A2A (Agent-to-Agent Protocol)
A2A (Agent-to-Agent Protocol) is a communication protocol that enables different AI agents to perform capability discovery, task delegation, and state synchronization, published by Google in April 2025.
If MCP is a protocol that "connects agents and tools," then A2A is a protocol that "connects agents with each other." A practical scenario makes this easier to understand. A sales support agent analyzes prospective customers, passes the results to a marketing agent to design an optimal campaign, and then a CRM agent updates the customer database — the A2A specification is what enables this kind of multi-agent coordination. A2A defines four main functional areas: **Capability Discovery**: A mechanism by which each agent publishes metadata about "what it can do," allowing other agents to search and reference that information. **Task Delegation**: The request/response format used when one agent assigns a task to another agent. **State Synchronization**: A streaming mechanism for sharing the progress and intermediate results of long-running tasks in real time. **Authentication and Authorization**: A security layer that establishes trust relationships between agents and prevents unauthorized access. In February 2026, NIST announced the AI Agent Standards Initiative, marking the beginning of full-scale standardization efforts based on A2A and MCP. As multi-agent systems move into practical use, the importance of A2A will only continue to grow.
Related Terms
Agent Skills
Agent Skills are reusable instruction sets defined to enable AI agents to perform specific tasks or areas of expertise, functioning as modular units that extend the capabilities of an agent.
Agentic AI
Agentic AI is a general term for AI systems that interpret goals and autonomously repeat the cycle of planning, executing, and verifying actions without requiring step-by-step human instruction.
Ambient AI
Ambient AI refers to an AI system that is seamlessly embedded in the user's environment, continuously monitoring sensor data and events to proactively take action without requiring explicit instructions.
