Configuring AI agents

Each agent is a standalone AI assistant with its own identity, knowledge scope, and deployment channels.

Agent setup steps

1

Identity & tone

Set the agent's name, avatar, greeting message, and tone (Professional, Friendly, or Direct). You can also write custom base instructions to shape how it responds.

2

Attach knowledge bases

Choose which knowledge bases this agent can access. You can restrict an IT support agent to technical manuals only, while a sales agent reads product catalogs exclusively.

3

Add tools (optional) Technical

Attach custom API functions or webhooks the agent can call on behalf of users, such as looking up order status or submitting a support ticket.

4

Deploy to channels

Generate a web embed snippet, or connect to a Slack app. One agent can serve multiple channels simultaneously.

Every agent response passes through the AI Critic, a secondary model that checks answers against your knowledge base before they reach the user, blocking hallucinations automatically.

Model selection Technical

By default, agents use the model tied to your API key. You can override this per agent under Advanced Settings → Model, which is useful for routing complex tasks to a more capable model while keeping simpler agents on a faster, cheaper one.