HubOS

Multi-User AI Employee Management Platform
多用户 AI 员工管理平台

🌟 What is HubOS?

HubOS is an open-source, self-hosted AI employee management platform. It transforms large language models into a team of specialized digital employees that communicate through your existing messaging tools.

It's not a chatbot — it's an AI workforce.

Each agent has its own identity, skills, memory, and job responsibilities, coordinated by a central dispatcher (General Manager agent). You talk to them through Feishu, WeChat, Discord, or any of 14+ supported channels.

Why HubOS?

Problem	HubOS Solution
ChatGPT/Claude are single-agent, general-purpose	9 specialized agents with division of labor
SaaS AI tools store your data in the cloud	Everything runs locally, data never leaves your machine
One chatbot can't handle complex business workflows	Multi-agent orchestration with parallel/pipeline/DAG modes
AI doesn't learn from experience	Self-evolving memory system extracts lessons automatically
Every AI tool is a separate subscription	One platform, 24 built-in skills, 18 tools

Key Numbers

Metric	Value
Code	150,000+ lines (Python + TypeScript)
API Endpoints	170+
Built-in Tools	18
Skills	24
Messaging Channels	14+
Agent Roles	9
Frontend Pages	28
Languages	4 (EN / ZH / JA / RU)

✨ Features

🔥 What Makes HubOS Different

Most AI platforms are single-agent chatbots. HubOS is fundamentally different:

Feature	Description
Multi-Task Concurrency	Multiple users can chat simultaneously — each request spawns an independent agent session via FastAPI async. No queuing, no blocking.
Agent Instance Pooling	LRU cache reuses initialized agent instances across requests. Second request for the same agent is ~200× faster (no skill/MCP re-registration).
Sub-Agent Spawning	Any agent can independently spawn child agents (`spawn_subagents`) for parallel work — and those children can spawn their own children (up to configurable depth).
DAG Orchestration	Complex multi-step workflows with dependencies: step B uses step A's output, step C+D run in parallel after B. Full pipeline coordination.
Self-Evolving Memory	After each task, the system automatically reflects on what happened, extracts lessons, and merges them into reusable methodology cards. The more you use it, the smarter it gets.
Multi-User Isolation	Each user gets independent workspaces, sessions, and memory. File locks (`fcntl.flock`) prevent write conflicts. Agent sandboxing prevents cross-user access.
Hot-Swappable Models	Per-agent model configuration — assign GPT-4o for complex reasoning, Claude for writing, Gemini for multimodal tasks, or local Ollama models for privacy. Switch models without restarting.
Channel Multiplexing	One agent team serves all channels simultaneously. A message on WeChat and a message on Discord go to the same agent with the same memory.

🤖 Multi-Agent Team

9 specialized agents, each with its own personality, skills, and model configuration:

Agent	Department	Role
HubOS	General Manager	Task dispatch, coordination, decision-making
Sam	Sales	Lead generation, outreach, quotations
Mavis	Marketing	Content creation, competitive analysis, branding
Iris	Research	Market research, data analysis, industry reports
Rex	R&D	Development, system maintenance, automation
Felix	Finance	Billing, financial reports, cost analysis
Harper	HR	Documentation, scheduling, process management
Clara	Customer Success	Support, feedback handling, FAQ maintenance
Oscar	Operations	Task tracking, scheduled jobs, monitoring

💡 Fully Customizable: The 9 built-in agents are just a starting point. You can create any number of agents with custom roles, personalities, skills, and model configurations to match your specific business needs — from a solo assistant to a 50-person digital company.

