Understand Responses and Agent Behavior
As you work in a chat, Agent Hub provides information that helps you understand how your request is being handled. Depending on your request and how you start the chat, Agent Hub can use an agent or tool to process the request or provide a direct response using its configured large language model (LLM).
Understanding how Agent Hub handles requests can help you interpret responses and decide whether to continue the current conversation or start a new chat.
For information about how Agent Hub determines how to handle a request, see How Agent Hub Processes Requests.
Identify the Agent Associated with a Chat
When you start a chat by selecting or tagging an agent, that agent remains associated with the conversation. To work with a different agent, start a new chat.
When you start a chat without selecting or tagging an agent, the orchestrator can invoke different agents for successive requests within the same conversation. To see which agent handled a particular request, click the processing status indicator in the chat to open Thinking Activity.
Understand Direct Responses from Agent Hub
Not every request requires a specialized agent or tool. When you enter a prompt without first selecting an agent, the orchestrator evaluates your request and determines how to handle it.
If a specialized agent or tool isn't required, Agent Hub can use its configured LLM to provide a direct response.
A direct response is different from a response provided by a registered agent. An agent can have specialized instructions, tools, and capabilities configured for a particular business purpose, while a direct response uses the LLM configured for Agent Hub.
For information about configuring LLM, see How Agent Hub Uses Large Language Models (LLMs).
See How Agent Hub Processes Requests.
Understand When Agent Hub Uses a Tool
Agent Hub can identify and use an appropriate tool when processing a request. Unlike agents, tools aren't resources that you browse and select directly from the Registry.
To see which tool Agent Hub invoked, click the processing status indicator (for example, Completed in 18s) in the chat to open the Thinking Activity panel. The panel displays the name of the tool used to process your request.
Continue Working with an Agent
Enter follow-up prompts to continue the conversation. In an orchestrated conversation, Agent Hub can invoke different agents as needed and provide relevant information from earlier prompts and responses to an agent handling a follow-up request. The receiving agent determines how to use and retain that information.
If you selected or tagged an agent when starting the chat, that agent remains associated with the conversation and is responsible for maintaining its own conversation context.
For example, after asking an agent to summarize your open purchase orders, you might follow up with a request such as Show only the purchase orders that are more than 30 days overdue.
Know When to Start a New Chat
Start a new chat when you want to begin a separate task or explicitly select a particular agent. If you started the current chat by selecting or tagging an agent, you must start a new chat to work with a different agent.
In an orchestrated conversation, you can continue submitting requests without starting a new chat when your work requires different agents.
Review Responses
Review Agent Hub responses before relying on the information provided or taking further action. Depending on the request and the capabilities used to process it, a response can include text, supporting information, generated files, or interactive content.
If a response doesn't provide the result you need, use a follow-up prompt to clarify or refine your request.
When Agent Hub requests approval before performing an action, review the requested action before deciding whether to approve it. See Understand Approvals in Agent Hub.
For information about the processing information displayed while Agent Hub handles a request, see Understand Request Activity.
For help resolving problems with a response or agent behavior, see Troubleshooting.
Why this matters
Understanding how Agent Hub handles a request can help you interpret the response you receive and decide what to do next. You can refine your request with a follow-up prompt, continue an orchestrated conversation that uses different agents as needed, or start a new chat when you want to begin a separate task or explicitly select a different agent.