tgo

TGO Introduction

TGO is an open-source AI agent customer service platform dedicated to helping enterprises "Build AI Agent Teams for Customer Service". It integrates multi-channel access, agent orchestration, knowledge base management (RAG), and human agent collaboration.

🚀 Quick Start

One-Click Deployment

Run the following command on your server to check requirements, clone the repository, and start the services:

REF=latest curl -fsSL https://raw.githubusercontent.com/tgoai/tgo/main/bootstrap.sh | bash

For users in China (using Gitee and Aliyun mirrors):

REF=latest curl -fsSL https://gitee.com/tgoai/tgo/raw/main/bootstrap_cn.sh | bash

Local Development

Start the full development environment with Docker Compose:

cp .env.dev.example .env.dev
make dev

Useful variants:

make dev PROFILES=monitoring
make dev DISABLE=tgo-rag-beat,tgo-workflow-worker

For more details, please visit the Documentation.

✨ Features

🤖 AI Agent Orchestration

• Multi-Agent Support - Configure multiple AI agents for different business scenarios
• Multi-Model Integration - Connect with various LLM providers (OpenAI, Anthropic, etc.)
• Streaming Response - Real-time AI responses via SSE for smooth conversation experience
• Context Memory - Maintain conversation history for coherent dialogue

📚 Knowledge Base (RAG)

• Document Knowledge Base - Upload documents to enhance AI response accuracy
• Q&A Knowledge Base - Create question-answer pairs for quick knowledge expansion
• Website Knowledge Base - Crawl website content to keep information up-to-date
• Smart Retrieval - Vector-based semantic search for precise answers

🔧 MCP Tools Integration

• Tool Store - Rich library of MCP tools, enable on demand
• Custom Tools - Project-level tool configuration and management
• OpenAPI Schema - Auto-parse schemas to generate interactive forms

🌐 Multi-Channel Access

• Web Widget - Embeddable chat widget for websites
• WeChat Integration - Official Account and Mini Program support
• Unified Management - Manage all channels from a single dashboard

💬 Real-time Communication

• WuKongIM Integration - Stable and reliable instant messaging
• WebSocket Connection - Efficient bidirectional communication
• Message Sync - Read/unread status, delivery confirmation
• Rich Media - Support for text, images, files and more

👥 Human-AI Collaboration

• Smart Handoff - Seamlessly transfer to human agents when needed
• Visitor Management - Collect visitor info, assign sessions, track history
• Agent Workspace - Unified interface for human agents

🎨 UI Widget System

• Structured Display - Render orders, products, logistics as beautiful cards
• Rich Components - Order cards, logistics tracking, product display, price comparison
• Action Protocol - Standardized URI protocol for interactions

📦 Repository Structure

| Repository | Description | Tech Stack | |:---|:---|:---| | tgo-ai | AI/ML operations service for managing agents, tool bindings, knowledge bases, and usage analytics | Python / FastAPI | | tgo-api | Core business logic service handling user management, visitor tracking, assignment, and communication | Python / FastAPI | | tgo-cli | CLI tool & MCP Server enabling AI agents to execute customer service operations with 40+ built-in tools | TypeScript / Node.js | | tgo-device-agent | Embedded agent running on managed devices providing file and shell capabilities via TCP JSON-RPC | Go | | tgo-device-control | Device control service managing TCP/JSON-RPC connections for remote device management with MCP Agent | Python / FastAPI | | tgo-platform | Multi-channel message intake service supporting WeChat, Feishu, DingTalk, Telegram, Slack, email, etc. | Python / FastAPI | | tgo-plugin-runtime | Plugin lifecycle management and execution service with dynamic tool synchronization | Python / FastAPI | | tgo-rag | RAG service providing document processing, hybrid semantic/full-text search, and async processing | Python / FastAPI | | tgo-web | Admin frontend with real-time chat, AI agent management, knowledge base, and MCP tool integration | TypeScript / React 19 | | tgo-workflow | AI Agent workflow execution engine supporting DAG topology with LLM, API, condition, and tool nodes | Python / FastAPI |

Widget SDKs

| Repository | Description | Tech Stack | |:---|:---|:---| | tgo-widget-js | Embeddable customer service chat widget (Intercom-style) for websites | TypeScript / React 18 | | tgo-widget-ios | Native iOS customer service chat SDK with SwiftUI views and UIKit bridging | Swift / SwiftUI | | tgo-widget-flutter | Cross-platform customer service chat widget for iOS and Android | Dart / Flutter | | tgo-widget-cli | Visitor-facing CLI tool & MCP Server providing customer service interface | TypeScript / Node.js | | tgo-widget-miniprogram | WeChat Mini Program chat widget component with AI streaming responses and Markdown rendering | TypeScript |

🏗️ System Architecture

Product Preview

| | | |:---:|:---:| | Dashboard | Agent Orchestration | | Knowledge Base | Q&A Debugging | | MCP Tools | Platform Admin |

System Requirements

• CPU: >= 4 Core
• RAM: >= 8 GiB
• OS: macOS / Linux / WSL2