Three orchestration modes:

Parallel Delegation (spawn_subagents) — Independent tasks run simultaneously
Pipeline Coordination (coordinate_workflow) — Sequential DAG with dependencies
Background Tasks (delegate_task) — Long-running jobs with progress tracking

📡 14+ Messaging Channels

All channels use a unified message format. One agent serves multiple channels simultaneously:

Console (built-in web UI) · Feishu · WeChat · DingTalk · Discord · Telegram · WeCom · QQ · Matrix · Mattermost · iMessage · MQTT · XiaoYi · Voice

🧠 Three-Layer Memory System

Layer	Loading	Purpose
Long-term Memory (MEMORY.md)	Auto-loaded every session	Persistent knowledge, tool configs, lessons learned
Work Experience v4	Auto-retrieved by task type	Methodology cards — one card per workflow
Daily Notes (memory/)	On-demand search	Detailed session logs, troubleshooting records

Work Experience v4 — Self-Evolving Engine:

Automatically reflects after task completion → extracts lessons → merges into cards
LLM semantic matching replaces keyword matching for higher accuracy
Promotion pipeline: candidate → approved → mature
Users choose their own reflection model via the UI

🛠️ 24 Built-in Skills

Category	Skills
Document Processing	PDF · Word (.docx) · Excel (.xlsx) · PowerPoint (.pptx)
Web & Search	Web crawling · Tavily search · Browser automation
Communication	Channel messaging · Email (himalaya) · Multi-agent collaboration
Business	E-commerce price search · News aggregation · Cron scheduling
System	HubOS setup guide · Frontend design · File reading
Platform Integration	Feishu (Bitable/Doc/Wiki/Drive) · DingTalk channel setup

🔒 Security

Feature	Description
File Locking	`fcntl.flock` per-file granularity, 30s timeout — prevents write conflicts
Agent Sandbox	Each agent can only write to its own workspace
Write Whitelist	Configurable external directory access (e.g., R&D agent → project dir)
Tool Guard	Risk-level tool control with human-in-the-loop approval
RBAC	Role-based access control (admin / user / viewer)
JWT Auth	Web API and WebSocket authentication

⚡ Performance

Feature	Implementation
Async Architecture	FastAPI + Uvicorn, each request as independent asyncio.Task
Agent Instance Pooling	LRU cache with concurrency-safe borrowing, ~200× speedup on repeat requests
200K Context Window	GLM-5.1 supports 200K tokens; auto-compression when exceeded
SSE Heartbeat	15s ping intervals to prevent proxy timeouts

🆕 What's New in v1.1.0

Feature	Description
🎮 Real-Time Run Guidance	Interrupt and redirect agents mid-task — type guidance while they work and they pivot immediately
📊 Task Workflow Visualization	Watch multi-agent tasks execute in real-time with a full-page monitor and chat-side panel
🔧 Status Messages	Context understanding & experience matching now shown as 🔧 tool-call indicators before each response
⌨️ Slash Commands	24 built-in commands with EN/ZH labels, accessible via `/` in chat input
⚡ Faster Response	WE card generation moved to background thread, context compression uses dedicated compact model
🛡️ Reliability	Fixed async `await` bug, timeout floor guarantee, instant SSE release, friendly cancel prompts

📋 Full Changelog

New Modules: RunControl API · TaskMonitor (backend + frontend) · TaskPlan (plan/autogen/executor/risk) · Slash Commands · UI Language

Improvements: Compact model config · WE async fire-and-forget · Retriever fuzzy matching · Reflection prompt rewrite · Timeout max() semantics · MCP Streamable HTTP · Absolute stdio paths · Desktop port-based restart

Fixes: memory_compaction AttributeError · memory.add() missing await · AGENT_ERROR on cancel · Dark mode text · delegate_task timeout override · SSE heartbeat blocking

📸 Screenshots

Coming soon — screenshots of the Web Console, Agent Management, and Chat interfaces.

🚀 Quick Start

Prerequisites

Python 3.10 - 3.13
Node.js 18+ (for frontend build)
At least one LLM API key (Zhipu GLM, MiniMax, OpenAI, Anthropic, etc.)

