Repolode
Submit
mindscale-noah

mindscale-noah/MindMemOS

A memory layer for AI agents that stores user and task context, shares it across agents, and distills experience into reusable Skills.

View on GitHubWebsite
Stars
918
Forks
100
Open issues
0
Last commit
4 days ago

What is it?

MindMemOS is a memory service for AI agents. It persists user profiles, preferences, project facts, and tool experience as long-term memory that can be searched and injected into later interactions, so an agent does not start from zero in each session. Memories are exposed through an HTTP API, a Python SDK with a CLI, and plugins for OpenClaw and the DeepSeek Harness (dsh) agent framework.

The system includes a schema-based memory extraction flow, offline consolidation (dreaming), and feedback-driven optimization of add and search. Experience memories can be distilled into Skill candidates, and skill execution results and user feedback feed back into memory. It can run as a local self-hosted stack with Qdrant, Neo4j, and Kafka via Docker, or be used through the official cloud service with an API key.

Key features

  • 01Stores user profiles, preferences, and project facts as reusable long-term memory
  • 02Exposes add, search, and feedback operations over HTTP, Python SDK, and CLI
  • 03Ships plugins for OpenClaw and DeepSeek Harness that auto-recall and write back memories
  • 04Distills experience memories into Skill candidates and supports Skill evolution
  • 05Runs fully self-hosted with Docker, or against the official cloud service

Use cases

  • Giving a coding or personal agent memory of user preferences across sessions
  • Sharing the same long-term memory between different agents
  • Storing project facts and tool experience for agents working on recurring tasks
  • Running an agent memory backend on-premises or offline

In collections

Open-source AI agent frameworks & appsAI agent skill libraries

Topics

#agent-memory#ai-agents#rag#skills#llm#openclaw-plugin#python#agent#agent-skills#agentic

FAQ

What is MindMemOS?
A memory layer for AI agents that stores user and task context, shares it across agents, and distills experience into reusable Skills. MindMemOS is a memory service for AI agents. It persists user profiles, preferences, project facts, and tool experience as long-term memory that can be searched and injected into later interactions, so an agent does not start from zero in each session. Memories are exposed through an HTTP API, a Python SDK with a CLI, and plugins for OpenClaw and the DeepSeek Harness (dsh) agent framework.
Is MindMemOS free to use? What license does it have?
MindMemOS has no declared license. Its code is public, but without a license you have no explicit permission to reuse it; ask the author before relying on it.
Is MindMemOS still maintained?
Actively maintained. The last commit to MindMemOS was on Oct 19, 2026.
What can I use MindMemOS for?
Giving a coding or personal agent memory of user preferences across sessions Sharing the same long-term memory between different agents Storing project facts and tool experience for agents working on recurring tasks Running an agent memory backend on-premises or offline
Where can I get MindMemOS?
The source code and installation instructions are on GitHub: https://github.com/mindscale-noah/MindMemOS

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects