CaP-X Agentic Robotics

LLM agents that write Python code to control real robots — from zero-shot manipulation to RL-trained coding agents with sim-to-real transfer.

Based on CaP-X (NVIDIA, Berkeley, Stanford, CMU). arXiv: 2603.22435. MIT License.

Quick Start

1. Clone and Install

git clone https://github.com/capgym/cap-x.git
cd cap-x

# Requires Python 3.10 (BEHAVIOR) or 3.12 (RL/LIBERO)
uv sync          # uses uv (Astral) for package management

2. Start Perception Services