Agent Memory

About

Agent Memory RS offers a sophisticated memory management system designed for LLM agents, built on modern AI research and cognitive science. It provides three distinct memory types: episodic memory for interaction events, semantic memory for consolidated knowledge, and procedural memory for learned workflows. Key capabilities include auto-consolidation of memories, intelligent decay based on composite scoring, hybrid search, and hierarchical retrieval, ensuring agents have efficient access to relevant information through its MCP server support and CLI tools.

Key Features

• Intelligent Decay (composite scoring for memory archival)
• Hierarchical Retrieval (multi-level memory access from synopsis to archived)
• MCP Server (Model Context Protocol integration for AI assistants)
• Hybrid Search (BM25 keyword + vector semantic search with RRF fusion)
• Auto-Consolidation (nightly pattern extraction, daily synopsis generation)
• 0 GitHub stars

Use Cases

• Enabling LLM agents to learn and adapt workflows over time
• Providing persistent, context-aware memory for long-running AI applications
• Integrating advanced memory capabilities into custom AI assistants