A2A Protocol = Agents sending prompts to other agents

A2A Protocol Is Just HTTP With Extra Structure

Everyone’s talking about Google’s A2A protocol like it’s revolutionary. It’s not. It’s just HTTP with extra structure. Let me break it down, it’s not that complicated.

Agent Discovery

Agents publish what they can do at a known URL. Think of it like a business card for bots. The problem: without a marketplace or registry, no one’s going to find your agent by randomly guessing URLs.

Standardized Format

A JSON-RPC wrapper around messages and artifacts. It’s a shared language so agents from different vendors can talk to each other. Useful? Sure. Required? No. We’ve been passing structured JSON over HTTP for decades.

Streaming

SSE and webhooks for real-time updates. Nothing new here. These patterns existed long before A2A.

Authentication

OAuth, API keys, mutual TLS. Already solved. A2A just says “use these” - which you were already doing.

Task Lifecycle

States like in-progress, completed, waiting-for-input. Helpful for long-running async work. But also just a state machine - something web apps have done forever.

That’s it. That’s the whole protocol. Don’t overcomplicate it. And don’t let the hype make it sound harder than it is.