Agent memory for LLM agents: 7 neuroscience-inspired layers (working, episodic, semantic, procedural) with FSRS-style spaced repetition and memory consolidation. TypeScript, no third-party dependencies in the core packages; Model Context Protocol (MCP) server, Vercel AI SDK middleware. Paper: arXiv 2604.23878 · reproduction packages on Zenodo.
This repository also configures its own agents. See what zenbrain tells them →
Plugin Claude Code
Memory for Claude Code that lasts across sessions: the ZenBrain MCP server with five tools (store, recall, forget, consolidate, health) over one local SQLite file. No account, no network call of its own.