Engram — a brain-inspired context database for AI agents

Engram is a brain-inspired context database that stores AI agent knowledge as atomic "bullets" in a concept graph, allowing any agent to share context across different LLMs like Claude, GPT, or Gemini. The system mimics human memory through reinforcement learning, where useful knowledge bullets grow stronger while unhelpful ones fade away through reconsolidation and active forgetting. Unlike traditional agent frameworks that lose context through repeated summarization, Engram's delta operations ensure context persists across sessions and improves with every use.