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

# Create Your First Agent

> Step-by-step guide to building your first AI agent in PLai Framework

# Create Your First Agent

This comprehensive tutorial will guide you through creating your first AI agent in PLai Framework. You'll learn the fundamentals of agent creation, configuration, and testing.

<Frame>
  <img src="https://mintcdn.com/plaisolutions/NF5mtbzLHH-8qtIK/images/hero-light.png?fit=max&auto=format&n=NF5mtbzLHH-8qtIK&q=85&s=82cff8cb4b83372de52239abca668308" alt="Create Your First Agent" width="2064" height="1104" data-path="images/hero-light.png" />
</Frame>

## What You'll Build

In this tutorial, you'll create a **Customer Support Agent** that can:

<CardGroup cols={2}>
  <Card title="Answer Questions" icon="comments">
    Provide helpful responses to customer inquiries
  </Card>

  <Card title="Search Information" icon="magnifying-glass">
    Look up real-time information using Perplexity AI
  </Card>

  <Card title="Process Requests" icon="gear">
    Handle various customer service tasks
  </Card>

  <Card title="Escalate Issues" icon="arrow-up">
    Know when to escalate complex issues to humans
  </Card>
</CardGroup>

## Prerequisites

Before you begin, ensure you have:

* [ ] A PLai Framework account with Pro access (for Perplexity AI)
* [ ] Access to a project (create one if needed)
* [ ] Basic understanding of AI agents and customer service

## Step 1: Create the Agent

<Steps>
  <Step title="Access Your Project">
    1. Log in to [PLai Framework Dashboard](https://framework.plaisolutions.com)
    2. Select your organization and project
    3. Navigate to the **Agents** section
  </Step>

  <Step title="Start Agent Creation">
    Click the **Create Agent** button (+ icon) in the top right
  </Step>

  <Step title="Configure Basic Information">
    Fill in the agent details:

    **Name**: `Customer Support Agent`
    **Description**: `Helpful customer support agent that can answer questions and provide assistance`
  </Step>

  <Step title="Set System Instructions">
    Copy and paste this system prompt:

    ```text theme={null}
    You are a helpful and professional customer support agent. Your role is to:

    1. Provide accurate and helpful information to customers
    2. Maintain a friendly and professional tone
    3. Search for current information when needed
    4. Escalate complex issues to human agents when appropriate
    5. Always prioritize customer satisfaction

    Guidelines:
    - Be concise but thorough in your responses
    - Ask clarifying questions when needed
    - Provide specific solutions and next steps
    - If you can't help with something, explain why and suggest alternatives
    - Always end conversations on a positive note
    ```
  </Step>
</Steps>

## Step 2: Configure the Model

<Steps>
  <Step title="Select LLM Provider">
    Choose **OpenAI** as your provider (recommended for customer support)
  </Step>

  <Step title="Choose Model">
    Select **GPT-4 Turbo** for the best performance, or **GPT-3.5 Turbo** for cost efficiency
  </Step>

  <Step title="Adjust Parameters (Optional)">
    * **Temperature**: 0.3 (more focused responses)
    * **Max Tokens**: 1000 (sufficient for most support responses)
  </Step>

  <Step title="Save Configuration">
    Click **Save** to store your model configuration
  </Step>
</Steps>

## Step 3: Add Tools

Now let's enhance your agent with the Perplexity AI tool for real-time information lookup.

<Steps>
  <Step title="Navigate to Tools">
    Go to the **Tools** section in your project
  </Step>

  <Step title="Create Perplexity AI Tool">
    1. Click **Create Tool**
    2. Select **Perplexity AI**
    3. Configure the tool:

    **Name**: `Web Search Tool`
    **Model**: `sonar-pro`
    **System Message**: `Search for current information to help answer customer questions. Focus on official sources and recent information.`
    **Search Domain Filter**: Leave empty for general search
  </Step>

  <Step title="Test the Tool">
    Click **Test** and try a sample query like "latest iPhone features"
  </Step>

  <Step title="Add Tool to Agent">
    1. Return to your agent's settings
    2. Go to the **Tools** tab
    3. Click **Add Tool** and select your "Web Search Tool"
  </Step>
</Steps>

## Step 4: Configure Agent Settings

<Steps>
  <Step title="Set Agent Avatar (Optional)">
    Upload a professional avatar image or use the default
  </Step>

  <Step title="Enable Features">
    Turn on these recommended features:

    * **File Upload**: Allow customers to share screenshots or documents
    * **Voice Input**: Enable voice conversations (if supported)
  </Step>

  <Step title="Configure Guardrails">
    Set up basic safety measures:

    * Enable **Content Filtering**
    * Set **Rate Limiting** to 10 messages per minute
    * Enable **Inappropriate Content Detection**
  </Step>

  <Step title="Activate Agent">
    Toggle the agent status to **Active**
  </Step>
</Steps>

## Step 5: Test Your Agent

Now let's test your agent with realistic customer support scenarios.

