> ## Documentation Index
> Fetch the complete documentation index at: https://fliqr.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Create and train an AI Agent

> Create a Fliqr AI Agent, connect a model provider, add business data via prompts and files, and start answering customer messages.

An AI Agent replies from your business data — not only the model’s general knowledge. Create the agent, attach a provider, then train it with a prompt and documents.

## Prerequisites

* A Fliqr AI workspace
* A connected model provider — see [AI Providers](/docs/core-concepts/ai-agents/ai-providers)
* Source content: policies, FAQs, product docs, or support macros

## Create the Agent

<Steps>
  <Step title="Connect a provider">
    Go to **Settings → Integrations** and connect **OpenAI**, **Google Gemini**, **Claude**, **DeepSeek**, or **xAI** with your API key (or use included quota where available).
  </Step>

  <Step title="Open AI Agents">
    Go to **AI Agents** (or **AI Center**) and create a new agent — or use the default agent created when you first connect a provider.
  </Step>

  <Step title="Name the agent and pick a model">
    Use a clear name (`Support — Tier 1`). Start with a fast/cheap model for drafting, then evaluate a stronger model before go-live.
  </Step>

  <Step title="Write the system prompt">
    Add role, constraints, and language rules. See [Prompt engineering](/docs/core-concepts/ai-agents/prompt-engineering).
  </Step>

  <Step title="Add business data">
    Either expand the system prompt with key facts, or attach [knowledge sources](/docs/core-concepts/ai-agents/knowledge-sources) and [AI files](/docs/core-concepts/ai-agents/files).
  </Step>

  <Step title="Test in the Playground">
    Ask real customer questions, including ones that should refuse or hand over.
  </Step>

  <Step title="Activate">
    Set the agent as the channel default, or call it from a Flow / Default Reply. See [Keywords and priority](/docs/core-concepts/ai-agents/keywords-and-priority).
  </Step>
</Steps>

## Two ways to provide business information

| Method                    | Best for                                      |
| ------------------------- | --------------------------------------------- |
| **System prompt**         | Tone, hard rules, short product facts         |
| **Documents / knowledge** | Long policies, catalogs, help-center articles |

Upload files under **AI Center → AI Tools → AI Files**, then select them on the agent. Details: [Files](/docs/core-concepts/ai-agents/files).

<Tip>
  Put irreversible rules in the prompt (“never invent pricing”). Put bulky reference material in knowledge/files so you can update docs without rewriting the prompt.
</Tip>

## Next

<CardGroup cols={2}>
  <Card title="Knowledge sources" icon="book" href="/docs/core-concepts/ai-agents/knowledge-sources">
    Index URLs, PDFs, and YouTube transcripts.
  </Card>

  <Card title="Functions & actions" icon="bolt" href="/docs/core-concepts/ai-agents/functions-actions">
    Let the agent fetch live data and trigger Flows.
  </Card>
</CardGroup>