Awesome AI AgentsAI Agent Operating System

MemTensor/MemOS

⭐ 11308 Python added to this list on 2025-07-30 repository created 2025-07-06

MemOS is an innovative operating system designed specifically for Large Language Models (LLMs) to enhance their capabilities with long-term memory. This system enables LLMs to store, retrieve, and manage information effectively, which results in more context-aware, consistent, and personalized interactions. MemOS introduces a modular memory architecture called MemCube, which supports multiple types of memory including textual memory for unstructured or structured text knowledge, activation memory for caching key-value pairs to speed up inference and context reuse, and parametric memory for storing model adaptation parameters like LoRA weights. The system provides a unified API for memory operations, facilitating seamless integration with LLMs to improve chat and reasoning tasks through memory-augmented generation (MAG). MemOS demonstrates significant performance improvements over baseline memory solutions in various reasoning tasks, especially excelling in temporal reasoning with an accuracy improvement of 159% compared to OpenAI's baseline. The project supports extensibility, allowing users to customize memory modules, data sources, and LLM integrations. It is cross-platform, supporting Linux, macOS, and Windows, and is available as a Python package. The project also offers comprehensive documentation, API references, and community support through GitHub discussions and Discord. Installation is straightforward via pip, with optional dependencies for additional features like tree memory, memory reader, and memory scheduler. MemOS also supports integration with external tools such as Ollama and the transformers library for enhanced functionality. Overall, MemOS represents a significant advancement in the field of AI memory systems, providing a robust framework for improving the intelligence and usability of LLMs through sophisticated memory management.

https://github.com/MemTensor/MemOS

activation-memoryagentaiai-memory-systemapicontext-awarecross-platformextensiblekv-cachelanguage-modellarge-language-modelsllmllm-memorylong-term-memoryloramagmemcubememorymemory-augmented-generationmemory-managementmemory-operating-systemmemory-retrievalmemory-schedulingmemory-storagememosmodular-memory-architectureneo4jollamaparametric-memorypersonalized-interactionspython-packageragreasoning-tasksretrieval-augmented-generationtemporal-reasoningtextual-memorytransformerstree

Also in AI Agent Operating System

RightNow-AI/openfang

OpenFang is an open-source agent operating system, built in Rust, designed to facilitate the management and orchestration of AI agents.

Arindam200/awesome-ai-apps

A comprehensive repository showcasing practical examples, tutorials, and tools for building powerful AI applications using large language models and various AI agent frameworks.

OpenByteInc/QuantDinger

QuantDinger is an open-source AI trading operating system designed for independent traders, Python strategy authors, and small teams, offering backtesting, live trading, and multi-agent research ca...

holaboss-ai/holaOS

HolaOS is a local-first, AI-powered desktop agent that learns your work context, integrates with over 100 workplace tools, and manages memory efficiently to provide proactive assistance without for...

mcp-use/mcp-use

mcp-use is an open-source unified client library that enables seamless integration of any LangChain-supported LLM with MCP servers to build custom agents with tool access.

astrid-runtime/astrid

Astrid is a portable, capability-secure operating system for composable software components, providing an OS-grade boundary for AI agents and other untrusted code via WebAssembly capsules.

HKUDS/AutoAgent

AutoAgent is a fully-automated, zero-code framework for creating and deploying LLM agents using natural language, featuring top GAIA benchmark performance and broad LLM support.

Q00/ouroboros

Ouroboros is an Agent OS for AI coding, providing a local-first runtime layer to transform non-deterministic agent work into replayable, observable, and policy-bound execution contracts.