Per the AgentForce Integration and AI Architecture Guide, the Model Context Protocol (MCP) enables agents to interact dynamically with external predictive or AI models. The documentation states:
“Through MCP, agents can discover, connect, and invoke external models via standardized schema definitions. This allows agents to use third-party tools—like recommendation engines or classifiers—without pre-coding fixed API calls.”
Option A (MCP) supports dynamic interoperability with predictive models, making it the correct answer.
Option B (Hugging Face) refers to a model hosting platform, not a Salesforce integration mechanism.
Option C (A2A protocol) supports agent-to-agent communication, not external model invocation.
Therefore, Option A correctly reflects the Salesforce-recommended method for integrating external predictive APIs.
References (AgentForce Documents / Study Guide):
AgentForce Architecture Guide: “Integrating Predictive Models with MCP”
AgentForce Technical Overview: “Model Context Protocol for External Tools”
AgentForce Study Guide: “Using MCP for Recommendation and AI Model Access”
