Agent2Agent (A2A) Protocol

Definition

What it is

An open, Linux-Foundation-hosted protocol (originally announced by Google in April 2025, donated to the Linux Foundation in 2025) for **peer-to-peer agent collaboration** — allowing autonomous AI agents built on different frameworks (LangGraph, crewAI, LlamaIndex, AutoGen, custom) and operating in heterogeneous environments to dynamically discover each other's capabilities, exchange contextual state, and delegate sub-tasks. A2A operates on HTTP + Server-Sent Events and uses capability-based "Agent Cards" as the discovery primitive.

Why it exists

MCP standardized how an agent talks to its *tools* (databases, file systems, APIs); it was not designed for how one agent talks to *another agent*. In a vendor-fragmented multi-agent ecosystem — Salesforce's inventory agent collaborating with SAP's procurement agent collaborating with an in-house pricing agent — every pair-wise integration becomes O(N²) glue code. A2A is the universal-translator layer that makes that integration O(N): every agent advertises its capabilities once via an Agent Card, every other agent reads the Card and can negotiate.

Recent developments

Latest signals

A2A donated to the Linux Foundation, governance maturing. Google donated the spec in 2025; a Linux Foundation working group now manages it alongside the MCP working group, with explicit interop alignment. Per Google Developers Blog — Announcing A2A.
A2A + MCP positioned as complementary, not competing. Industry consensus (IBM, AWS, Anthropic, Google) is that the canonical stack uses MCP for agent→tool I/O and A2A for agent→agent communication. Per IBM — What is A2A protocol and AWS — Open Protocols for Agent Interoperability.
Academic survey formalizes the four-protocol taxonomy. Ehtesham et al. arXiv:2505.02279 surveys MCP / A2A / ACP / ANP as the four interoperability tiers — MCP for tools, A2A for cross-org peer agents, ACP for high-throughput local multi-agent, ANP for trust-decentralized federation. Per arXiv 2505.02279.
Production case study: inventory→procurement→supplier agent chains. Enterprise pilots demonstrate A2A-mediated chains where a warehouse inventory agent detects a supply shortage, delegates to an internal procurement agent, which negotiates with external supplier agents — all over standardized A2A messages with no per-vendor integration. Per Google Developers Blog.