Manage agents¶

With the Insights Hub Copilot Studio, a user can directly create custom agents using a comfortable wizard.

An agent is an AI-powered assistant that helps users to complete tasks. It is typically invoked by a user question (prompt). To answer the question, it can leverage knowledge from an LLM, pre-built skills/sub-agents provided by Insights Hub, or custom skills/sub-agents that support dedicated use cases. The agent instructions are the guidelines for the agent on how to answer the question using the provided skills.

A skill is a tool used to customize behavior with dedicated logic, data, and analytics. VFC skills run in the user's context. This means that if, for example, the read asset node is used and a user does not have access to all assets, the node will provide only the relevant subset of assets.

A sub-agent wraps an existing agent and enables it to be called within a nested agent structure. This allows building advanced, multi-step workflows where a parent agent can delegate tasks to one or more sub-agents. For more details, see Manage Sub Agents.

A user who is a tenant admin can create their own agents to solve specific problems.

After the agent is created, the configuration can be checked in Copilot Studio.

A user of Production Copilot can choose one of the available/enabled agents via a dropdown inside a supported application, such as Insights Hub Monitor.

Pre-built and custom agents¶

• The Manage agents view shows available pre-built or custom agents.

  • Custom agents can be created within the tenant context using pre-built or custom skills.
  • Pre-built agents are provided by Insights Hub and cannot be changed. A user can copy pre-built agents and adapt them for custom needs.

• By selecting an agent, the details can be viewed or parameters copied for further work in the APIs.

• A graphical view visualizes the relationships between the agent, its assigned sub-agents, and skills, providing a clear overview of the agent's structure.

• Agents can be enabled for productive use. Only enabled agents will be shown in the dropdown of the copilot chat - for example used in Insights Hub Monitor.

Create custom agent¶

• A wizard helps create or edit agents.

Wizard Screen	Description
	The agent name will be used e.g. in the dropdown menu in Insights Hub Monitor.
	The instructions are most important in determining how your agent will work and respond to questions. Be inspired by pre-built agents, and review their instructions. Documentation can also be found here.
	Assign sub-agents or skills the agent can use. You can also refer to them in the instructions like "Use skill xyz when asked for abc."
	Predefined prompts are shown in the Copilot chat at the bottom when selecting this agent. This provides an easier entry point for users to start with a common or frequently asked question.
	With clients, developers can control in which application the agent should appear. No client means it could appear in every application. Setting it to monitor will limit the use of the agent in Insights Hub Monitor.
	Different models can be used for an agent. Currently available are - gpt-5.4 (for advanced agents) - gpt-5.4-mini (for faster responses) - gpt-4o (deprecated / please switch to newer models) The timeout controls how long an agent can run before a timeout is shown in the chat.
	A summary validating the overall input.

Test an agent¶

• An agent can be tested when it is selected.
• After changing an agent instruction, a new chat must be started for the changes to take effect.

Export an agent¶

Agents can be exported, including all referenced sub-agents and skills. This allows you to transfer agent configurations between tenants or environments. For details on the export and import process, see Export & Import Agents and Skills.