MemTensor/MemOS

Cross-references: mentioned by 1 indexed plugin →

★ 12.1k GitHub stars · TypeScript · Memory & Context

About 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.

MemOS is one of 29 projects listed under Memory & Context on DSHToolkit. Memory layers, context managers and retrieval tools that give DSH agents persistent recall.

Installation

git clone https://github.com/MemTensor/MemOS.git
cd MemOS
# follow the README for install steps

Most DeepSeek Harness plugins install as npm packages or DSH extensions. After installing, restart DSH so the plugin registers. If the project ships a desktop client or a plugin market, you can install it with one click instead.

Install and configure MemOS

# try the package directly if it is published
npx MemOS

# otherwise work from source
git clone https://github.com/MemTensor/MemOS.git
cd MemOS
npm install
npm run build

This is a TypeScript project. If MemOS is published on npm, npx is the fastest path. If it is not, build from source — most DSH plugins in this language expect a build step before the runtime can load them.

Configuration

// dsh.config.js — typical registration shape
export default {
  plugins: [
    // confirm the exact export name in the README before using this
    { name: "MemOS" },
  ],
};

Registration shapes vary between plugins. The snippet above shows the common case — " copy the exact name from the README rather than trusting this example verbatim.

Is it the right choice?

• Choose it if you want the path other people already took. At 12,053 stars this is one of the most adopted projects in the directory, so problems are likely to already have answers.
• Licensing is Apache-2.0, which is permissive enough to vendor into a commercial product without a legal review.
• Stack: written in TypeScript, tagged agent, agentic-ai, ai, ai-agents. If your team does not work in that language, every fix becomes a research task.

Things that commonly go wrong

• Version drift. DSH is young and moves fast. A plugin that worked three months ago may target an API that has since changed. Check the last commit date before you plan around it.
• No sandbox guarantee. Plugins run with the runtime's permissions. Read the source of anything that touches your filesystem, shell or API keys before installing it.
• Single maintainer risk. Most projects here are one person's side work. That is fine for a utility, risky for infrastructure you depend on.

Star history

Recorded on 4 days. Over that window the project moved +0 stars (+0.0%).

Window	Change
2026-10-24 → 2026-10-25	+0
2026-10-25 → 2026-10-26	+0
2026-10-26 → 2026-10-28	+0

These numbers are our own daily snapshots, not an estimate. A directory that only shows a single static star count hides whether a project is alive.

Details

Language	TypeScript
License	Apache-2.0
Stars	12,053
Last updated	2026-10-23
Category	Memory & Context
Homepage	https://memos.openmem.net

Topics

FAQ

What is MemOS?

MemOS is a DeepSeek Harness plugin by MemTensor. 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.

How do I install MemOS?

Clone or install from the repository at https://github.com/MemTensor/MemOS, follow the README install steps, then restart DeepSeek Harness. Plugins are usually npm packages or DSH extensions.

Is MemOS free?

MemOS is published under the Apache-2.0 license. Check the repository for the current terms.

How popular is MemOS?

MemOS has 12,053 GitHub stars and was last updated on 2026-10-23.

Related plugins