An open-source AI-native operating system that enables intelligent agent team orchestration, cloud code execution, and automated task scheduling through Ghost/Shell/Model architecture and Pipeline/Route/Coordinate/Collaborate patterns.
Core Architecture#
Bot Assembly Formula: Ghost (Persona) + Shell (Executor) + Model = Bot
- Ghost (Persona): Defines AI role and behavior logic
- Shell (Executor): Defines execution environment and methods
- Model: Pluggable LLM brain (OpenAI, Anthropic, DeepSeek, GLM, etc.)
Five Core Modules#
- Chat (AI Dialogue): Multi-model compatible dialogue, group chat, attachment parsing, Follow-up/Correction modes, mem0 long-term memory, sandbox execution
- Code (Cloud Coding Engine): Parallel cloud coding tasks, requirement clarification, Git integration (GitHub/GitLab/Gitea/Gerrit), MCP/Skill configuration
- Follow (AI Task Trigger): Scheduled/event-driven execution, info stream generation, conditional filtering
- Knowledge (AI Document Library): Multi-format document management, NotebookLM mode Q&A, knowledge citation
- Customization (Fully Configurable): Web-based agent creation, agent wizard, group sharing
Four Collaboration Patterns#
| Pattern | Description | Example |
|---|---|---|
| Pipeline | Bots work sequentially, passing results | Developer → Reviewer → Tester |
| Route | Leader Bot dispatches tasks to best-fit Bot | Leader → {Frontend | Backend | Database} |
| Coordinate | Leader coordinates parallel work and merges | Leader → [Analyst, Data, Report] → combine |
| Collaborate | All Bots share context and freely discuss | [Bot A ↔ Bot B ↔ Bot C] |
Built-in Agent Teams#
- chat-team: General AI assistant + Mermaid diagrams
- translator: Multi-language translation
- dev-team: Complete Git workflow (branch → code → commit → PR)
- wiki-team: Repository Wiki documentation generation
Executor Types#
| Executor | Best Use Case |
|---|---|
| Chat | Quick dialogue, Q&A |
| Code | Programming tasks, code generation |
| Agno | Multi-agent collaboration |
| Dify | Workflow automation |
Quick Deployment#
# One-click install
curl -fsSL https://raw.githubusercontent.com/wecode-ai/Wegent/main/install.sh | bash
# Optional: Enable RAG feature
docker compose --profile rag up -d
Access: http://localhost:3000
Configuration Methods#
- Web UI: Visual configuration of Prompts, MCP, Skills, and multi-agent collaboration
- YAML (CRD style): Kubernetes-style definitions for Ghost/Bot/Team/Skill
- API: OpenAI-compatible interface for easy system integration
Tech Stack#
- Backend: Python, FastAPI
- Frontend: TypeScript, Next.js
- Runtime: Docker, E2B (Sandbox)
- Protocol Support: MCP (Model Context Protocol), E2B Sandbox
- License: Apache-2.0