Agents

The Agent is the core building block of the Live Hub AI Agents framework. Each agent defines, through natural-language instructions, how it should behave and what it can do.

An agent is built from the following:

• Prompt – instructions that define the agent’s identity, objectives, and behavior.
• Model – the large language model that powers the agent, together with settings such as temperature and output limits.
• Knowledge – documents that extend the agent’s knowledge and ground its answers in your own content.
• Tools – actions the agent can take on external systems, for example to look up an order or book an appointment.
• Variables – values reused across the prompt, welcome message, and tools.
• Post-call analysis – tools that summarize the conversation or extract structured data once the call ends.

A single agent can handle a complete task on its own — for example, an 'Appointments Agent' for scheduling, or a 'Q&A Agent' for business inquiries — each reachable on its own phone number.

For more complex use-cases, several agents can collaborate, exchanging messages so that one agent can route the caller to the right specialist. For details, see Multi-agent topologies.