Two agents built by different teams must delegate work to each other. What are they designed to use?
Why D is the answer
An agent-to-agent protocol
Agent-to-agent protocols define discovery and task exchange.
In detail
- Agent-to-agent protocols define discovery and task exchange.
- MCP connects tools, while this connects agents.
Use case
An illustrative scenario, not a real incident.
Two teams' agents delegate work through an agent-to-agent protocol, with no custom integration.
See it in different ways
flowchart TD A["Supervisor"] --> B["Researcher"] A --> C["Analyst"] A --> D["Writer"] B --> A C --> A D --> A
mindmap
root((Standards))
Tools and data
Model Context Protocol
Other agents
Agent-to-agent protocols
Long tasks
Asynchronous steps
Durable stateflowchart TD
Q{"What is being connected?"}
Q -->|"Agent to tool"| A0["MCP"]
Q -->|"Agent to agent"| A1["Agent-to-agent protocol"]Key point: Agent-to-agent protocols cover delegation.
Why the other options are wrong
- A. A shared GPU memory region
- Shared GPU memory is not an interface.
- B. A single shared system prompt
- One prompt does not connect two systems.
- C. A manual copy of each other's logs
- Log copies do not delegate work.