Engram empowers AI agents with a robust, persistent memory system, overcoming the common issue of context loss after conversations. It integrates advanced hybrid search techniques, combining BM25 keyword matching, semantic embeddings, and a knowledge graph, to ensure comprehensive and nuanced recall. Modeled on biological memory, Engram incorporates temporal decay and memory strengthening, allowing AI to remember important information longer while consolidating older, less crucial data. Designed as a local-first MCP server, it keeps user data private and offers a suite of tools for storing, recalling, and managing AI's evolving understanding of its world.
Key Features
01Comprehensive MCP Toolset for memory management (remember, recall, forget, create_entity)
02Temporal Decay & Memory Strengthening based on Ebbinghaus forgetting curve
03Local-first data storage for enhanced privacy and performance
043 GitHub stars
05Hybrid Search (BM25 + Semantic Embeddings + Knowledge Graph)
06Memory Consolidation with AI-generated summaries and contradiction detection
Use Cases
01Enhancing AI agents with long-term, contextual conversational memory.
02Allowing AI assistants to remember user preferences, facts, and relationships across interactions.
03Building AI systems that can connect disparate pieces of information for deeper understanding and smarter responses.
