What is the A2A protocol? Agent2Agent explained
Also called Agent2Agent protocol, A2A, agent-to-agent protocol
Definition
The A2A (Agent2Agent) protocol is an open standard, introduced by Google in April 2025, that lets AI agents built on different frameworks discover each other, exchange messages, and delegate tasks.
The A2A (Agent2Agent) protocol is an open standard for communication between AI agents. Google announced A2A on April 9, 2025, with support from more than 50 technology partners, and later donated it to the Linux Foundation; the project has since joined the Agentic AI Foundation. A2A lets an agent built with one framework or vendor hand work to an agent built with another. A2A treats the remote agent as a black box: the client sees its declared skills, not its internal logic, memory, or tools.
How the A2A protocol works#
The A2A protocol works by having a client agent send tasks to a remote agent that advertises its skills. The core objects in the A2A specification are:
- Agent Card: a JSON document, published at
/.well-known/agent-card.jsonon the agent's domain, that lists the agent's identity, endpoint, skills, supported features, and authentication requirements. - Task: the unit of work, with a server-generated ID and a lifecycle of states including submitted, working, input-required, completed, failed, and canceled.
- Message and Part: a message is one turn of communication, made of parts that carry text, file references, or structured data.
- Artifact: an output the remote agent produces for a task.
Version 1.0 of the specification defines protocol bindings for JSON-RPC 2.0, gRPC, and HTTP+JSON. Streaming updates use Server-Sent Events, and long-running tasks can report progress through push notifications to a client-registered webhook.
A2A vs MCP#
A2A and MCP are complementary: MCP connects an agent to tools and data, and A2A connects an agent to other agents. The A2A project describes the two the same way and states that A2A is not a replacement for MCP.
| MCP | A2A | |
|---|---|---|
| Connects | Agent to tools and data sources | Agent to agent |
| Remote side | A server exposing tools, resources, and prompts | An opaque agent exposing skills |
| Unit of work | A tool call | A task with a lifecycle |
| Discovery | tools/list and other list methods |
Agent Card |
| Origin | Anthropic, November 2024 | Google, April 2025 |
A multi-agent system can use both: each agent reaches its own tools through MCP and delegates to specialist agents through A2A. A2A vs MCP compares the protocols in depth, and the MCP topic hub covers the wider protocol landscape.
Related terms
Guides that cover a2a protocol
- A2A protocol vs MCP: differences and when to use each
- What is the Model Context Protocol (MCP)? How it works
- Bifrost vs LiteLLM: architecture, features, and performance compared
- AI agent frameworks compared: LangGraph, CrewAI, ADK, and more
- What is an AI gateway? How it works and when you need one
- What are AI agents? How agents work and when to use one
Cite this definition
The AI Engineering Editors. "What is the A2A protocol? Agent2Agent explained." The AI Engineering, updated Nov 06, 2026. https://theaiengineering.com/glossary/a2a-protocol
- Title
- What is the A2A protocol? Agent2Agent explained
- Author
- The AI Engineering Editors
- Publisher
- The AI Engineering
- Updated