From Source

# Clone
git clone https://github.com/hubos-ai/HubOS.git
cd HubOS

# Set up virtual environment
python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate

# Install Python dependencies
pip install -e .

# Build frontend
cd console
npm ci
npm run build
cd ..

# Initialize (creates ~/.hubos/ directory)
hubos init

# Start the server
hubos app

Open http://localhost:8088 and start chatting.

First-Time Setup

Open http://localhost:8088
Register an account
Go to Settings → Models and configure your LLM provider (API key + model)
Go to Settings → Agents and verify agent configurations
Start chatting with your AI team!

🏗️ Architecture

┌─────────────────────────────────────────────────────────┐
│                    User Interface                        │
│  Web UI │ Electron Desktop │ Feishu │ WeChat │ ...      │
└──────────────────────┬──────────────────────────────────┘
                       │
┌──────────────────────▼──────────────────────────────────┐
│               FastAPI Backend (Async)                    │
│  ┌──────────┐  ┌───────────┐  ┌──────────┐  ┌────────┐ │
│  │ Auth/RBAC │  │ Sessions  │  │ Tracker  │  │ Locks  │ │
│  └──────────┘  └───────────┘  └──────────┘  └────────┘ │
└──────────────────────┬──────────────────────────────────┘
                       │
┌──────────────────────▼──────────────────────────────────┐
│              Agent Dispatch Layer (GM)                   │
│  ┌────────────────────────────────────────────────────┐ │
│  │ Agent Instance Pool (LRU + concurrency-safe)       │ │
│  └────────────────────────────────────────────────────┘ │
│  Parallel │ Pipeline │ Background   ← Task Modes       │
└──────────────────────┬──────────────────────────────────┘
                       │
┌──────────────────────▼──────────────────────────────────┐
│              9 Specialized Agents                        │
│  Sales │ Marketing │ Research │ R&D │ Finance │ ...     │
│  Each has: Identity + Skills + Memory + Model Config    │
└──────────────────────┬──────────────────────────────────┘
                       │
┌──────────────────────▼──────────────────────────────────┐
│              Infrastructure                              │
│  Skills │ LLM Router │ Memory │ WE v4 │ MCP │ Cron     │
│  Security │ DAG Engine │ Tool Guard │ File Guard        │
└─────────────────────────────────────────────────────────┘

Directory Structure

src/hubos/
├── app/                    # Application layer
│   ├── runner/             # Agent runners (with instance pooling)
│   ├── routers/            # 22 API router modules
│   ├── channels/           # 14+ channel adapters
│   ├── mcp/                # MCP client management
│   └── workspace/          # Workspace lifecycle
├── agents/                 # Agent core
│   ├── react_agent.py     # HubOSAgent (ReAct reasoning)
│   ├── tools/              # 18 built-in tools
│   ├── skills/             # 24 skill definitions
│   ├── hooks/              # Agent lifecycle hooks
│   └── memory/             # Memory managers
├── config/                 # Configuration system (45+ classes)
├── core/                   # Core engine
│   ├── llm/                # LLM provider routing
│   ├── memory/             # Memory storage (ChromaDB)
│   ├── work_experience/    # Experience system (v4)
│   ├── execution/          # Task execution engine
│   ├── dag/                # DAG scheduling engine
│   └── infra/              # RBAC, feature flags, metrics
├── security/               # Tool guard, skill scanner
└── cli/                    # CLI commands

console/                    # React frontend (28 pages)
desktop/                    # Electron desktop app

Tech Stack

Component	Technology
Backend	Python 3.10+ · FastAPI · Uvicorn
Agent Framework	AgentScope 1.0.18
Frontend	React 18 · Vite · TypeScript · Ant Design 5
Desktop	Electron
Vector Store	ChromaDB (SQLite backend)
UI Components	@agentscope-ai/chat · @agentscope-ai/design

🤝 Supported LLM Providers

HubOS supports multiple LLM providers with per-agent model configuration:

