NevaMind-AI/memU

Health v2 · maintenance only

Flags: no_license

Adoption not part of the score

14425 stars · 1069 forks observed · 2026-09-23

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

memU is a lightweight, agent-driven memory system that gives LLM agents a shared, inspectable personal memory layer across sessions, agents, and devices, stored as readable Markdown wiki files. It automatically distills reusable skills from agent history and supports both a cloud backend and self-hosting with SQLite or Postgres/pgvector.

Use cases

• give my coding agent persistent memory across sessions
• share context between Claude Code, Cursor, and Codex
• extract reusable skills from my agent chat history
• self-host an agent memory store with embeddings
• remember project preferences across devices
• inject relevant past decisions before new agent tasks

When to choose

• you use multiple AI coding agents and want one shared memory layer
• you want memory stored as human-readable Markdown you can inspect and edit
• you want automatic skill extraction from agent history
• you want a self-hosted memory backend with SQLite or Postgres

When to avoid

• you need a general-purpose vector database for application data
• you need a fully supported memory API for a production product rather than agent-driven workflows
• your agent is not among the supported hosts (e.g. Claude chat mode is unsupported)

Facets

library · maturity active

agent-framework rag mcp cli sdk large-language-models developer-tools python windows cli self-hosted cross-platform agent-memory memory-layer skill-extraction claude-skills markdown-memory cross-agent sqlite pgvector openclaw ai-agents macos linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="NevaMind-AI/memU")