In this video tutorial, you’ll learn how to build, test, and deploy a multi-agent system using the Visual Builder in Oracle AI Data Platform (AIDP). By coordinating Supervisor and Executor Agents in a single workflow, you can automate specialized business tasks such as expense compliance while applying guardrails, memory controls, and governed access to enterprise data.
In this video tutorial, you’ll learn how to:
- Build a multi-agent workflow with a Chat Trigger, Supervisor Agent, and specialized Executor Agents.
- Connect an Executor Agent to Oracle Analytics Cloud using an MCP connection to answer spending and budget questions.
- Apply guardrails to help block unsafe content and prompt injection attempts.
- Configure Private State Isolation and conversation history to control how memory is shared across agents.
- Test the multi-agent system in the Playground and use traces to understand how tasks are delegated across agents.
- Validate the complete workflow or individual agents and verify that guardrails are applied to incoming messages and generated responses.
- Deploy the agent to AI Compute and access endpoint URLs for communicating with the deployed agent.
By combining specialized agents, governed enterprise data, guardrails, and approved tools, Oracle AI Data Platform provides a visual, low-code approach to building multi-agent systems for business workflows. Watch the full tutorial to see how to build, test, and deploy an expense compliance agent from end to end.
For more details on building and configuring multi-agent systems, explore the Oracle Help Center documentation on Agent Creation and Guardrails.
