TeamClaw

Enterprise OpenClaw AI Agent Management Platform

企业级 OpenClaw AI Agent 管理平台

English | 中文

TeamClaw 是什么？

TeamClaw 是基于 OpenClaw（🦞）构建的全功能管理平台，提供 OpenClaw 目前不具备的企业级能力。

核心功能

AI 对话

多会话管理 — 每个 Agent 支持创建多个独立对话
流式输出 — 逐 Token 实时显示回复内容
思考过程 — 可折叠展示 LLM 的推理链路
图片附件 — 支持发送图片（PNG/JPEG/GIF/WebP，最大 5MB）
上下文管理 — 对话历史快照与上下文重置

Agent 管理

跨实例 Agent 浏览与创建，支持克隆到不同实例
分类体系 — DEFAULT / DEPARTMENT / PERSONAL 三级分类
文件管理 — 树形浏览与在线编辑 Agent 配置文件
可视化配置编辑器 — Schema 驱动的表单，覆盖所有 OpenClaw 模块

Skills 市场

ClawHub 集成 — 从公共市场搜索、安装和更新技能包
技能开发 — IDE 风格的文件编辑器，本地开发后发布到 ClawHub
版本管理 — 安装追踪、版本检查与一键升级
作用域控制 — 支持 PERSONAL / DEPARTMENT / GLOBAL 三级作用域

知识库 (RAG)

文档上传 — 支持 PDF、DOCX、TXT、Markdown，自动 OCR 与分块
语义检索 — 基于 pgvector 向量相似度搜索，支持流式问答
作用域管理 — PERSONAL / DEPARTMENT / GLOBAL 三级知识库隔离
智能复用 — 同一知识库内自动复用已有文档的 OCR 结果

多实例管理

Docker 一键创建 — 配置镜像、端口、绑定即可部署
外部网关接入 — 通过 URL + Token 连接已有 OpenClaw 实例
健康监控 — 60 秒周期检查，自动故障检测与实例恢复
生命周期管理 — 启动、停止、重启，实时日志查看

组织与权限

RBAC 角色体系 — SYSTEM_ADMIN / DEPT_ADMIN / USER 三级权限
部门隔离 — 按部门分配实例和 Agent 访问权限
审计日志 — 全量操作追踪，支持筛选与 CSV 导出

平台能力

完整国际化 — 中英文界面一键切换
移动端适配 — 手机浏览器完整对话体验，侧边栏和文件面板以抽屉形式展开
PWA 支持 — 支持添加到主屏幕，独立窗口运行
多模型支持 — Anthropic、OpenAI、MiniMax、Groq 等
Docker 部署 — 一条命令启动全栈服务

快速开始

⚠️ 重要提示：TeamClaw 依赖 PostgreSQL 和 Redis，不支持单独 docker run 启动。请使用下面的 Docker Compose 方式部署。

方式一：Docker Compose 部署（推荐）

git clone https://github.com/szsip239/teamclaw.git
cd teamclaw
bash setup.sh

脚本会自动：

生成 JWT 密钥对和加密密钥
询问是否启用 Nginx HTTPS 反向代理（可选）
通过 Docker Compose 启动 PostgreSQL、Redis 和 TeamClaw
初始化数据库并创建默认管理员账号

访问 http://localhost:3100 — 账号：admin@teamclaw.local / Admin@123456

启用 HTTPS（可选）

setup.sh 会交互式询问是否启用 Nginx 反向代理。你也可以手动配置：

将 SSL 证书文件放入 nginx/cert/ 目录

编辑 .env 设置 Nginx 变量：

NGINX_SERVER_NAME="your-domain.com"
NGINX_SSL_CERT="your-cert.crt"
NGINX_SSL_KEY="your-cert.key"

