What it does

The Nex MemPalace Memory System provides a sophisticated, long-term semantic memory architecture for AI agents. It transforms a standard Raspberry Pi or Linux server into a centralized "Memory Palace" using a structured hierarchy of wings (projects/people), halls (facts/preferences), and drawers (verbatim logs). Built on ChromaDB and SQLite, it enables agents to recall past conversations, project details, and user preferences with high precision across separate sessions.

Why use this skill

Standard LLM memory is often transient or limited to simple key-value pairs. This skill implements a full Memory Protocol that instructs your agent to search for context before responding and store insights after interacting. By running as an MCP (Model Context Protocol) server, it allows multiple bots (Telegram, OpenClaw, or CLI) to share a unified knowledge base. It is specifically optimized for ARM64 architecture, ensuring smooth deployment on Raspberry Pi hardware.

Supported tools

• Database: ChromaDB (Vector) and SQLite (Relational)
• Frameworks: MCP (Model Context Protocol), OpenClaw, Telegram Bot API
• Hardware: Raspberry Pi (ARM64), Linux Servers
• Languages: Python-based venv deployment

The Output

The skill generates a complete deployment package including an idempotent shell script for automated Pi installation, a small "Identity Layer" for session-start persona loading, and specific JSON configurations for your memory wings and keyword routing. It also provides a dedicated 'Memory Protocol' document that teaches your agent the logic of when to search and when to store.