Agent Memory is an MCP server designed to empower AI agents with robust, persistent, and cross-session memory capabilities. It enables agents to store, retrieve, search, and summarize various types of information across different interactions and agent boundaries, ensuring long-term knowledge retention and continuity. The server includes a suite of tools for remembering facts, preferences, interactions, and learned insights, alongside advanced functionalities to aggregate user profiles, summarize past sessions, and perform comprehensive searches across all agents. By classifying memories into distinct types like `fact`, `preference`, `interaction`, `learning`, and `context`, Agent Memory facilitates structured and highly effective knowledge management crucial for developing sophisticated and context-aware AI applications.
Key Features
01Store and classify memories with optional time-to-live (TTL).
02Aggregate knowledge about a user or entity from all connected agents.
03Delete memories using exact keys, wildcard prefixes, or clear all.
04Retrieve memories by keyword search, sorted by relevance and recency.
05Search and access memories comprehensively across all managed agents.
060 GitHub stars
