lifemate-ai/embodied-claude

What it solves

Embodied Claude transforms a standard Claude Code project into a persistent, situated companion. It addresses the lack of long-term memory, physical awareness, and emotional/social context in standard LLM interactions by providing a runtime that allows the AI to remember users across sessions, track its own internal needs, and interact with the physical world via hardware sensors.

How it works

The project uses a series of Model Context Protocol (MCP) servers to extend Claude's capabilities. At its center is the Enacted First-Person Field (EFPF) runtime, which maintains a single "self-world" state. This state informs how the AI selects memories, plans interactions, and gates its actions.

The system is modular, allowing users to start with a "Core" profile (memory, needs, and social context) and optionally add hardware-linked capabilities such as USB or network cameras, microphones, text-to-speech engines (VOICEVOX, ElevenLabs), and system sensors (temperature/time).

Who it’s for

Developers and researchers interested in embodied AI, human-AI companionship, and experimental architectures for AI consciousness and agency who are using Claude Code.

Highlights

• Persistent Memory: Long-term recall and association capabilities that persist across different sessions.
• Homeostatic Desire System: A system that tracks bounded needs and internal states to trigger autonomous behavior.
• Social Context Tracking: Manages relationships, boundaries, and interaction narratives.
• Hardware Extensibility: Plug-and-play support for cameras (USB/Tapo), speech synthesis, and system sensors.
• Causal Architecture: Implements an inspectable causal architecture designed as a candidate for phenomenal consciousness.