使用 --profile nginx 启动：

docker compose -f docker-compose.prod.yml --profile nginx up -d

访问 https://your-domain.com

方式二：本地开发

# 1. 克隆并安装依赖
git clone https://github.com/anthropics/teamclaw.git
cd teamclaw
npm install

# 2. 启动数据库服务
docker compose up -d

# 3. 配置环境变量
cp .env.example .env
node scripts/generate-keys.mjs

# 4. 初始化数据库
npx prisma generate
npx prisma db push
npx tsx prisma/seed.ts

# 5. 启动开发服务器
npm run dev

首次使用指南

部署完成后，按以下步骤开始你的第一次 AI 对话：

1. 登录管理面板

访问 http://localhost:3100，使用默认管理员账号登录：

邮箱：admin@teamclaw.local
密码：Admin@123456

建议首次登录后立即修改默认密码。

2. 配置模型 API 密钥

进入 资源管理 页面，创建模型资源：

提供商	API 协议	说明
Anthropic	`anthropic-messages`	Claude 系列，默认提供商
OpenAI	`openai-completions`	GPT / o 系列
Google	`google-generative-ai`	Gemini 系列
DeepSeek	`openai-completions`	DeepSeek V3 / R1
MiniMax	`anthropic-messages`	Anthropic 兼容协议
Groq / xAI / Mistral	`openai-completions`	OpenAI 兼容协议
Ollama / vLLM	`ollama` / `openai-completions`	本地部署

完整支持 20+ 提供商（智谱、豆包、千帆、Moonshot 等），详见资源管理页面。

点击 创建资源 → 选择提供商 → 填入 API Key → 保存。可勾选"设为默认"供所有实例使用。

3. 部署 OpenClaw 实例

进入 实例管理 页面，选择以下任一方式：

Docker 容器（推荐） — 点击"创建实例"，选择 Docker 模式，填写实例名称，选择镜像（默认 alpine/openclaw:latest），点击创建即可自动部署。

外部网关 — 如果已有运行中的 OpenClaw，选择外部网关模式，填入 WebSocket URL 和 Token 即可连接。

等待实例状态变为 🟢 ONLINE（通常 5-10 秒）。

4. 开始对话

进入 AI 对话 页面 — 实例上线后，默认 Agent 会自动出现在左侧栏。点击 Agent 即可开始对话。

✅ 登录 → 配置 API Key → 创建实例 → 开始对话（约 3 分钟）

系统架构

graph TB
    subgraph Client["浏览器"]
        Chat["AI 对话"]
        AgentUI["Agent 管理"]
        SkillUI["Skills 市场"]
        ConfigUI["配置编辑器"]
        OrgUI["组织管理"]
    end

    NextJS["Next.js 16 (App Router)"]

    subgraph Backend["后端服务"]
        API["REST API (57 个路由)"]
        GW["Gateway Registry<br/>WebSocket 连接池"]
        Health["健康监控<br/>60s 检查 + 自动恢复"]
    end

    subgraph Storage["存储"]
        PG["PostgreSQL 17 + pgvector<br/>14 个数据模型"]
        RD["Redis 7<br/>限流 + 健康计数"]
    end

    RAG["RAG Service<br/>文档解析 + 向量检索"]

    subgraph Instances["OpenClaw 实例"]
        OC1["实例 1 (Docker)"]
        OC2["实例 2 (Docker)"]
        OCN["实例 N (外部网关)"]
    end

    DK["Docker Engine"]

    Client --> NextJS
    NextJS --> API
    API --> GW
    API --> PG
    API --> RD
    API -- "文档/检索" --> RAG
    RAG --> PG
    GW --> Health
    GW -- "WebSocket" --> OC1
    GW -- "WebSocket" --> OC2
    GW -- "WebSocket" --> OCN
    API -. "容器管理" .-> DK
    DK -. "创建/启停" .-> OC1
    DK -. "创建/启停" .-> OC2

    style Client fill:#e3f2fd
    style NextJS fill:#fff3e0
    style Backend fill:#e8f5e9
    style Storage fill:#fce4ec
    style RAG fill:#fff9c4
    style Instances fill:#f3e5f5

