Model Context Protocol (MCP)
- MCP
brings authoritative external data to agents, while A2A protocol lets
agents in Fusion Applications collaborate with other Oracle and third
party agents.
Agent-to-Agent (A2A)
- MCP
and A2A are implemented in Oracle AI Agent Studio with Fusion Applications
security and role-based access controls, so you can expand AI adoption
with confidence.
MCP for data retrieval
MCP allows an agent to securely access external tools or data sources in a structured way. Rather than relying on the general knowledge of a large language model, the agent can use MCP to retrieve authoritative information from a third-party system.
For example,
a financial management agent can use this protocol to pull foreign exchange rates from a service provider to complete a financial transaction. This ensures that the financial management agent remains in control while grounding its responses in accurate, up-to-date data from an authoritative third party outside of Fusion Applications.
A2A for agent collaboration
In contrast, A2A allows agents running in different systems to communicate securely so they can work together. With A2A, one agent can delegate tasks to another and then incorporate the output into a broader workflow. Those agents can work together across systems (i.e., agents deployed in Fusion Apps working with non-Fusion agents) to complete more complex, longer-running requests.
For example,
consider the Oracle Fusion Cloud HCM
Interview Management Assistant that helps automate scheduling with candidates.
In cases where a site visit is required, the Interview Management Assistant
could use A2A to call a Navan agent or Amex GBT agent to coordinate flight and
hotel bookings that align with the interview schedule, without the recruiter
leaving the Fusion UI.