What is A2A (Agent-to-Agent Protocol)?
TL;DR
An open protocol for safe communication and collaboration between AI agents from different vendors. Proposed by Google.
A2A (Agent-to-Agent Protocol): Definition & Explanation
A2A (Agent-to-Agent Protocol) is an open protocol proposed by Google in 2025 that enables AI agents from different vendors and organizations to communicate and collaborate in a secure, standardized way. While Anthropic's MCP (Model Context Protocol) standardizes connections between AI and tools/data sources, A2A standardizes horizontal connections between agents themselves. It provides capabilities including agent discovery (Agent Cards), task delegation and progress tracking, and message exchange. Over 50 partner organizations — including Salesforce, SAP, Atlassian, and LangChain — have announced their participation, and A2A is expected to play a pivotal role in building the enterprise AI ecosystem. MCP and A2A are complementary rather than competitive: MCP handles 'vertical' connections between AI and tools, while A2A handles 'horizontal' connections between agents, together forming the infrastructure for multi-agent systems.