This new learn guide is inspired by Olivia Zhu’s A Minimal Self-Perceiving Embodiment for Large Language Models (2026, zenodo, GitHub).

The core idea of the project is to give an LLM agent access to a device that can be used to sense the environment that its human user is in, and to express itself in that environment using sensor feedback loops to confirm the physical effects. This version of the project uses an Adafruit ESP32-S3 Reverse TFT Feather as the central hardware.

The feedback loop is what makes this ’embodied’ compared to just a bot with sensors because it will be ‘grounded’ in the reality of being able to actuate-then-sense and thus have verifiable behaviors instead of just an ‘agent with sensors’ that can drift from reality.

The guide includes demonstrations, assembly instructions, software setup, and usage pages. Read more at LLM Agent Embodiment Kit