Skip to content
BENIBIX Learn

Agent2Agent Protocol (A2A)

AI🤖 AI AgentsAgent Orchestration

Also known as A2A, A2A Protocol, Agent-to-Agent Protocol, Agent2Agent

Rules that let agents from two firms talk and work as one team.

Think of it like

Think of it like a shared language. Robots from different shops can use it to work as a team.

Example

Jack's travel agent needs a hotel. It reads a booking agent's card, sends it the dates and city, and gets back three options to show him.

Why it matters

With no shared standard, each pair of agents needs its own code just to talk.

Definition

The Agent2Agent protocol, or A2A, is an open standard that lets AI agents work together. It works even when the agents were built by different makers. Each agent posts an agent card, a small JSON file with its name, what it can do and where to reach it. Another agent reads the card, sends it a task and gets back updates and a result. A2A links agent to agent, while MCP links an agent to its tools. It came out in 2025 and is now run by the Linux Foundation.

Example code

json
{
  "name": "Hotel Booking Agent",
  "description": "Finds and books hotel rooms.",
  "url": "https://hotels.example.com/a2a",
  "skills": [
    { "id": "find-room", "name": "Find a room" }
  ]
}
An agent card says who the agent is, what it can do and where to send it work.
One agent reads the other's card, then sends it a task and gets back a result.

Common mistake

Mixing it up with MCP. A2A is for agents talking to other agents, and MCP is for giving one agent tools and data.

Key takeaways

  1. 1
    Agent to agent. Across makers.
  2. 2
    Agent cards. Say what each can do.
  3. 3
    Tasks and results. Sent back and forth.

Compare