# MemTensor/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 Harness support.

Repository: https://github.com/MemTensor/MemOS
Canonical: https://ross.abutalabs.com/products/memtensor-memos
Homepage: https://memos.openmem.net
Language: TypeScript
License: Apache-2.0
License Family: permissive
Topics: agent, llm, memory, long-term-memory, memory-management, rag, skills, openclaw, agentic-ai, ai, ai-agents, chatgpt, claude, hermes, self-evolving, self-hosted, token-savings, mcp, deepseek-harness, dsh-plugin
Last push: 2026-08-26T17:18:47+00:00

## Health v2 (maintenance only)
Score: 85/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 99, release rhythm 99, longevity 30
- inputs: {"age_days": 423, "days_push": 7, "days_rel": 9, "gap_med": 6.5, "n_releases_24m": 43}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 11012, forks 1009 (observed 2026-08-28T04:10:45.858072+00:00)

## What it is
MemOS is a memory operating system for LLMs and AI agents that unifies storing, retrieving, and managing long-term, multi-modal memory through a single API. It supports graph-structured, editable memory, multi-cube knowledge bases, hybrid retrieval, and cross-task skill reuse, with self-hosted and cloud deployment options.

## Use cases
- add persistent long-term memory to my AI agent
- reduce token usage in LLM chat applications
- give my chatbot personalized memory across sessions
- manage multiple knowledge bases for different users and projects
- self-host a memory backend for LLM apps
- connect memory to agent frameworks via MCP
- store and retrieve tool traces and multi-modal memories

## When to choose
- you need production-grade, persistent, multi-modal memory for LLM apps or agents
- you want an editable, inspectable memory graph rather than a black-box embedding store
- you need self-hosted deployment with data control or a managed cloud option
- you want token savings and hybrid retrieval across long conversations

## When to avoid
- you only need simple in-process caching or short-lived context windows
- you want a fully turnkey product with no integration work
- your project has no LLM or agent component

## Facets
- artifact type: framework
- maturity: active
- function: rag, vector-database, agent-framework, mcp, llm-inference, search-engine, sdk
- domain: large-language-models, databases, developer-tools, self-hosted
- platform: python, self-hosted, cross-platform, cloud
- tags: long-term-memory, memory-management, memory-os, token-savings, agentic-ai, hybrid-retrieval, knowledge-base, multi-modal-memory, personalization, deepseek-harness, ai-agents, retrieval-augmented-generation, typescript, docker

## Member repositories
- MemTensor/MemOS (main) score 85

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:45.858072+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:17:13.056975+00:00, confidence not recorded.
  - readme: https://github.com/MemTensor/MemOS (fetched 2026-08-28T04:10:45.858072+00:00, sha 46e19d3c3f2e)
  - homepage: https://memos.openmem.net (fetched 2026-08-29T08:16:09.491816+00:00, sha 00b378c9fb91)
  - site_page: https://memos-docs.openmem.net/ (fetched 2026-08-29T08:16:09.503184+00:00, sha d726357390bc)
  - site_page: https://memos.openmem.net/pricing (fetched 2026-08-29T08:16:09.501243+00:00, sha b30f102bd26c)
- Data as of 2026-08-30T08:39:29.467469+00:00.