Provider	Models	Best For
Zhipu (智谱)	GLM-4, GLM-5.1	Complex reasoning, long context (200K)
MiniMax	M2.7-highspeed	Fast structured tasks
OpenAI	GPT-4o, GPT-4-turbo	General purpose
Anthropic	Claude 3.5/4	Analysis, writing
Google	Gemini 1.5/2.0	Multimodal tasks
DeepSeek	DeepSeek-V3	Cost-effective reasoning
Qwen (通义千问)	Qwen 2.5/3	Chinese language tasks
Ollama	Local models	Privacy-first, no API needed

🔌 MCP (Model Context Protocol)

HubOS supports MCP client integration with hot-reload:

Configure MCP servers via Web UI or config files
Each skill can define its own MCP dependencies
Built-in support for Zhipu MCP Server (Vision, Search, webReader, ZRead)

📋 CLI Reference

hubos app          # Start FastAPI server
hubos init         # Initialize working directory
hubos agent        # Agent management & communication
hubos cron         # Cron job management
hubos channel      # Channel configuration
hubos models       # Model provider configuration
hubos skills       # Skill management
hubos auth         # Authentication management
hubos shutdown     # Stop all services

🗺️ Roadmap

Plugin marketplace for community skills
Multi-language agent templates (DE, FR, ES, AR)
GPU-accelerated local model support (MLX, llama.cpp)
Collaborative workspaces for teams
Mobile companion app
REST API playground / Swagger UI

🤝 Contributing

We welcome contributions! Please see CONTRIBUTING.md for:

Development environment setup
Code style guidelines
PR submission process
How to add new skills, channels, and tools

📄 License

HubOS is released under the Apache License 2.0.

🙏 Acknowledgments

HubOS builds upon the following excellent open-source projects:

Project	Usage	License
AgentScope	Core agent framework, model routing, MCP integration	Apache-2.0
AgentScope Runtime	Frontend UI components (@agentscope-ai/chat, @agentscope-ai/design)	Apache-2.0
FastAPI	Async web framework	MIT
React	Frontend UI library	MIT
Ant Design	UI component library	MIT
Playwright	Browser automation	Apache-2.0
ChromaDB	Vector storage	Apache-2.0
ReMe	Long-term memory management	MIT
Uvicorn	ASGI server	BSD-3-Clause
Electron	Desktop application framework	MIT

Special thanks to the open-source community and everyone who contributes to making AI more accessible.

🌟 HubOS 是什么？

HubOS 是一个开源、自托管的 AI 员工管理平台。它将大语言模型转化为一支专业的数字员工团队，通过你已有的沟通工具与你协作。

它不是聊天机器人 — 它是 AI 劳动力。

每个 Agent 有自己的身份、技能、记忆和工作职责，由总经理 Agent 统一调度。你可以通过飞书、微信、钉钉等 14+ 渠道与团队沟通。

为什么需要 HubOS？

问题	HubOS 的解决方案
ChatGPT/Claude 是单 Agent、通用型	9 个专业 Agent，分工明确
SaaS AI 工具把数据存在云端	完全本地运行，数据不离开你的机器
单个聊天机器人无法处理复杂业务流程	多 Agent 编排：并行/流水线/DAG 三种模式
AI 不会从经验中学习	自进化记忆系统，自动提取教训
每个 AI 工具都是独立订阅	一个平台，24 个内置技能，18 个工具

