A2A Protocol — Agent-to-Agent Communication Standard

What it is

The A2A (Agent-to-Agent) Protocol is Google's open standard for AI agents to discover each other's capabilities, communicate, and collaborate regardless of the framework they were built with. Each agent publishes an Agent Card describing its skills, and other agents use the protocol to delegate tasks, exchange data, and coordinate multi-agent workflows. SDKs are available for Python, TypeScript, Java, Go, and .NET.

Developers building multi-agent systems where agents from different frameworks need to interoperate benefit from A2A. It solves the problem of vendor lock-in in agent orchestration by providing a shared communication layer.

How it saves time or tokens

Without A2A, connecting agents built on different frameworks requires custom integration code for each pair. A2A standardizes this with a single protocol, reducing the integration work from O(n^2) to O(n). Agents can also discover each other's capabilities dynamically, avoiding the overhead of hardcoded routing logic.

How to use

1. Install the A2A SDK for your language
2. Create an Agent Card describing your agent's capabilities
3. Implement the A2A server interface and connect to other A2A agents

Example

# Python SDK
from a2a import A2AServer, AgentCard, Task

card = AgentCard(
    name='code-reviewer',
    description='Reviews Python code for bugs and style issues',
    skills=['code-review', 'python']
)

class MyAgent(A2AServer):
    def handle_task(self, task: Task):
        code = task.input['code']
        # Review the code...
        return {'review': 'LGTM, no issues found'}

agent = MyAgent(card=card)
agent.start(port=8080)

# TypeScript/JavaScript
npm install @a2a-js/sdk

# Go
go get github.com/a2aproject/a2a-go

Related on TokRepo

• Multi-agent frameworks — Compare multi-agent orchestration tools
• AI tools for agents — Browse agent frameworks and protocols

Common pitfalls

• A2A is a communication protocol, not an orchestration framework; you still need logic to decide which agent handles which task
• Agent discovery requires a registry or known endpoints; there is no global agent directory built into the protocol
• Protocol adoption is early; expect breaking changes in the SDK as the specification evolves