Skip to main content

AI Agent Setup Tutorial

This guide shows how to create a production-ready assistant on the Flametree platform. You’ll build an AI agent that answers user questions, uses an FAQ or knowledge base for accurate replies, collects contact details, and suggests times for calls.

You’ll define the agent’s identity and goal, choose how it communicates, connect data sources, and set up a simple workflow to manage conversations. By the end, you’ll have a working prototype ready to use.

Before you start, make sure you have a Flametree account and know the basics of YAML. If possible, prepare some FAQ content or documents for the knowledge base.

Agent Build Flow​

  1. Create your agent — Set up a new AI agent and choose the correct type. You’ll start with a text-based agent that can handle written conversations.
  2. Define the agent’s role — Describe who the agent is and what it helps users with. This determines its tone and behavior in chats.
  3. Set the agent’s task — Specify the main goal — what the agent should achieve in a session and under what conditions it acts.
  4. Choose communication style & language — Choose the languages and tone your agent will use — professional, friendly, or neutral — and allow automatic language detection.
  5. Add FAQ — Create short question-and-answer pairs for common topics to make responses faster and more accurate.
  6. Set up the knowledge base — Upload documents or structured data so the agent can answer based on reliable information.
  7. Configure the workflow — Define how the conversation flows — how the agent starts, what tools it can use, and when it should finish.
  8. Define data collection — Decide which information the agent should capture, such as name, email, or product interest.
  9. Test your agent — Use the Playground to simulate user conversations and check how your agent handles questions and actions.
  10. Monitor and analyze — Review session logs and analytics to understand user behavior and improve responses over time.