Built by open-source people.
Learning robot skills moves faster when the tools are shared. The core neuracore library is MIT-licensed and free, and our team builds and maintains teleoperation tools, robot examples, and manipulation benchmarks in the open, on top of the community standards we already rely on.
MIT-licensed core
The neuracore SDK and CLI are free and open. Read the code, fork it, and see exactly how your data is handled.
Standing on the community
We build on the open standards roboticists already use, and contribute back the tools and benchmarks we depend on.
Examples, not black boxes
Runnable examples for real robots and a Colab notebook, so you can reproduce the whole workflow before you commit.
Repositories we build and back
Fork it, star it, send a PRThe core library and CLI: stream data logging from any robot, visualize datasets, launch cloud training, and run policy inference locally or on an endpoint.
View on GitHub ↗Read hand position and button state from a Meta Quest headset to teleoperate robots and collect VR demonstrations straight into Neuracore.
View on GitHub ↗A worked example of collecting demonstrations and training a policy on AgileX hardware with the neuracore SDK, from robot setup to inference.
View on GitHub ↗End-to-end example on the low-cost, open-source SO-101 arm: teleoperate, log demonstrations, train, and deploy a learned skill.
View on GitHub ↗A mobile bi-manual, demonstration-driven benchmark for imitation learning and offline reinforcement learning, built on MuJoCo with human-collected demonstrations across everyday tasks.
View on GitHub ↗Start with one line: pip install neuracore
Clone a repo, run the Colab notebook, or read the docs. Everything you need to log your first demonstration is open.