Introduction
Custom, code-first agents become genuinely useful only when they can act on real business data and work alongside the agents your organization already runs in production. Safely connecting a code-first agent to live Microsoft Power Platform data, and coordinating it with an existing Microsoft Copilot Studio agent, determines two things. It determines whether automation reflects current reality instead of stale assumptions, and whether employees get one coherent answer instead of several conflicting ones.
Scenario
Consider an internal equipment-provisioning system built on a custom Dataverse solution. Employee, device, request, and allowance tables track who's eligible for replacement equipment, what's currently available, and how a request moves through approval and fulfillment.
A Microsoft Copilot Studio agent already answers employee questions about request status, and a supplied code-first Microsoft Foundry agent coordinates the fulfillment logic behind the scenes. Someone still has to connect that Foundry agent to live Dataverse records, and decide where a broader option like Work IQ fits.
They also need to link the Foundry agent to the status agent, and let other applications consume that status agent directly. Treat this scenario as illustrative: the pattern it demonstrates applies to whatever data and agents you connect in your own work.
What you learn
You explore how Microsoft Power Platform Model Context Protocol (MCP) servers, including the Dataverse MCP server, give a code-first agent direct, governed access to Dataverse data. You also examine where Work IQ fits as a broader option for that same integration. You then learn how to connect a custom agent to an existing Copilot Studio agent, and how application code consumes a Copilot Studio agent through the Copilot Studio client library.
By the end of this module, you can connect a custom agent to live Power Platform data and to an existing Copilot Studio agent. You can also enable applications to consume that Copilot Studio agent directly.