Skill Transfer Platform

Develop Transferable
Robot Skills

Train on one robot, deploy on many. FoodforThought is the platform for building, labeling, and sharing robot skills that transfer across different robot platforms.

How Robot Skills Are Made

Train once, deploy everywhere. Here's how FoodforThought turns robot demonstrations into reusable skills.

1

Record & Upload

Record your robot doing a task (pick up mug, fold towel, etc.)

ate upload video.mp4 --robot franka --task pick_place
2

Community Labels

Contributors mark actions: "approach object", "grasp", "lift", "move", "release"

0:02 grasp
0:04 lift
0:06 move
3

Export Skill

Download labeled data for ML training in your preferred format

ate pull pick_place --format rlds
4

Check Transfer

See if your Franka skill works on UR5, Unitree, or other robots

ate check-transfer --from franka --to ur5
50+
Skills Developed
12+
Robots Supported
85%+
Avg Transfer Rate
100+
Contributors

Why FoodforThought?

The infrastructure you need to develop, share, and transfer robot skills

Cross-Robot Transfer

See exactly how skills transfer between robots with our compatibility scoring system

Community Labeling

Leverage gamified data labeling to turn demonstrations into training-ready data

Fork & Extend

Start from community skills and adapt them to your robot and use case

Use Skills in Your Code

pip install kindly-robotics
from kindly import load_skill, check_compatibility

# Pull labeled skill data for training
skill = load_skill("pick_place", robot="franka-panda", format="rlds")

# Check transfer compatibility before deploying
compat = check_compatibility(
    skill_name="pick_place",
    source_robot="franka-panda",
    target_robot="unitree-r1"
)

print(f"Transfer score: {compat.overallScore * 100:.1f}%")
print(f"Adaptation type: {compat.adaptationType}")

Ready to Build Transferable Skills?

Join researchers and developers building the future of robot skill sharing.

© 2025 Kindly Robotics