Google launched an Agent2Agent protocol recently. https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/

I have been struggling to understand who actually benefits by conforming their Agents to this spec.

The way I understood it, MCP as a protocol took off because there were two established Clients with very large developer distribution (Claude Desktop and Cursor) that adopted the standard. This mean if you're a developer, MCP now democratizes your ability to make the tools you already use (Claude & Cursor), more powerful!

A2A doesn't seem to have an analagous benefit. Current clients with established distribution for A2A is 0, as I understand it. Therefore there are no developers today that would benefit from building an A2A server. Perhaps A2A is designed to motivate the big clients with massive existing distribution (ChatGPT, Claude, Cursor) to support A2A so they can get access to other people's Agents. But all those clients are also competitors of A2A's maker (Google). So even though a Client might value being able to provide a lot of other people's agents easily in their interface, it seems like adopting Google's protocol might be a risky choice.