Open-Source Memory for the Agents You Already Use
One shared, inspectable memory across sessions, agents, and devices.
Host adapters capture new sessions, your agent decides what is worth turning into memory or reusable skills, and memU stores, embeds, and retrieves the Markdown it prepared. The memory service itself makes no LLM or chat calls.
Why memU Is Different
Agent-Driven
Judgment and synthesis stay in the host agent instead of a separate memory LLM pipeline.
Cross-Agent
Connected host adapters share one configured Cloud or local memory backend.
Readable by Default
Memory and skills remain inspectable Markdown files that people and agents can understand.
Small and Adaptable
A compact Python core, pluggable storage, and host-specific adapters keep the system easy to inspect and extend.
One Project, Two Deployment Modes
The memU repository contains the memory service, CLI, installation skill, and host adapters. Choose one shared backend, then connect the agents you already use.
memU Cloud
Use a managed, cross-device backend. Get an API key from memu.so and let the installation skill configure each host adapter.
Self-Hosted
Keep memory on your own machine or infrastructure with SQLite or Postgres. Configure an embedding provider, then run the same host adapters locally.
Host Adapters
Dedicated CLIs support Codex, Claude Code, Cursor, OpenClaw, Hermes, and WorkBuddy. The generic adapter detects compatible agents.
Join the Growing memU Open-Source Community
For more information please contact info@nevamind.ai.
GitHub Issues
Report bugs, request features, and track development. Submit an issue.
Discord
Get real-time support, chat with the community, and stay updated. Join us.
X (Twitter)
Follow for updates, AI insights, and key announcements. Follow us.
Give Every Agent the Same Memory
Start with the installation skill, connect one host, then add other agents to the same shared backend.