify the entire ecosystem.

LayerAvailability / license
softwareavailable · Apache-2.0
Root license; individual notices and dependencies retain their own terms. [1] [3]
model weightsavailable · License not fully assessed
Published artifacts exist. Licenses and suitability must be checked on each selected artifact; no blanket license is assigned. [6]
datasetsavailable · License not fully assessed
Published artifacts exist. Licenses and suitability must be checked on each selected artifact; no blanket license is assigned. [6]
documentationavailable · License not fully assessed
Documentation source is available; no separate blanket documentation/media license determination is made in this profile. [1] [5]

What the license does—and does not—cover

The software repository carries Apache-2.0 plus third-party notices. That license does not establish the terms for models and datasets. Check the selected artifact's card for license information before reuse. [3] [6]

References

Each source links to the exact revision we read. “Used for” shows where this page relies on it.

Project sources

Videos

Correction guidance · Discussion is not enabled · Profile format 0.5.0-draft.1

Change history

2026-09-30 — Edited the opening and removed repeated qualifications; clarified source labels, configuration costs and further reading. Source-check dates are unchanged.

2026-10-04 — Added 3 YouTube reference(s) (S007, S008, S009): official channel or videos linked from the pinned upstream sources. Existing claims, source-check dates and evidence are unchanged.