Add a connected agent (Customer Conversation) (preview)

[This article is prerelease documentation and is subject to change.]

Use connected agents to let your primary agent delegate a request to an independent agent hosted in Microsoft Foundry. Unlike a sub-agent, a connected agent keeps its own configuration and lifecycle management, separate from your primary agent.

Tip

Connected agents can't reference tools, topics, events, sub-agents, or variables from the primary agent's orchestration. See best practices for working with sub-agents for a full comparison of what primary agents, sub-agents, and connected agents can each reference.

Prerequisites

Prepare your Foundry agent

Before you can add a connected agent, locate and publish the agent in the Microsoft Foundry portal, then copy the values you'll need to create the connection:

  1. In the Microsoft Foundry portal, locate the agent you want to connect.
  2. Publish the agent.
  3. Copy the Azure AI Project Endpoint for the agent's project.
  4. Copy the agent's Agent ID (the agent's name).

Add a connected agent

  1. Sign in to Copilot Studio.

  2. Open the primary agent you want to add a connected agent to.

  3. On the Build tab, select the primary agent's card to open its configuration panel.

  4. In the components panel, select Connected agents to open the Add a connected agent dialog.

  5. Select the connection to use for Microsoft Foundry, or create a new connection by providing the Foundry project endpoint URL in the format https://{account-name}.services.ai.azure.com/api/projects/{project-name}, then select Next.

  6. Enter a Name and Description for the connected agent. The description should explain the purpose of the Foundry agent so the orchestrator knows when to route requests to it.

  7. Enter the Agent Id for the Microsoft Foundry agent you want to connect to.

  8. Select Add Agent.

  9. Confirm the connected agent appears in the components panel.

  10. Publish your primary agent. Publishing is required so it can communicate with Azure; you don't need to deploy it to a specific channel yet. See Deploy an agent (Customer Conversation).

Caution

The model you select for the agent in Microsoft Foundry affects how long it takes to respond to a customer's question. Choose a model with response latency appropriate for your scenario, especially if the agent will be published to a voice channel.