Content
# Tool List
**LAN-level Intelligent Asset Management Platform** — Skill Warehouse + Security Inspection Pipeline + Onboarding Manual, all-in-one.
## Product Positioning
Agent Hub does not handle task collaboration scheduling between agents, but focuses on three main tasks:
1. **Unified Skill Management** — Install once, and all agents automatically discover and use
2. **Unified Security Detection** — Scan skills upon import, with traceable risks
3. **Unified Knowledge Distribution** — Onboarding manual and development norms entered once, shared by all agents
## Quick Start
### Start Service
```bash
# Clone repository
git clone <repo-url>
cd agent-hub
# Start (Docker Compose)
docker-compose up -d
# Access Web Panel
# http://localhost:8201
# Default admin password: admin123
```
### One-Click Agent Access
```bash
curl -s http://localhost:8200/api/v1/onboard/claim?token=<TOKEN> | bash
```
Or in the Web Panel:
1. Go to "Agent Management" → Create Agent
2. Copy one-click access command
3. Execute on the agent machine
## Core Features
### Web Panel (port 8201)
| Page | Function |
|------|------|
| 📊 Dashboard | Agent/skill/compliance/security overview |
| 🤖 Agent Management | Create/delete/view agents, credential management |
| 📦 Skill Library | Upload/assign/public-private switching |
| 🌐 Public Market | ClawHub/SkillHub search import |
| 🛡️ Security Inspection | Scan records/Tier 1 & Tier 2 inspection |
| ✅ Compliance Check | Mandatory skill list + missing alerts |
| 📋 Audit Log | Full operation audit tracing |
| 📚 Knowledge Base | Onboarding manual/development norm management |
### MCP Protocol (main access method)
| Tool | Description |
|------|------|
| `skills.list` | Get skill list |
| `skills.get` | View skill details |
| `skills.download` | Download skill package |
| `agent.heartbeat` | Agent heartbeat report |
| `knowledge.search` | Search knowledge base |
| `knowledge.get` | Get knowledge entry full text |
| `hub.onboarding` | Get platform usage norms |
### REST API (fallback)
| Endpoint | Description |
|------|------|
| `POST /api/v1/agent/heartbeat` | Heartbeat report |
| `GET /api/v1/agent/skills` | Skill list |
| `GET /api/v1/agent/skills/{name}/download` | Download skill package |
| `GET /api/v1/agent/knowledge/search` | Search knowledge base |
## Security Detection System
| Level | Method | Description |
|------|------|------|
| Tier 1 | Static analysis | Regular expression matching of dangerous patterns (shell/network/file/privilege escalation, etc.) |
| Tier 2 | Sandbox execution | Docker isolated execution, dynamic behavior detection |
| Tier 3 | Threat modeling | Reserved |
## Tech Stack
- **Backend**: Python FastAPI + SQLAlchemy + SQLite + MCP (FastMCP)
- **Frontend**: React + TypeScript + Ant Design
- **Deployment**: Docker Compose
- **SDK**: Python / Node.js
## Project Structure
```
agent-hub/
├── backend/ Backend service (FastAPI + MCP)
│ ├── app/
│ │ ├── api/ REST API routing
│ │ ├── services/ Business logic layer
│ │ ├── models/ SQLAlchemy models
│ │ ├── schemas/ Pydantic data models
│ │ ├── mcp/ MCP Server
│ │ └── utils/ Utility functions
│ └── tests/ Test suite (85 tests)
├── frontend/ Web management panel (React + Ant Design)
├── agent-sdk/ Agent access SDK
│ ├── python/ Python SDK
│ └── nodejs/ Node.js SDK
├── docker-compose.yml Deployment configuration
└── Development Document/ PRD and technical plan
```
## Development Norms
- Follow RESTful API design specifications
- Code cleanliness, modularity, and maintainability prioritized
- Test before implementation, with unit tests for all new features
- Comply with "Alibaba Java Development Manual" ideas (Python version)
## License
MIT
