Agent2Agent (A2A) Example

This directory demonstrates a simple implementation of Google�s Agent2Agent (A2A) protocol with three specialized agents and an orchestrator.

Directory Structure

agent2agent/
├── README.md           # this file
└── a2a_example.py      # example script showing A2A workflow

Overview

The A2A protocol enables AI agents to discover each other�s capabilities, negotiate tasks, and collaborate through structured JSON messages.

Components

• AgentCard: Advertises an agent�s name and capabilities.
• Task: Encapsulates task metadata, status, and resulting artifact.
• Message: Carries sender, receiver, content, and optional parts for UI negotiation.
• Agent (base class): Provides advertise and default handle_message (echo).
• SearchAgent: Implements search_candidates to return dummy profiles.
• FilterAgent: Implements filter_candidates to filter by skill.
• ScheduleAgent: Implements schedule_interview to assign interview times.
• Orchestrator: Acts as the client agent, discovers capabilities, sends messages to complete a candidate-sourcing workflow.

Prerequisites

• Python 3.7 or higher

No external packages are required (uses Python standard library).

Running the Example

1. Clone or navigate to this directory:

   cd agent2agent

2. Run the example script:

   python a2a_example.py

3. Observe:
   • Agents advertise their capabilities.
   • Orchestrator issues tasks: search, filter, schedule.
   • Agents respond with Task artifacts.
   • Final output: list of found, filtered candidates and interview schedule.

Extending this Example

• Add new agents: Create classes inheriting from Agent and implement handle_message for new actions.
• Real data: Replace dummy logic with database/API calls or LLM-based processing.
• Network transport: Integrate HTTP/WebSocket layers for remote agent communication.
• UI integration: Use Message.parts to negotiate UI components (forms, iframes).

References

• Google Developers Blog: Announcing the Agent2Agent Protocol (A2A)
• A2A Spec and Samples: https://github.com/google/A2A