Your first agent

An agent is a brain (its instructions) and connections (your channels, your data). Here's how to build one, tab by tab.

The agent editor is organized into tabs: Discover, Brain, Skills, Knowledge, Scenarios and Deployment. You work through them in this order.

Discover

The Discover tab sums up what your agent does: it replies in the customer's language, draws on your knowledge, qualifies leads and hands off to your team for complex cases. It's the starting point, with an overview of what the AI has already handled across your conversations.

Write the brain

In Brain, you write the agent's instructions: what it knows, its tone, its rules. Type "/" to insert a capability (human handoff, product card, knowledge base…). The details are in The brain and instructions.

Choose the model

Still in Brain, choose the model that powers the agent and its level of reasoning. Free models work out of the box; premium models are enabled with your own key.

Test, then publish

Validate in the sandbox and the Scenarios, then click Publish to take the agent live. An Online / Offline switch lets you turn it off at any time.