Northwind Traders is integrating an existing inventory specialist agent built by another team with a new customer-service orchestrator. The inventory agent already exposes skills over a standard agent-to-agent interface. The orchestrator also needs to call a warehouse REST API for stock counts. Which protocol assignment is most appropriate?
Select an answer to reveal the explanation.
Short Explanation
Two different integration problems need two different contracts. Agent-to-agent is A2A (or equivalent agent discovery/invoke), so the orchestrator should integrate the inventory specialist via Agent2Agent. Calling a warehouse REST API is tool territory — MCP servers/clients or classic function tools, not peer-agent chat. That is option C. Blob drop files (A) are not a production multi-agent protocol. Swapping MCP and A2A roles (B) gets the model backward. Stuffing API secrets into prompts (D) is unsafe and not a real tool interface.
Full Explanation
AI-500 covers specifying appropriate protocols for multi-agent communication and integrating existing agents using A2A and/or MCP. Option C assigns A2A to agent-to-agent integration of an existing specialist and MCP/tool interfaces to structured external system access — the correct separation of concerns.
Option A is incorrect: asynchronous blob handoffs lack typed discovery, capability negotiation, and reliable request/response semantics needed for production agent coordination.
Option B reverses the roles. MCP is designed for tool and context resource access; A2A targets agent interoperability. Using MCP as the primary agent-to-agent fabric and A2A for raw REST is inverted.
Option D is incorrect and insecure. Credentials must not live in prompts, and models should not freestyle HTTP without mediated tool boundaries, auth, and validation.
Exam tip: A2A connects agents; MCP (and function tools) connect agents to tools/data sources.