Tech.brewed

RecallMatrix is an AI-powered long-term memory engine that solves the problem of forgotten information for both humans and AI agents. Every day we absorb thousands of data points—notes, chats, ideas, links, screenshots—and both humans and AI assistants lose them within minutes. RecallMatrix acts as a persistent memory layer by capturing user inputs (text, links, and media), converting them into embeddings, storing them with context, and retrieving them through natural-language queries using RAG. It demonstrates an autonomous agent capable of organizing, indexing, and intelligently recalling memories without manual effort. The full system runs end-to-end: users store a memory → embeddings are created → relationships form in a knowledge graph → the agent retrieves the exact memory when queried.

Key demos include:

Saving ideas, notes, links, and files in real time.

Universal semantic search retrieving forgotten memories (“What startup idea did I write last week?”).

The MCP agent connecting RecallMatrix with external tools like ChatGPT, Gmail, or voice assistants to auto-update or fetch memories.

A live Knowledge Graph showing how the user’s memories connect across people, topics, and contexts.

Dashboard showing embedding activity, logs, and memory growth.