✨ 核心特性

🔥 HubOS 的独特之处

大多数 AI 平台是单 Agent 聊天机器人。HubOS 从根本上不同：

特性	说明
多任务并发	多个用户可以同时对话 — 每个请求通过 FastAPI 异步机制创建独立的 Agent 会话。无需排队，互不阻塞。
Agent 实例池化	LRU 缓存复用已初始化的 Agent 实例。同一 Agent 的第二次请求快 ~200 倍（无需重新注册 skill/MCP）。
子 Agent 派生	任何 Agent 都可以独立派生子 Agent（`spawn_subagents`）并行工作 — 子 Agent 还可以继续派生自己的子 Agent（深度可配置）。
DAG 工作流编排	复杂的多步骤依赖工作流：步骤 B 使用步骤 A 的输出，步骤 C+D 在 B 完成后并行执行。完整的流水线协调。
自进化记忆	每次任务完成后，系统自动反思、提取教训，合并为可复用的方法论卡片。越用越聪明。
多用户隔离	每个用户拥有独立的 workspace、会话和记忆。文件锁（`fcntl.flock`）防止写入冲突。Agent 沙箱防止跨用户访问。
模型热切换	按 Agent 配置模型 — 复杂推理用 GPT-4o，写作用 Claude，多模态用 Gemini，隐私场景用 Ollama 本地模型。用户自由选择供应商和模型，无需重启即可切换。
渠道多路复用	一个 Agent 团队同时服务所有渠道。微信和 Discord 上的消息访问同一个 Agent 和同一套记忆。

🤖 多 Agent 团队协作

9 个专业 Agent，各有自己的人格、技能和模型配置：

Agent	部门	职责
HubOS	总经理	任务调度、协调、决策
Sam 张	销售	客户开发、报价、合同推进
Mavis 王	市场	内容创作、竞品分析、品牌传播
Iris 周	调研	市场调研、数据分析、行业报告
Rex 陈	技术	开发、系统维护、自动化
Felix 刘	财务	账单、财务报表、成本分析
Harper 赵	人力	文档管理、日程协调、流程规范
Clara 孙	客服	售后支持、反馈处理、FAQ 维护
Oscar 吴	运维	任务追踪、定时作业、监控

💡 完全可定制：内置的 9 个 Agent 只是起点。你可以创建任意数量的 Agent，自定义角色、人格、技能和模型配置，匹配你的业务需求 — 从单人助手到 50 人的数字公司都可以。

三种协作模式：

并行委派（spawn_subagents）— 独立任务同时执行
流水线协作（coordinate_workflow）— 有依赖的串行 DAG 工作流
后台任务（delegate_task）— 长时间运行的任务，带进度追踪

📡 14+ 渠道接入

所有渠道统一消息格式，同一个 Agent 同时服务多个渠道：

Web 控制台（内置）· 飞书 · 微信 · 钉钉 · Discord · Telegram · 企业微信 · QQ · Matrix · Mattermost · iMessage · MQTT · 小蚁 · 语音

🧠 三层记忆系统

层	加载方式	用途
长期记忆（MEMORY.md）	每次自动加载	持久化知识、工具配置、经验教训
经验技巧（Work Experience v4）	按任务类型自动检索	方法论卡片 — 一卡一流程
每日笔记（memory/）	按需搜索	详细会话日志、踩坑记录

Work Experience v4 — 自进化引擎：

任务完成后自动反思 → 提取教训 → 合并去重到现有卡片
LLM 语义匹配替代关键词匹配，准确度大幅提升
晋升管线：candidate → approved → mature
用户可在前端自选反思模型

🛠️ 24 个内置技能

类别	技能
文档处理	PDF · Word · Excel · PowerPoint
网络搜索	网页抓取 · Tavily 搜索 · 浏览器自动化
沟通协作	渠道消息 · 邮件 · 多 Agent 协作
业务工具	电商比价 · 新闻聚合 · 定时任务
系统管理	安装指南 · 前端设计 · 文件读取
平台集成	飞书（多维表格/文档/知识库/云盘）· 钉钉接入

🔒 安全体系

特性	说明
文件写锁	`fcntl.flock` per-file 粒度，防止多用户写入冲突
Agent 沙箱	每个 Agent 只能写自己的 workspace
工具防护	按风险级别控制工具调用，支持人机协同审批
RBAC	基于角色的访问控制（admin / user / viewer）
JWT 认证	Web API 和 WebSocket 认证

