Meridian's pet-insurance partner has built its own external claims-adjustment AI system, and wants that system to be able to ask Meridian's deployed agent 'what is the status of claim X' on demand, without any Meridian staff mediating the exchange. Which mechanism fits this integration?
Select an answer to reveal the explanation.
Short Explanation
Follow the arrow: the insurer's system is reaching IN to ask Meridian's agent a question, not the other way around. Whenever an outside application wants to call your agent, that's an Agent API job, full stop.
Full Explanation
Direction is the whole trick in this scenario. The insurer's external claims-adjustment system wants to initiate the query and call into Meridian's already-deployed agent - that is an external application consuming an Agentforce agent, which is exactly what the Agent API is built for.
MCP is the closest-looking trap because it also connects Salesforce to something external, but MCP describes the opposite direction: Meridian's own agent reaching out to draw on an external model or tool as context, not an outside system calling in to query Meridian's agent. The agent-to-agent protocol only applies when the other side is itself an Agentforce agent reasoning over its own topics and actions; nothing here indicates the insurer's claims-adjustment system is an Agentforce agent, and even if it were, A2A is for task delegation and collaboration, not a simple external query-response pattern. Exposing a custom Apex invocable action as a public object would not give the insurer a governed, purpose-built way to query the agent's conversational capability, and it sidesteps the supported integration surface entirely.
Remember: who is calling whom decides the mechanism - external app calling an agent is Agent API territory.