MemOS

What MemOS does

MemOS unifies store, retrieve and manage operations for long-term memory. The README lists a unified memory API structured as an inspectable graph, multi-modal memory covering text, images, tool traces and personas, multi-cube knowledge base management, asynchronous ingestion through MemScheduler, and natural-language feedback that corrects or replaces memories over time.

For Hermes Agent, the README news section describes an official local memory plugin launched in April 2026 with hybrid retrieval (FTS5 plus vector), smart dedup, tiered skill evolution and multi-agent collaboration, running fully locally with no cloud dependency. In May 2026 it announced memos-local-plugin 2.0, an official local plugin for Hermes Agent and OpenClaw built on L1 traces, L2 policies, L3 world models and crystallized Skills. OpenClaw also has cloud and local plugins, and a DeepSeek Harness plugin adds automatic recall and a local Memory Viewer.

Key features

• Unified memory API for adding, retrieving, editing and deleting memory
• Multi-modal memory for text, images, tool traces and personas
• Multi-cube knowledge base management with isolation and controlled sharing
• Asynchronous ingestion through MemScheduler
• Official local memory plugins for Hermes Agent and OpenClaw
• Natural-language feedback to correct stored memories

When to use it

• Giving Hermes Agent local, persistent memory with hybrid retrieval
• Sharing memory between several agents through separate knowledge cubes
• Building assistants and support bots that recall earlier context

Who it is for: Developers and agent users who want a graph-based, editable long-term memory system with ready-made agent plugins.

How it fits with Hermes Agent

The README announces official MemOS memory plugins for Hermes Agent: a local plugin in April 2026 and memos-local-plugin 2.0 for Hermes Agent and OpenClaw in May 2026. The repository is also tagged hermes.

FAQ