# mujocolab/mjlab

Isaac Lab API, powered by MuJoCo-Warp, for RL and robotics research

Repository: https://github.com/mujocolab/mjlab
Canonical: https://ross.abutalabs.com/products/mjlab
Homepage: https://mujocolab.github.io/mjlab/
Language: Python
License: Apache-2.0
License Family: permissive
Topics: isaaclab, mujoco, mujoco-warp, reinforcement-learning, robotics-simulation
Last push: 2026-08-26T10:02:01+00:00

## Health v2 (maintenance only)
Score: 85/100 (v2, computed 2026-09-03T02:39:23.370411+00:00)
- activity 99, release rhythm 96, longevity 32
- inputs: {"age_days": 450, "days_push": 7, "days_rel": 25, "gap_med": 16.5, "n_releases_24m": 11}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2837, forks 501 (observed 2026-08-28T04:07:24.601823+00:00)

## What it is
mjlab is a Python framework that combines Isaac Lab's manager-based API with MuJoCo Warp, a GPU-accelerated version of MuJoCo, for reinforcement learning and robotics research. It provides composable building blocks for environment design with minimal dependencies and direct access to native MuJoCo data structures.

## Use cases
- train a humanoid robot to follow velocity commands with reinforcement learning
- simulate thousands of robot environments in parallel on GPU
- train a humanoid to imitate reference motions
- run multi-GPU distributed RL training for robotics
- evaluate trained robot policies from checkpoints
- prototype robot learning environments using an Isaac Lab-style API without NVIDIA Isaac Sim

## When to choose
- you want Isaac Lab's API ergonomics but prefer MuJoCo physics over Isaac Sim
- you need fast GPU-parallel robot simulation for RL training
- you work with humanoid robots like the Unitree G1 and need velocity tracking or motion imitation tasks
- you want minimal dependencies and direct access to native MuJoCo data structures
- you need multi-GPU scaling for large-scale robot learning experiments

## When to avoid
- you need CPU-only simulation, since training requires an NVIDIA GPU
- you rely on Isaac Sim-specific features like ray-traced rendering or Omniverse integrations
- you need non-robotics physics simulation outside MuJoCo's supported feature set
- you need a long-established, battle-tested framework rather than a fast-moving research project

## Facets
- artifact type: framework
- maturity: active
- function: simulation, machine-learning, reinforcement-learning, gpu-computing, robotics
- domain: robotics, reinforcement-learning, simulation, machine-learning, gpu-computing
- platform: windows, python
- tags: mujoco, mujoco-warp, isaac-lab, robot-learning, gpu-simulation, humanoid-robots, motion-imitation, sim-to-real, linux, macos, gpu, docker

## Member repositories
- mujocolab/mjlab (main) score 85

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:24.601823+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:37:45.612836+00:00, confidence not recorded.
  - readme: https://github.com/mujocolab/mjlab (fetched 2026-08-28T04:07:24.601823+00:00, sha c9484f1ccce8)
  - homepage: https://mujocolab.github.io/mjlab/ (fetched 2026-08-29T09:53:05.567108+00:00, sha 44136fa355b3)
  - registry_pypi: https://pypi.org/pypi/mjlab/json (fetched 2026-08-29T09:53:05.569414+00:00, sha 26f7e217a712)
- Data as of 2026-08-30T08:39:29.467469+00:00.