Engram

Engram is a persistent memory layer designed for AI agents. It operates on the GitHub platform and utilizes semantic similarity to retrieve relevant context from stored observations. The agent includes a self-pruning mechanism based on a time-to-live (TTL) setting. Engram is written in TypeScript and supports integration with technologies like ChromaDB and Claude. It provides a way to maintain and access historical data for AI agents.

Engram solves the problem of managing and retrieving context for AI agents over time. Manually tracking and searching through observations becomes impractical as agent interactions increase, making Engram a useful tool for maintaining agent memory.