A2A-Demo: Cross-Departmental AI Agent Integration

July 22, 2025 · View on GitHub

A demonstration project showcasing the integration between AG-UI and A2A protocols, enabling seamless communication between AI agents across different departments and systems.

🎯 Overview

This demo features an HR Agent as the primary interface that can intelligently delegate tasks to specialized departmental agents:

💰 Finance Agent - Handles budgets, expenses, payroll, and financial queries
💻 IT Agent - Manages technical requests, system access, and IT support
🏢 Buildings Management Agent - Oversees facilities, maintenance, and office resources

The HR agent acts as a central coordinator, automatically routing complex requests to the appropriate specialists and orchestrating multi-departmental workflows.

🎬 Demo

🌐 Try the Live Demo

A2A Demo

See the A2A protocol in action as agents seamlessly communicate across departments

🔗 Protocol Integration

AG-UI Protocol

Provides the user interface and conversation management, enabling natural language interactions with the agent system.

A2A (Agent-to-Agent) Protocol

Enables secure, structured communication between different AI agents, allowing them to:

Share context and information
Delegate specialized tasks
Coordinate complex, cross-departmental workflows
Maintain conversation continuity across agent handoffs

🚀 Key Features

🤝 Cross-Departmental Coordination: HR agent seamlessly communicates with specialized agents
🔄 Intelligent Task Routing: Automatically determines which agent(s) can best handle specific requests
📋 Context Preservation: Maintains conversation context across agent interactions
⚡ Real-time Communication: Live streaming of agent responses and tool calls
🎛️ Multi-Agent Orchestration: Handles complex workflows requiring multiple departments

🛠️ Example Use Cases

Employee Onboarding

User: "Set up a new employee John Doe starting Monday"
HR Agent → IT Agent: "Create user account and email for John Doe"
HR Agent → Buildings: "Assign desk and access card for new hire"
HR Agent → Finance: "Add John Doe to payroll system"

Budget Request with IT Requirements

User: "I need approval for a \$5000 software license and server setup"
HR Agent → Finance: "Review budget availability for \$5000 expense"
HR Agent → IT: "Assess technical requirements for software deployment"

Facility Issue with Security Implications

User: "The keycard reader is broken in the secure server room"
HR Agent → Buildings: "Schedule repair for keycard reader"
HR Agent → IT: "Implement temporary security measures for server room"

🏗️ Architecture

┌─────────────────┐    ┌─────────────────┐
│   Web Frontend  │    │   AG-UI Client  │
│    (Next.js)    │◄──►│   Integration   │
└─────────────────┘    └─────────────────┘
         │                       │
         ▼                       ▼
┌─────────────────────────────────────────┐
│            HR Agent                     │
│        (Central Coordinator)            │
└─────────────────────────────────────────┘
         │           │           │
         ▼           ▼           ▼
    ┌─────────┐ ┌─────────┐ ┌─────────┐
    │Finance  │ │   IT    │ │Buildings│
    │ Agent   │ │ Agent   │ │  Agent  │
    └─────────┘ └─────────┘ └─────────┘

🚦 Getting Started

Prerequisites

Node.js 18+ and pnpm
Python 3.12+
OpenAI API Key

1. Start the Web Application

# Install dependencies
pnpm install

# Set your OpenAI API key
export OPENAI_API_KEY=your_openai_api_key_here

# Start the development server
pnpm dev

The application will be available at http://localhost:3000

2. Setup and Run the Agents

# Navigate to agents directory
cd agents

# Create virtual environment (one-time setup)
uv venv .venv --python 3.12
source .venv/bin/activate

# Install dependencies (one-time setup)
uv pip install -r pyproject.toml --all-extras

# Set your OpenAI API key
export OPENAI_API_KEY=your_openai_api_key_here

# Start all agents
./run_agents.sh

This will start all three departmental agents simultaneously:

Finance Agent on http://localhost:8001
IT Agent on http://localhost:8002
Buildings Management Agent on http://localhost:8003

Press Ctrl+C to stop all agents.

3. Try It Out

Open the web application at http://localhost:3000
Start a conversation with the HR agent
Try cross-departmental requests like:
- "Help me onboard a new employee"
- "I need to request a budget for new office equipment"
- "Set up IT access for our new intern"

🧪 Development

Adding New Agents

Create a new agent in the agents/ directory
Update run_agents.sh to include the new agent
Configure the agent URL in the frontend integration