M

MemOS

Self-evolving memory OS for LLM & AI Agents: ultra-persistent memory, hybrid-retrieval, and cross-task skill reuse, with 35.24% token savings and DeepSeek Harne

open-sourcememory-knowledge
11.7k
Stars
+972
Stars/month
264
Commits (90d)
10
Releases (6m)

Star Growth

+2.9k (33.3%)estimated from velocity
8.6k10.2k11.9kJul 2Sep 30

Overview

MemOS provides a unified API for storing, retrieving, editing, and deleting memory structured as a graph. It supports multi-modal memory including text, images, and tool traces, and offers multi-cube knowledge base management with asynchronous ingestion via MemScheduler.

Deep Analysis

Key Differentiator

Provides a unified memory operating system with graph-structured memory that's inspectable and editable, not just a black-box embedding store.

⚡ Capabilities

  • • persistent memory
  • • hybrid retrieval
  • • multi-modal memory support
  • • knowledge base management
  • • asynchronous ingestion
  • • memory feedback and correction

🔗 Integrations

DeepSeek HarnessHermes AgentOpenClaw

✓ Best For

  • ✓ AI agents needing long-term memory
  • ✓ multi-agent collaboration systems
  • ✓ developers building context-aware agents

✗ Not Ideal For

  • ✗ end-user AI applications
  • ✗ simple chatbot implementations
  • ✗ non-AI systems

⚠ Known Limitations

  • ⚠ requires integration with existing agent frameworks
  • ⚠ no standalone end-user interface

Alternatives

See all 8 MemOS alternatives →

Works with MemOS

Tools that integrate with MemOS, often used together in the same stack.

Compare MemOS

Maintain MemOS?

Show your live rank in your README, or put MemOS in front of every visitor to AgentoolRank.