AI Assistant

Live Hub has a built-in AI Assistant — a chat interface in which you describe what you want in plain language, and the assistant does it: building and editing your configuration, testing it, troubleshooting past conversations, and answering questions about how the platform works.

To open the AI Assistant:

1. In the AI Agents section, click Build with AI at the top of the screen.

2. The assistant opens as a panel docked to the right of the screen. Type your request in the input field at the bottom and press Enter.

The panel stays open while you work inside the AI Agents section, so you can keep using the screens next to it. Drag its left edge to resize it — the width is remembered for your next visit.

What you can ask it

• Configuration — list, create, edit, rename, clone and delete agents, flows, tools, documents, post-call analysis configurations and test suites. For example: "Create an appointment-booking agent that uses my booking-api tool", or "Add https://example.com/faq as a document and attach it to the support agent".

• Learning the platform — the assistant has access to the latest version of this documentation and answers from it rather than from memory, so you can ask what a feature, field or setting does and how to accomplish something. For example: "What is active listening mode and when should I enable it?"

• Testing — start a test chat with one of your agents or flows on your behalf and report what happened, and run your test suites.

• Monitoring and troubleshooting — find past conversations, read their transcripts together with the tool calls and agent hand-offs, read post-call-analysis results, and explain why an agent behaved the way it did. For example: "Why did the last call end without a transfer?"

• Research — when a request needs information the assistant doesn't already have, it searches the web and reads the relevant pages as part of carrying it out. For example: "Build an agent that tells callers the current weather in the city they ask about, using a free weather API" — the assistant finds a suitable public API, reads its documentation pages to learn the request format, then creates the tool with the right URL and parameters and wires it into the new agent.

Working with the assistant

While the assistant works, it shows its reasoning and each action it takes as a collapsed group of steps; expand it to see the individual operations and their results.

The assistant is instructed to look before it acts — inspecting the relevant entity and checking with you before it creates, changes or deletes anything — though on an explicit request it may simply go ahead and do it. The steps it shows are the record of what it actually did, so it's worth reviewing them. The assistant acts as the signed-in user and is limited to that user's account and permissions.

Under the hood it uses the same MCP server that external MCP clients connect to, so it has the same set of capabilities.

Closing the panel ends the conversation. Opening it again starts a fresh one: the assistant doesn't carry over previous chats, so include any context it needs in your request.

What the assistant doesn't cover

The assistant covers the AI Agents area of Live Hub. Other parts of the platform — bot connections, speech services (STT/TTS), routing rules and voice channels (SIP connections, phone numbers, WhatsApp, Teams) — are not part of the configuration it can change; you set those up yourself in the relevant Live Hub screens.

Charging

Conversations with the AI Assistant are charged in the same way as chat sessions with your own agents — based on the number of tokens consumed.