Build Multi-Agent Systems using A2A SDK

⚠️ DISCLAIMER: THIS DEMO IS INTENDED FOR DEMONSTRATION PURPOSES ONLY. IT IS NOT INTENDED FOR USE IN A PRODUCTION ENVIRONMENT.

⚠️ Important: A2A is a work in progress (WIP) thus, in the near future there might be changes that are different from what demonstrated here.

This document describes a web application demonstrating the integration of Agent2Agent (A2A), Google Agent Development Kit (ADK) for multi-agent orchestration with Model Context Protocol (MCP) clients. The application features a host agent coordinating tasks between remote agents that interact with various MCP servers to fulfill user requests.

Architecture

The application utilizes a multi-agent architecture where a host agent delegates tasks to remote agents (Airbnb and Weather) based on the user's query. These agents then interact with corresponding MCP servers.

App UI

Setup and Deployment

Prerequisites

Before running the application locally, ensure you have the following installed:

1. Node.js: Required to run the Airbnb MCP server (if testing its functionality locally).
2. uv: The Python package management tool used in this project. Follow the installation guide: https://docs.astral.sh/uv/getting-started/installation/
3. Python 3.13 Python 3.13 is required to run a2a-sdk
4. set up .env

• Create a .env file in airbnb_agent and weather_agent folder with the following content:

  GOOGLE_API_KEY="your_api_key_here" 
  

• Create .env file in host_agent/ folder with the following content:

  GOOGLE_GENAI_USE_VERTEXAI=TRUE
  GOOGLE_CLOUD_PROJECT="your project"
  GOOGLE_CLOUD_LOCATION=global
  AIR_AGENT_URL=http://localhost:10002
  WEA_AGENT_URL=http://localhost:10001
  

1. Run Airbnb Agent

Run the airbnb agent server:

cd samples/python/agents/airbnb_planner_multiagent/airbnb_agent
uv run .

2. Run Weather Agent

Open a new terminal and run the weather agent server:

cd samples/python/agents/airbnb_planner_multiagent/weather_agent
uv run .

3. Run Host Agent

Open a new terminal and run the host agent server

cd samples/python/agents/airbnb_planner_multiagent/host_agent
uv run .

4. Test at the UI

Here are example questions:

• "Tell me about weather in LA, CA"

• "Please find a room in LA, CA, June 20-25, 2025, two adults"

References

• https://github.com/google/a2a-python
• https://codelabs.developers.google.com/intro-a2a-purchasing-concierge#1
• https://google.github.io/adk-docs/

Disclaimer

Important: The sample code provided is for demonstration purposes and illustrates the mechanics of the Agent-to-Agent (A2A) protocol. When building production applications, it is critical to treat any agent operating outside of your direct control as a potentially untrusted entity.

All data received from an external agent—including but not limited to its AgentCard, messages, artifacts, and task statuses—should be handled as untrusted input. For example, a malicious agent could provide an AgentCard containing crafted data in its fields (e.g., description, name, skills.description). If this data is used without sanitization to construct prompts for a Large Language Model (LLM), it could expose your application to prompt injection attacks. Failure to properly validate and sanitize this data before use can introduce security vulnerabilities into your application.

Developers are responsible for implementing appropriate security measures, such as input validation and secure handling of credentials to protect their systems and users.