mjviser 0.0.4

Web-based MuJoCo viewer powered by viser.

mjviser

Web-based MuJoCo viewer powered by Viser.

Quick start

View any MuJoCo model instantly, no install needed:

uvx mjviser path/to/model.xml

Or install and use:

pip install mjviser
mjviser path/to/model.xml

Fuzzy path matching finds models in the current directory tree:

mjviser humanoid        # finds **/humanoid*.xml
mjviser shadow_hand     # finds **/shadow_hand*.xml

When there's a single match it's used automatically. Multiple matches show a numbered list to pick from.

Browse 57 robots from robot_descriptions without cloning anything:

uvx --with robot_descriptions mjviser go1
uvx --with robot_descriptions mjviser shadow_hand
uvx --with robot_descriptions mjviser aloha

Python API

import mujoco
from mjviser import Viewer

model = mujoco.MjModel.from_xml_path("robot.xml")
data = mujoco.MjData(model)
Viewer(model, data).run()

Open the printed URL in your browser. Most of the MuJoCo native viewer's functionality is available: simulation controls, joint and actuator sliders, contact and force visualization, camera tracking, keyframes, and more.

Extension points

The Viewer accepts three callbacks for injecting custom logic:

• step_fn(model, data): Called each simulation step. Use this to apply a controller, external forces, or any per-step logic. Defaults to mujoco.mj_step.

• render_fn(scene): Called each render frame. Use this to push custom state to the scene, add ghost overlays, or render debug geometry. Defaults to scene.update_from_mjdata(data).

• reset_fn(model, data): Called on reset. Use this to restore custom simulation state.

For full control over the loop, use ViserMujocoScene directly. The server is a standard viser server, so you can add custom GUI elements, scene overlays, or anything else viser supports.

server = viser.ViserServer()
scene = ViserMujocoScene(server, model, num_envs=1)
scene.create_visualization_gui()

# Add your own GUI.
with server.gui.add_folder("My Controls"):
    slider = server.gui.add_slider("Force", min=0, max=100, initial_value=0)

while True:
    mujoco.mj_step(model, data)
    scene.update_from_mjdata(data)

Examples

• active_viewer.py: simplest usage with playback controls
• active_viewer_with_controller.py: custom step_fn with random torques
• passive_viewer.py: manual simulation loop with ViserMujocoScene
• multi_env.py: 4 humanoids in parallel via mujoco-warp
• ghost_overlay.py: custom render_fn that overlays a time-delayed ghost
• motion_playback.py: recorded trajectory with timeline scrubber, speed control, and contact replay

Limitations

• No mouse interaction: clicking/dragging bodies and keyboard callbacks are not yet supported; this requires upstream support in viser.
• Performance with many bodies: models with lots of independently-moving bodies can be slower than the native MuJoCo viewer due to per-body websocket message overhead.
• Cubemap textures: approximated via per-vertex colors rather than true cubemap rendering.