A2A (Agent-to-Agent) is a communication protocol introduced by Google that enables autonomous agents to interact with each other through standardized message passing, goal negotiation, and task coordination. Rather than treating agents as isolated LLM wrappers or task-specific bots, A2A enables them to:
- Exchange structured messages (intents, plans, results, etc.)
- Negotiate roles and responsibilities
- Form dynamic workflows through coordination and planning
- Collaborate asynchronously on complex objectives
The protocol is inspired by the need for multi-agent collaboration at scale—an essential step for complex use cases like automated research, enterprise automation, and multi-modal AI systems.
This repository gathers concrete implementations and usage examples of the A2A protocol across popular agentic AI frameworks. The goal is to lower the barrier to entry for developers and researchers looking to adopt A2A in their own systems.
- Provide plug-and-play examples of A2A in action
- Demonstrate how to adapt A2A for various frameworks
- Highlight best practices for multi-agent messaging, intent resolution, and coordination
- 🪄 Agno Agents
- 📚 LlamaIndex
- 🧠 LangGraph
- 🛠️ AutoGen
- 📊 CrewAI
- 💡 Pydantic AI
- 📖 OpenAI Agent SDK
Each example illustrates the core A2A flow:
- Agent Registration: Define agents with roles, goals, capabilities.
- Intent Messaging: Agents send/receive intent messages like
"propose_plan","evaluate_output","request_input". - Coordination: Agents negotiate or agree on task boundaries (e.g.,
"you validate my output"). - Execution Loop: Messages are processed and results are passed to the next agent.
- Finalization: Agents may jointly agree on when a task is considered “done”.
Simple go to each folder and follow the instructions and explore the examples