🆕 v1.1.0 更新

功能	说明
🎮 实时引导打断	任务执行中随时发指令，Agent 立即转向 — 像和真人同事协作一样
📊 任务流程可视化	全页任务监控 + 聊天内侧边面板，多 Agent 协作进度实时可见
🔧 状态消息	上下文理解、经验匹配以 🔧 工具调用样式显示在每次响应前
⌨️ 快捷命令	24 个内置命令，中英双语，输入 `/` 即可呼出
⚡ 响应更快	经验卡片后台异步生成，上下文压缩使用独立轻量模型
🛡️ 稳定性提升	修复 async await 丢失、超时覆盖、取消报错等关键 bug

📋 完整更新日志

新模块：运行控制 API · 任务监控（后端+前端）· 任务计划（生成/自动/执行/风控）· 快捷命令 · UI 语言模块

改进：压缩模型独立配置 · 经验卡片异步生成 · 检索器模糊匹配 · 反思 prompt 重写 · 超时 max() 语义 · MCP Streamable HTTP · stdio 绝对路径 · 桌面端按端口重启

修复：memory_compaction AttributeError · memory.add() 漏 await · 取消引导显示 AGENT_ERROR · 深色模式文字不可见 · delegate_task 超时覆盖 · SSE 心跳阻塞

🚀 快速开始

前提条件

Python 3.10 - 3.13
Node.js 18+（前端构建需要）
至少一个 LLM API key（智谱 GLM、MiniMax、OpenAI、Anthropic 等）

从源码安装

git clone https://github.com/hubos-ai/HubOS.git
cd HubOS

python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate

pip install -e .

cd console
npm ci
npm run build
cd ..

hubos init
hubos app

打开 http://localhost:8088 开始使用。

首次配置

打开 http://localhost:8088
注册账号
进入 设置 → 模型，配置 LLM 供应商（API key + 模型）
进入 设置 → Agent，确认 Agent 配置
开始和你的 AI 团队对话！

🤝 支持的 LLM 供应商

HubOS 支持多种 LLM 供应商，每个 Agent 可独立配置模型：

供应商	模型	适用场景
智谱（Zhipu）	GLM-4, GLM-5.1	复杂推理、长上下文（200K）
MiniMax	M2.7-highspeed	快速结构化任务
OpenAI	GPT-4o, GPT-4-turbo	通用场景
Anthropic	Claude 3.5/4	分析、写作
Google	Gemini 1.5/2.0	多模态任务
DeepSeek	DeepSeek-V3	高性价比推理
通义千问	Qwen 2.5/3	中文场景
Ollama	本地模型	隐私优先，无需 API

🏗️ 系统架构

┌─────────────────────────────────────────────────────────┐
│                    用户接入层                              │
│  Web 控制台 │ 桌面应用 │ 飞书 │ 微信 │ 钉钉 │ 14+ 渠道    │
└──────────────────────┬──────────────────────────────────┘
                       │
┌──────────────────────▼──────────────────────────────────┐
│               FastAPI 后端 (异步)                         │
│  认证/权限 │ 会话管理 │ 任务追踪 │ 文件锁                 │
└──────────────────────┬──────────────────────────────────┘
                       │
┌──────────────────────▼──────────────────────────────────┐
│              Agent 调度层 (总经理)                         │
│  Agent 实例池 (LRU 缓存 + 并发安全)                       │
│  并行委派 │ 流水线协作 │ 后台任务                          │
└──────────────────────┬──────────────────────────────────┘
                       │
┌──────────────────────▼──────────────────────────────────┐
│              9 个专业 Agent (数字员工)                      │
│  每个 Agent: 身份 + 技能 + 记忆 + 模型配置                 │
└──────────────────────┬──────────────────────────────────┘
                       │
┌──────────────────────▼──────────────────────────────────┐
│              基础设施层                                    │
│  技能 │ LLM 路由 │ 记忆系统 │ 自进化 │ MCP │ 定时任务     │
│  安全防护 │ DAG 引擎 │ 工具防护 │ 文件防护                 │
└─────────────────────────────────────────────────────────┘

