Learn how agents discover, communicate, and delegate tasks to each other using the A2A protocol — the open standard (now Linux Foundation) for agent-to-agent interoperability.

The AI ecosystem is converging on two complementary standards: MCP (Model Context Protocol) for connecting agents to tools and data, and A2A (Agent-to-Agent) for connecting agents to each other. Together they form the complete agent interoperability stack.

A2A solves a critical gap: how do independently built agents — potentially from different vendors, frameworks, and platforms — discover each other's capabilities and collaborate on tasks? Without A2A, every multi-agent system invents its own bespoke integration layer.

This track focuses on:

• understanding the A2A protocol specification and its design principles
• building agents that publish discoverable Agent Cards
• implementing task lifecycle management with streaming updates
• securing agent-to-agent communication with OAuth2 and identity verification
• combining A2A with MCP to create the full agent ecosystem architecture

• repository: a2aproject/A2A
• stars: about 24.9k
• GitHub release reference: v1.0.1 (checked 2026-07-20; release metadata on GitHub)

flowchart LR
    subgraph MCP ["MCP (Agent → Tool)"]
        A1[AI Agent] -->|calls| T1[Tool Server]
        A1 -->|reads| R1[Resource / Data]
    end

    subgraph A2A ["A2A (Agent → Agent)"]
        A2[Client Agent] -->|discovers| AC[Agent Card]
        A2 -->|sends task| A3[Remote Agent]
        A3 -->|streams updates| A2
        A3 -->|returns artifact| A2
    end

    User[User / Host App] --> A1
    User --> A2
    A3 -->|uses tools via MCP| T2[MCP Server]

    classDef mcp fill:#e1f5fe,stroke:#01579b
    classDef a2a fill:#fff3e0,stroke:#ef6c00
    classDef user fill:#e8f5e8,stroke:#1b5e20

    class A1,T1,R1 mcp
    class A2,AC,A3 a2a
    class User user
    class T2 mcp

• How to read and implement the A2A protocol specification
• How to create Agent Cards that advertise capabilities and skills
• How to manage the full task lifecycle with streaming and push notifications
• How to authenticate and authorize agent-to-agent communication
• How to build A2A servers and clients using the Python SDK
• How to design multi-agent delegation and composition patterns
• How to combine MCP (tools) and A2A (agents) into a unified architecture

• A2A Protocol README
• A2A Specification
• Agent Card Schema
• A2A Python SDK
• A2A Samples
• A2A Technical Documentation

• MCP Specification Tutorial — The complementary protocol for agent-to-tool communication
• Composio Tutorial — Tool integration platform that bridges agent frameworks
• CrewAI Tutorial — Multi-agent orchestration framework
• Taskade Tutorial — AI-native productivity with agent capabilities

Start with Chapter 1: Getting Started.

• Start Here: Chapter 1: Getting Started
• Back to Main Catalog
• Browse A-Z Tutorial Directory
• Search by Intent
• Explore Category Hubs

1. Chapter 1: Getting Started
2. Chapter 2: Protocol Specification
3. Chapter 3: Agent Discovery
4. Chapter 4: Task Management
5. Chapter 5: Authentication and Security
6. Chapter 6: Python SDK
7. Chapter 7: Multi-Agent Scenarios
8. Chapter 8: MCP + A2A

Generated by AI Codebase Knowledge Builder