Equips AI agents with an intelligent memory layer to learn, consolidate, and recall context effectively.
Sponsored
Engram acts as a sophisticated intelligence layer for AI agents, transforming them from 'amnesiac' to perpetually learning entities. Unlike traditional memory solutions that merely store and retrieve facts, Engram excels at detecting behavioral patterns, consolidating insights, and surfacing relevant context without explicit prompting. Its unique architecture prioritizes intelligence at read time, offering three tiers of memory—Explicit, Implicit, and Synthesized—to provide agents with a comprehensive and adaptive understanding of their interactions and tasks, demonstrated by its strong performance on the LOCOMO benchmark.
主要功能
01MCP Server integration for Claude Code, Cursor, and any MCP client
02Comprehensive REST API and TypeScript SDK for flexible agent integration
03Model-agnostic support for various LLMs (Gemini, OpenAI, Ollama, Groq, Cerebras)
04Zero infrastructure requirement using SQLite for easy deployment
05LLM-powered memory consolidation, contradiction detection, and pattern discovery
0626 GitHub stars
使用案例
01Improving context recall and relevance for large language model applications
02Enhancing AI agents with persistent, intelligent memory capabilities
03Developing AI assistants capable of learning, consolidating information, and detecting contradictions
