Microsoft Foundry (formerly Azure AI Foundry) and the Power Platform can work together when a service is reachable through a suitable API. You plan authentication, data access, networking, runtime and operations for the specific use case.
A traceable architecture
A typical setup consists of an app or flow, a managed API endpoint and an Azure service. Azure API Management can expose APIs as Power Platform custom connectors. Microsoft documents this route for Power Apps, Power Automate and Copilot Studio. A custom connector describes operations and authentication, among other things; it replaces neither the API nor its access controls.
Before implementation, clarify:
- Which data may the solution send and return?
- Which identity authenticates the call, and how are secrets managed?
- Which environments, regions and network routes are involved?
- Which quotas, costs and failure cases need handling?
- Who reviews changes to the API, connector and flow before release?
Choosing the integration path
Use an existing connector if it covers the required operations and authentication types. A custom connector fits when your own API should appear as a reusable action in Power Apps, Power Automate or Copilot Studio. A direct HTTP call can suit a limited prototype, but it also needs secure authentication, error handling and a controlled release.
Governance and testing
Power Platform environments, roles and data policies control who builds solutions and which connectors may be used. On their own, they do not ensure that an agent or flow processes only suitable data. Check permissions with test accounts, log relevant calls and test failures, denied access and changes to the API contract.
Plan a small pilot with synthetic or approved data. Measure answer quality, response time and cost against criteria set in advance. Roll the solution out more widely only after security and user testing. Microsoft describes both exporting APIs from API Management as a custom connector and connectors and custom connectors in Copilot Studio.