技术栈

层级	技术
框架	Next.js 16 (App Router, Turbopack)
前端	React 19, Tailwind CSS 4, shadcn/ui
状态管理	Zustand 5, TanStack Query v5
数据库	PostgreSQL 17 + Prisma 7 (Driver Adapter) + pgvector
RAG	Python FastAPI + pgvector 向量检索
缓存	Redis 7 (ioredis)
认证	RS256 JWT (jose) + bcryptjs
网关通信	WebSocket (ws) + Docker API (dockerode)
数据验证	Zod 4

功能概览

模块	路由数	核心能力
对话	8	多会话、流式输出、思考展示、图片附件
Agent	6	CRUD、克隆、分类、文件管理
Skills	12	ClawHub 市场、安装/发布、版本管理、IDE 编辑
实例	13	Docker 创建、外部接入、健康监控、配置编辑
知识库	8	文档上传、OCR、语义检索、流式问答、作用域管理
认证	5	JWT 登录、Token 轮转、限流
组织	5	用户/部门 CRUD、RBAC 权限
审计	2	操作日志、CSV 导出
仪表盘	1	实例/会话/用户/技能统计
其他	5	资源密钥、实例访问

界面截图

仪表盘	AI 对话
Agent 管理	Skills 管理
实例管理	配置编辑器

贡献

详见 CONTRIBUTING.md，了解开发环境搭建、代码规范和 PR 流程。

许可证

MIT

What is TeamClaw?

TeamClaw is a full-featured management platform built on top of OpenClaw — 🦞the open-source AI Agent gateway🦞. It provides enterprise-grade capabilities that OpenClaw's native dashboard doesn't offer.

Core Features

AI Chat

Multi-conversation — create multiple independent sessions per agent
Streaming responses — real-time token-by-token display
Thinking process — collapsible LLM reasoning chain display
Image attachments — send images with messages (PNG/JPEG/GIF/WebP, max 5MB)
Context management — conversation snapshots and context reset

Agent Management

Cross-instance agent browsing and creation, with cloning to other instances
Classification — DEFAULT / DEPARTMENT / PERSONAL categories
File management — tree view with online editing of agent config files
Visual config editor — schema-driven forms covering all OpenClaw modules

Skills Marketplace

ClawHub integration — search, install, and update skill packages from public marketplace
Skill development — IDE-style file editor, develop locally and publish to ClawHub
Version management — installation tracking, version checks, and one-click upgrades
Scope control — PERSONAL / DEPARTMENT / GLOBAL skill scopes

Knowledge Base (RAG)

Document upload — PDF, DOCX, TXT, Markdown with automatic OCR and chunking
Semantic search — pgvector-powered similarity search with streaming Q&A
Scope management — PERSONAL / DEPARTMENT / GLOBAL knowledge base isolation
Smart reuse — automatically reuse OCR output from previous documents in the same KB

Multi-Instance Management

One-click Docker creation — configure image, ports, bind settings and deploy
External gateway — connect existing OpenClaw instances via URL + token
Health monitoring — 60-second periodic checks with automatic fault detection and recovery
Lifecycle control — start, stop, restart, with real-time log streaming

Organization & Permissions

RBAC — SYSTEM_ADMIN / DEPT_ADMIN / USER three-tier roles
Department isolation — assign instance and agent access per department
Audit logs — comprehensive operation tracking with filtering and CSV export

Platform

Full i18n — English and Chinese interface with one-click switching
Mobile-responsive — full chat experience on mobile browsers with sidebar and file panel as slide-in drawers
PWA support — add to home screen, runs in standalone mode
Multi-model support — Anthropic, OpenAI, MiniMax, Groq, and more
Docker deployment — one-command full-stack setup

Quick Start

⚠️ Important: TeamClaw requires PostgreSQL and Redis. Standalone docker run will NOT work. Use Docker Compose as described below.

