AI Agents

An AI Agent is the assistant you train, configure and deploy. It is the unit everything else attaches to — knowledge, channels, settings, analytics and team access.

What one agent contains

• Knowledge — the pages, documents and product data it has learned
• Settings — triggers, operating hours, collected fields
• Appearance — widget colours, avatar, welcome message
• Channels — website, WhatsApp, Telegram, Messenger, Instagram, Shopify, Gmail
• History — every conversation, lead, booking and issue it produced

How it answers

The agent does not guess from generic internet knowledge. It searches your trained content first, then answers in your tone. When it genuinely has no answer, it says so and offers a human instead of inventing something.

Multiple agents

Each agent is fully isolated. Common setups:

• One agent per website or brand
• A sales agent and a support agent on the same site, with different tone and triggers
• One agent per language or region
• A staging agent you experiment with before touching the live one

Your plan decides how many you can run — see Plans & Limits.

What makes it different from a chat widget

A widget forwards messages. An agent finishes jobs: it qualifies the lead, books the meeting, confirms the order, chases the abandoned cart and escalates only what actually needs a person.

Best practices

• Train on your full site before going live — thin knowledge is the top cause of weak answers
• Keep the welcome message short and specific to what you sell
• Use triggers wisely: 3–5 messages for lead capture
• Re-test in Playground after every retrain
• Give each agent one clear job instead of one agent doing everything