Same AI Agent Built on Google, AWS, and Azure to Test Cloud Framework Differences

A developer built an identical research agent three times using Google ADK, AWS Strands, and Microsoft Agent Framework, all communicating via the open A2A (Agent2Agent) protocol to a single coordinator. The experiment aimed to isolate what actually differs between the three major cloud providers' agent frameworks once cross-platform communication is established. To ensure fair comparison, all variables except the framework itself were kept constant, including the instruction set, search tool, model budget, and scoring rubric. A shared search function was deliberately used across all three platforms instead of each vendor's native tool, since only Google ships a built-in search tool while AWS Strands and Microsoft's framework require external setup. The project concludes that A2A protocol interoperability works as intended, but meaningful differences remain at the framework level beyond what the wire protocol alone can address.