Agent2Agent Protocol (A2A)
ay-too-ay
What is A2A?
Agent2Agent (A2A) is an open communication protocol that enables AI agents built on different frameworks and by different companies to communicate and collaborate with each other. Originally developed by Google in April 2025, A2A is now governed by the Linux Foundation under the Apache 2.0 license.
While MCP connects agents to tools and data sources, A2A enables agents to talk to each other—securely exchanging information and coordinating actions across organizational boundaries.
Why A2A Matters for Organizations
As enterprises deploy multiple AI agents across different departments and vendors, a critical problem emerges: how do agents built with different frameworks collaborate without manual integration?
A2A solves this by providing:
- Agent-to-agent communication - Agents can delegate tasks to specialized agents elsewhere
- Cross-platform interoperability - Works regardless of underlying framework (LangChain, ADK, etc.)
- Enterprise integration - Built on HTTP, SSE, and JSON-RPC standards that fit existing IT infrastructure
- Security model - Signed security cards and authentication for trust between agents
How A2A Relates to MCP
A2A and MCP are complementary, not competing:
| Protocol | Focus |
|---|---|
| MCP | Agent-to-tool communication |
| A2A | Agent-to-agent communication |
MCP provides agents with tools, APIs, and data access. A2A enables agents to delegate work to other specialized agents. In a complex agentic system, you typically need both.
Protocol Ecosystem
A2A fits into Google's broader agent protocol strategy:
- MCP - Tools and data access (Anthropic)
- A2A - Agent collaboration (Google → Linux Foundation)
- AP2 - Payment processing
- AGUI - Dynamic UI generation
- UCP - End-to-end commerce
Industry Adoption
A2A launched with support from over 50 technology partners including Atlassian, Box, Salesforce, ServiceNow, SAP, PayPal, MongoDB, and Workday. As of late 2025, the ecosystem has grown to over 150 organizations.
Version 0.3 introduced gRPC support, signed security cards, and extended Python SDK support.
See Also
- Model Context Protocol (MCP) - The complementary agent-to-tool protocol
- Universal Commerce Protocol (UCP) - Commerce-specific protocol built on A2A
- AI Agents - Autonomous systems using these protocols
Mentioned In
Zenchant
"Referenced as Google's agent-to-agent messaging protocol alongside MCP, AP2, AGUI, and UCP"