Option 1: Docker Compose (Recommended)

git clone https://github.com/szsip239/teamclaw.git
cd teamclaw
bash setup.sh

This will:

Generate JWT keys and encryption secrets
Ask whether to enable Nginx HTTPS reverse proxy (optional)
Start PostgreSQL, Redis, and TeamClaw via Docker Compose
Initialize the database with default admin account

Visit http://localhost:3100 — Login: admin@teamclaw.local / Admin@123456

Enable HTTPS (Optional)

setup.sh interactively asks whether to enable Nginx reverse proxy. You can also configure it manually:

Place SSL certificate files in nginx/cert/

Edit .env to set Nginx variables:

NGINX_SERVER_NAME="your-domain.com"
NGINX_SSL_CERT="your-cert.crt"
NGINX_SSL_KEY="your-cert.key"

Start with --profile nginx:

docker compose -f docker-compose.prod.yml --profile nginx up -d

Visit https://your-domain.com

Option 2: Local Development

# 1. Clone and install
git clone https://github.com/anthropics/teamclaw.git
cd teamclaw
npm install

# 2. Start databases
docker compose up -d

# 3. Configure environment
cp .env.example .env
node scripts/generate-keys.mjs

# 4. Setup database
npx prisma generate
npx prisma db push
npx tsx prisma/seed.ts

# 5. Start dev server
npm run dev

Architecture

graph TB
    subgraph Client["Browser"]
        Chat["AI Chat"]
        AgentUI["Agent Mgmt"]
        SkillUI["Skills Market"]
        ConfigUI["Config Editor"]
        OrgUI["Org Mgmt"]
    end

    NextJS["Next.js 16 (App Router)"]

    subgraph Backend["Backend"]
        API["REST API (57 routes)"]
        GW["Gateway Registry<br/>WebSocket Pool"]
        Health["Health Monitor<br/>60s Check + Auto Recovery"]
    end

    subgraph Storage["Storage"]
        PG["PostgreSQL 17 + pgvector<br/>14 Data Models"]
        RD["Redis 7<br/>Rate Limit + Health Counter"]
    end

    RAG["RAG Service<br/>Doc Parsing + Vector Search"]

    subgraph Instances["OpenClaw Instances"]
        OC1["Instance 1 (Docker)"]
        OC2["Instance 2 (Docker)"]
        OCN["Instance N (External)"]
    end

    DK["Docker Engine"]

    Client --> NextJS
    NextJS --> API
    API --> GW
    API --> PG
    API --> RD
    API -- "Docs/Search" --> RAG
    RAG --> PG
    GW --> Health
    GW -- "WebSocket" --> OC1
    GW -- "WebSocket" --> OC2
    GW -- "WebSocket" --> OCN
    API -. "Container Mgmt" .-> DK
    DK -. "Create/Control" .-> OC1
    DK -. "Create/Control" .-> OC2

    style Client fill:#e3f2fd
    style NextJS fill:#fff3e0
    style Backend fill:#e8f5e9
    style Storage fill:#fce4ec
    style RAG fill:#fff9c4
    style Instances fill:#f3e5f5

Tech Stack

Layer	Technology
Framework	Next.js 16 (App Router, Turbopack)
Frontend	React 19, Tailwind CSS 4, shadcn/ui
State	Zustand 5, TanStack Query v5
Database	PostgreSQL 17 + Prisma 7 (Driver Adapter) + pgvector
RAG	Python FastAPI + pgvector vector search
Cache	Redis 7 (ioredis)
Auth	RS256 JWT (jose) + bcryptjs
Gateway	WebSocket (ws) + Docker API (dockerode)
Validation	Zod 4

Feature Overview

Module	Routes	Key Capabilities
Chat	8	Multi-conversation, streaming, thinking display, image attachments
Agents	6	CRUD, clone, classify, file management
Skills	12	ClawHub marketplace, install/publish, version management, IDE editor
Instances	13	Docker create, external gateway, health monitoring, config editor
Knowledge Base	8	Document upload, OCR, semantic search, streaming Q&A, scope management
Auth	5	JWT login, token rotation, rate limiting
Org	5	User/department CRUD, RBAC
Audit	2	Operation logs, CSV export
Dashboard	1	Instance/session/user/skill metrics
Other	5	Resource keys, instance access

Getting Started Guide

After deployment, follow these steps to start your first AI conversation:

1. Login

Visit http://localhost:3100 and sign in with the default admin account:

Email: admin@teamclaw.local
Password: Admin@123456

Recommended: Change the default password after first login.

