Build ToddlerBot, a low-cost open humanoid for learning

Humanoid, Hard

ToddlerBot from Stanford is a low-cost, open-source humanoid for scalable policy learning, with sim, teleop and RL locomotion code.

View build on GitHub ↗Share on X

The repo includes MuJoCo sim, data collection via teleoperation, and policy training. Follow-up work (Locomotion Beyond Feet, 2026) builds on it.

Hardware & cost: Dynamixel servos, a Jetson, printed parts; see the paper and docs for the full BOM.

Difficulty: Hard.

Agent install prompt (paste into Claude Code, Cursor or Codex): "Set up https://github.com/hshi74/toddlerbot following its docs, run the MuJoCo simulation, and play back a walking policy."

Source: hshi74/toddlerbot on GitHub — https://github.com/hshi74/toddlerbot (facts checked against the project README, Oct 2026; prices vary by region).

Quick answers

What do you need to build ToddlerBot, a low-cost open humanoid for learning?: Dynamixel servos, a Jetson, printed parts; see the paper and docs for the full BOM.
How hard is it to build ToddlerBot, a low-cost open humanoid for learning?: Hard.

How your agent drives it

Copied from the hshi74/toddlerbot README

MuJoCo

Shell

$ python toddlerbot/policies/run_policy.py --policy replay --path motion/push_up_2xc.lz4 --vis view

Shell

$ python toddlerbot/locomotion/train_mjx.py --gin-file <skill_name> --env <skill_name>

Any coding agent that can run a shell can run these. Read the README before you run anything on a robot that moves.

Ideas to try with it

Our ideas, not the builder's claims.

Replay the push-up motion in simulation before you build
Train a new skill in MuJoCo and test it on the real robot
Use teleop to record a wave and replay it

1 more idea

Have your agent run training jobs and summarize which skills improved

Source: hshi74/toddlerbot on GitHub
Last commit: 2026-07-31
Builder: hshi74
Repo: hshi74/toddlerbot, 800 stars
Built with: ToddlerBot

What to build next

Builds that share a board, robot or agent with this one.