A telecom's Foundry Agent Service deployment uses a general front-line agent to greet customers and answer simple questions. When a customer asks a detailed billing dispute question requiring access to billing-system tools and account-adjustment actions, the team wants the front-line agent to route the conversation to a specialized billing agent built with those specific tools, rather than giving every tool to the one front-line agent. Which Foundry Agent Service capability supports this design?

The correct capability is connected agents, where one agent can hand off a conversation to another specialized agent that has its own distinct tool set. In this scenario, the front-line agent can recognize that a detailed billing dispute falls outside its scope and transfer the customer to a billing agent equipped with billing-system tools and account-adjustment actions. This gives the team the separation of responsibility they want instead of loading every tool onto the front-line agent. The single-agent approach with longer system instructions describing billing procedures is wrong because it only changes what the front-line agent says; it does not give that agent access to billing-system tools or adjustment actions it was never equipped with. Increasing the front-line agent's maximum token output is wrong because a larger output token limit affects response length, not the agent's ability to call billing tools it does not have. The content safety filter configured to block billing-related keywords is wrong because a content safety filter blocks unsafe content categories; it has no role in routing a conversation to a more capable specialized agent and would only prevent the front-line agent from discussing billing at all.