2. Configure Model API Key

Navigate to the Resources page to create a model resource:

Provider	API Protocol	Notes
Anthropic	`anthropic-messages`	Claude models, default provider
OpenAI	`openai-completions`	GPT / o series
Google	`google-generative-ai`	Gemini series
DeepSeek	`openai-completions`	DeepSeek V3 / R1
MiniMax	`anthropic-messages`	Anthropic-compatible protocol
Groq / xAI / Mistral	`openai-completions`	OpenAI-compatible protocol
Ollama / vLLM	`ollama` / `openai-completions`	Local deployment

20+ providers supported (OpenRouter, Together, HuggingFace, etc.). See the Resources page for the full list.

Click Create Resource → Select provider → Enter API key → Save. Toggle "Set as default" to use this key for all instances.

3. Deploy an OpenClaw Instance

Navigate to the Instances page and choose one of:

Docker Container (Recommended) — Click "Create Instance", select Docker mode, enter a name, choose an image (default: alpine/openclaw:latest), and create. The container deploys automatically.

External Gateway — If you already have a running OpenClaw, select external gateway mode, enter the WebSocket URL and token to connect.

Wait for instance status to become 🟢 ONLINE (typically 5-10 seconds).

4. Start Chatting

Navigate to the Chat page — once the instance is online, default agents appear in the left sidebar. Click an agent to start your conversation.

✅ Login → Configure API Key → Create Instance → Start Chatting (~3 minutes)

Screenshots

Dashboard	AI Chat
Agent Management	Skills Marketplace
Instance Management	Config Editor

Contributing

See CONTRIBUTING.md for development setup, coding standards, and PR guidelines.

License

MIT

Name		Name	Last commit message	Last commit date
.github		.github
.husky		.husky
docs/screenshots		docs/screenshots
nginx/templates		nginx/templates
postgres/init		postgres/init
prisma		prisma
public		public
rag-service		rag-service
scripts		scripts
src		src
.dockerignore		.dockerignore
.env.example		.env.example
.gitignore		.gitignore
.prettierignore		.prettierignore
.prettierrc		.prettierrc
CONTRIBUTING.md		CONTRIBUTING.md
Dockerfile		Dockerfile
Dockerfile.sale		Dockerfile.sale
LICENSE		LICENSE
README.md		README.md
SECURITY.md		SECURITY.md
components.json		components.json
docker-compose.prod.yml		docker-compose.prod.yml
docker-compose.yml		docker-compose.yml
eslint.config.mjs		eslint.config.mjs
next.config.ts		next.config.ts
package-lock.json		package-lock.json
package.json		package.json
postcss.config.mjs		postcss.config.mjs
prisma.config.ts		prisma.config.ts
setup.sh		setup.sh
tsconfig.json		tsconfig.json
vitest.config.ts		vitest.config.ts

Latest commit

History

Folders and files

Repository files navigation

TeamClaw

TeamClaw 是什么？

核心功能

快速开始

方式一：Docker Compose 部署（推荐）

启用 HTTPS（可选）

方式二：本地开发

首次使用指南

1. 登录管理面板

2. 配置模型 API 密钥

3. 部署 OpenClaw 实例

4. 开始对话

系统架构

技术栈

功能概览

界面截图

贡献

许可证

What is TeamClaw?

Core Features

Quick Start

Option 1: Docker Compose (Recommended)

Enable HTTPS (Optional)

Option 2: Local Development

Architecture

Tech Stack

Feature Overview

Getting Started Guide

1. Login

2. Configure Model API Key

3. Deploy an OpenClaw Instance

4. Start Chatting

Screenshots

Contributing

License

About

Resources

Contributing

Security policy

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages