Skip to main content
rfx has a minimal CLI focused on the complete robot learning workflow: 
rfx record   # Record demonstrations from hardware
rfx train    # Train a policy from collected data
rfx deploy   # Deploy a trained policy to a robot
rfx doctor   # Check your setup
All commands work with uv run if you don’t have rfx installed globally: 
uv run rfx record --robot so101 --repo-id demos --episodes 10
uv run rfx train --data demos/ --config train.yaml
uv run rfx deploy runs/my-policy --robot so101
uv run rfx doctor

​
Design Philosophy

The CLI is designed for the foundation model era workflow:
  1. Collect demos with rfx record
  2. Train a policy with rfx train
  3. Deploy to hardware with rfx deploy
  4. Iterate
Each command does exactly one thing and does it well.

​
Core Commands

record

Collect teleoperation demonstrations into LeRobot datasets

train

Train a policy from collected demonstration data

deploy

Load a trained policy and run it on hardware

doctor

Check system dependencies and hardware

​
Quick Examples

​
Record 10 demonstration episodes

rfx record --robot so101 --repo-id my-org/demos --episodes 10

​
Deploy a policy from a local checkpoint

rfx deploy runs/my-policy --robot so101

​
Deploy a policy from HuggingFace Hub

rfx deploy hf://rfx-community/go2-walk-v1 --robot go2 --duration 60

​
Test a policy without hardware

rfx deploy runs/my-policy --mock

​
Check your environment

rfx doctor

​
Python API

Every CLI command has a Python equivalent: 
import rfx

# Record
rfx.collection.collect("so101", "my-org/demos", episodes=10)

# Deploy
rfx.deploy("runs/my-policy", robot="so101")
rfx.deploy("hf://user/policy", robot="go2", duration=30)

​
Getting Help

Every command supports --help: 
rfx --help
rfx record --help
rfx deploy --help
rfx doctor --help

​
Version Information

rfx --version

​
Additional Commands

While the core workflow uses three commands, rfx also provides a runs subcommand for querying the training run registry: 
rfx runs list --limit 20
rfx runs show <run_id>
See the workflow documentation for details on training and managing runs.

Build docs developers (and LLMs) love

Get started for freeTalk to us