技术栈

组件	技术
后端	Python 3.10+ · FastAPI · Uvicorn
Agent 框架	AgentScope 1.0.18
前端	React 18 · Vite · TypeScript · Ant Design 5
桌面端	Electron
向量存储	ChromaDB (SQLite)
UI 组件	@agentscope-ai/chat · @agentscope-ai/design

📄 许可证

HubOS 基于 Apache License 2.0 开源。

🙏 致谢

HubOS 基于以下优秀开源项目构建：

项目	用途	许可证
AgentScope	核心 Agent 框架、模型路由、MCP 集成	Apache-2.0
AgentScope Runtime	前端 UI 组件 (@agentscope-ai/chat, @agentscape-ai/design)	Apache-2.0
FastAPI	异步 Web 框架	MIT
React	前端 UI 库	MIT
Ant Design	UI 组件库	MIT
Playwright	浏览器自动化	Apache-2.0
ChromaDB	向量存储	Apache-2.0
ReMe	长期记忆管理	MIT
Uvicorn	ASGI 服务器	BSD-3-Clause
Electron	桌面应用框架	MIT

特别感谢开源社区的每一位贡献者，是你们让 AI 变得更加普惠。

Name		Name	Last commit message	Last commit date
Latest commit History 81 Commits
.github		.github
assets		assets
console		console
deploy		deploy
desktop		desktop
docs		docs
experiments		experiments
logo		logo
scripts		scripts
src/hubos		src/hubos
templates		templates
tests		tests
.dockerignore		.dockerignore
.flake8		.flake8
.gitattributes		.gitattributes
.gitignore		.gitignore
.gitignore.lock		.gitignore.lock
.pre-commit-config.yaml		.pre-commit-config.yaml
.python-version		.python-version
CHANGELOG.md		CHANGELOG.md
CODE_OF_CONDUCT.md		CODE_OF_CONDUCT.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
OPEN_SOURCE_CHECKLIST.md		OPEN_SOURCE_CHECKLIST.md
README.md		README.md
README.md.lock		README.md.lock
SECURITY.md		SECURITY.md
docker-compose.yml		docker-compose.yml
pyproject.toml		pyproject.toml

Folders and files

Latest commit

History

Repository files navigation

HubOS

🌟 What is HubOS?

Why HubOS?

Key Numbers

✨ Features

🔥 What Makes HubOS Different

🤖 Multi-Agent Team

📡 14+ Messaging Channels

🧠 Three-Layer Memory System

🛠️ 24 Built-in Skills

🔒 Security

⚡ Performance

🆕 What's New in v1.1.0

📸 Screenshots

🚀 Quick Start

Prerequisites

From Source

First-Time Setup

🏗️ Architecture

Directory Structure

Tech Stack

🤝 Supported LLM Providers

🔌 MCP (Model Context Protocol)

📋 CLI Reference

🗺️ Roadmap

🤝 Contributing

📄 License

🙏 Acknowledgments

🌟 HubOS 是什么？

为什么需要 HubOS？

✨ 核心特性

🔥 HubOS 的独特之处

🤖 多 Agent 团队协作

📡 14+ 渠道接入

🧠 三层记忆系统

🛠️ 24 个内置技能

🔒 安全体系

🆕 v1.1.0 更新

🚀 快速开始

前提条件

从源码安装

首次配置

🤝 支持的 LLM 供应商

🏗️ 系统架构

技术栈

📄 许可证

🙏 致谢

About

Topics

Resources

License

Code of conduct

Contributing

Security policy

Uh oh!

Stars

Watchers

Forks

Releases 2

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages