MindMemOS
MindMemOS is a self-evolving memory operating layer for AI agents that enables persistent, portable memory across different agent platforms. It allows agents to accurately remember user preferences, project context, and task history, then reuse that information across multiple sessions and different agent systems like OpenClaw, DeepSeek Harness, and others. The system automatically distills experiences into reusable skills, consolidates memories through offline dreaming processes, and connects with file-based knowledge systems. Developers can integrate MindMemOS through HTTP APIs, a Python SDK, or agent-specific plugins that automatically recall and write memories during conversations.
Key Features
Use Cases
- 01Sharing user preferences and project context across multiple AI coding assistants without re-explaining requirements
- 02Enabling long-term memory for algorithm design agents to accumulate and reuse cross-task experience and domain knowledge
- 03Building conversational agents that remember user profiles, preferences, and interaction history across sessions
- 04Evolving agent skills automatically based on real usage patterns, execution results, and user feedback
- 05Consolidating fragmented knowledge from local files and project artifacts into searchable memory for agent retrieval
- 06Creating multi-agent systems where different specialized agents share a common memory layer
Related Skills
View moresuperpowers
An agentic skills framework & software development methodology that works.
skills
Skills for Real Engineers. Straight from my .agents directory.
skills
Public repository for Agent Skills
ponytail
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
MindMemOS — FAQ
What is MindMemOS?+
MindMemOS is a portable memory operating layer for AI agents that persists user preferences, project facts, and task experience as reusable assets. It enables different agent platforms (OpenClaw, DeepSeek Harness, Claude Code, etc.) to share the same long-term memory and automatically evolves that memory through ongoing interactions.
How do I install MindMemOS for OpenClaw?+
First install the Python SDK with 'pip install mindmemos-sdk' and run 'mindmemos auth' to configure your connection. Then install the OpenClaw plugin with 'openclaw plugins install @mindmemos/openclaw-plugin' and enable it with 'openclaw plugins enable mindmemos-memory'. You must also grant write permissions with 'openclaw config set plugins.entries.mindmemos-memory.hooks.allowConversationAccess true'.
Which AI clients and agents does MindMemOS work with?+
MindMemOS currently provides official plugins for OpenClaw and DeepSeek Harness, with integration support mentioned for Claude Code, OpenHands, and LLM4AD_NEXT. Any agent or application can integrate through the HTTP API or Python SDK.
Do I need API keys or other prerequisites?+
For the cloud service, you need an API key from the MindMemOS website. For local self-hosting, you need Docker, Python with uv, and at least three configured model routers (chat, embedding, and reranking models) in your dev.yaml configuration file.
Is MindMemOS free to use?+
MindMemOS is open source under the MIT License. The cloud service offers free access with automatic Pro quota upgrade when you star the GitHub repository, while local self-hosting is entirely free but requires your own infrastructure and model API access.
How does memory consolidation and dreaming work?+
MindMemOS automatically consolidates memories offline through a dreaming process that merges related memories, resolves conflicts, and improves retrieval quality. This runs asynchronously and achieves up to 92% accuracy on fact consolidation benchmarks with GPT-5-mini, with 23.5% of memories successfully archived.
How do I install MindMemOS?+
Open the source repository on GitHub and follow its README. MindMemOS is a skill — MCP Agents Market links you directly to the official repo.
Is MindMemOS free?+
MindMemOS is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.