A2A Doesn't need AI Agents

Is this just distributed computing re-packaged for AI?

What is A2A Without AI Agents?

I got into a debate with Gemini recently about Agent-To-Agent protocol (A2A). I said I thought it was a retread of existing distributed computing technologies like Service Discovery, Mesh, CORBA, etc. Perhaps Gemini took it personally, as Google (Gemini’s Creator) had announced A2A in April, and Gemini got a little “gushy” on how it was “a revolutionary new idea.” Also, perhaps “debate” is too strong a word. And I might want to consider getting out more often.

An experiment was needed. I wanted to explore A2A features and get a better idea of what A2A is bringing over and above APIs, MCP (Model Context Protocol), and other distributed computing paradigms of the past and present. The resulting project [1] can be found on GitHub, and the DeepWiki version [2] - where the below diagram comes from.

When I got done with this experiment (helped largely by Claude Code), it all worked very well, but I realized there was no “AI” in the solution! it was just routing messages and responses between client and remote JSON-RPC services. Sprinkle on a little bit of “discovery” in the form of “Agent Cards” (./well-known/agent-card.json), and this looks like YADS (Yet Another Distributed Service) architecture - minus the AI part.

It is early for A2A, so much of the value is yet to come. The other parts, the layering on of service negotiation, reputation, security, payments, and other features needed to realize a fully self-organizing Agentic AI collaboration infrastructure are still on the drawing board.

For now, MCP, and Agent frameworks have a head start.

References

[1] O’Shea, M. (2025). A2A Example Project. GitHub repository. https://github.com/oshea00/a2a-proj

[2] DeepWiki. (2025). A2A Project Wiki. https://deepwiki.com/oshea00/a2a-proj