Concept 7 · Google · AI Community 2026A2A Protocol
04a, 04b & 05d · Art #4 & #5d

What if we want to host our Suggeritore agent on a different machine, since it needs higher computation resources? Let us scale across network boundaries using the Agent to Agent Protocol (or A2A):

A2A Server (to_a2a): turns Suggeritore into an independent HTTP microservice (port=8001) publishing /.well-known/agent-card.json, so that it is searchable.

A2A Client (RemoteA2aAgent): Cameriere connects to Suggeritore reading its data from the agent card and delegates tasks to it over JSON-RPC!

adk_agent_app_suggeritore/agent.py & cameriere_agent.py (05d)
# Server (Port 8001): Expose Suggeritore as an A2A microservice
a2a_app = to_a2a(agent=suggeritore_soste_agent, port=8001)
# Client (Port 8000): Consume remotely via Agent Card
remote_suggeritore = RemoteA2aAgent(
name="suggeritore_soste",
agent_card="http://localhost:8001/.well-known/agent-card.json",
)
CameriereA2A client · :8000
JSON-RPC
agent-card.json
SuggeritoreA2A server · :8001