<Steps>
  <Step title="Start a Test Chat">
    Click the **Chat** button on your agent to open the chat interface
  </Step>

  <Step title="Test Basic Interaction">
    Try this conversation starter:

    **You**: `Hi, I need help with my account`

    Your agent should respond professionally and ask for more details.
  </Step>

  <Step title="Test Tool Integration">
    Ask a question that requires current information:

    **You**: `What are the latest features in the iPhone 15?`

    Your agent should use the Perplexity tool to search for current information.
  </Step>

  <Step title="Test Escalation">
    Present a complex scenario:

    **You**: `I've been charged twice for the same order and need an immediate refund`

    Your agent should recognize this needs human intervention.
  </Step>
</Steps>

## Step 6: Refine and Optimize

Based on your testing, let's optimize the agent's performance.

<Tabs>
  <Tab title="Improve System Instructions">
    If responses aren't quite right, refine your system prompt:

    ```text theme={null}
    You are a helpful customer support agent for [Company Name]. 

    Key responsibilities:
    - Answer product questions using current information
    - Help with account and billing issues
    - Provide troubleshooting guidance
    - Escalate complex issues to human agents

    Response guidelines:
    - Keep responses under 150 words when possible
    - Use bullet points for step-by-step instructions
    - Always include next steps or follow-up actions
    - For billing issues, always escalate to human agents
    - For technical issues, try to resolve first, then escalate if needed

    Tone: Professional, helpful, and empathetic
    ```
  </Tab>

  <Tab title="Adjust Tool Configuration">
    Fine-tune your Perplexity tool:

    ```text theme={null}
    Search for the most current and accurate information to help answer customer questions. 
    Prioritize:
    1. Official company documentation
    2. Recent product announcements
    3. Authoritative news sources
    4. User guides and help articles

    Provide concise, actionable information.
    ```
  </Tab>

  <Tab title="Add More Tools (Optional)">
    Consider adding these tools as you grow:

    * **API Request Tool**: Connect to your CRM or help desk
    * **External Data Source**: Access customer database
    * **Browser Tool**: Scrape your own documentation
  </Tab>
</Tabs>

## Common Test Scenarios

Here are realistic scenarios to test your agent:

<AccordionGroup>
  <Accordion title="Product Information Requests">
    **Test Input**: "What's the difference between your Pro and Enterprise plans?"

    **Expected Behavior**:

    * Use web search tool to find current pricing information
    * Provide clear comparison
    * Offer to help with plan selection
  </Accordion>

  <Accordion title="Technical Support">
    **Test Input**: "I can't log into my account, it says my password is incorrect"

    **Expected Behavior**:

    * Provide password reset instructions
    * Suggest checking for caps lock, typos
    * Offer alternative solutions
    * Escalate if issue persists
  </Accordion>

  <Accordion title="Billing Issues">
    **Test Input**: "I was charged \$99 but I cancelled my subscription last month"

    **Expected Behavior**:

    * Acknowledge the concern empathetically
    * Explain it needs human review
    * Provide escalation path
    * Give reference number or next steps
  </Accordion>

  <Accordion title="Feature Questions">
    **Test Input**: "Does your platform support integration with Slack?"

    **Expected Behavior**:

    * Search for current integration information
    * Provide specific details about Slack integration
    * Explain setup process if available
    * Offer additional resources
  </Accordion>
</AccordionGroup>

## Advanced Configuration Tips

### Performance Optimization

<Tip>
  **Response Speed**: Use GPT-3.5 Turbo for faster responses if speed is more important than complexity handling.
</Tip>

```json theme={null}
{
  "model_config": {
    "provider": "openai",
    "model": "gpt-3.5-turbo",
    "temperature": 0.2,
    "max_tokens": 500,
    "top_p": 0.9
  }
}
```

### Cost Management

Monitor your agent's usage:

* **Token Usage**: Track input/output tokens per conversation
* **Tool Calls**: Monitor Perplexity AI usage
* **Conversation Length**: Set reasonable limits
* **Rate Limiting**: Prevent abuse

### Quality Assurance

<CardGroup cols={2}>
  <Card title="Regular Testing" icon="check">
    Test your agent weekly with new scenarios
  </Card>

  <Card title="User Feedback" icon="message">
    Collect and analyze user feedback
  </Card>

  <Card title="Performance Metrics" icon="chart">
    Monitor response quality and resolution rates
  </Card>

  <Card title="Continuous Training" icon="graduation-cap">
    Update system prompts based on learnings
  </Card>
</CardGroup>

## Troubleshooting Common Issues

<AccordionGroup>
  <Accordion title="Agent Not Responding">
    **Possible Causes**:

    * Agent is set to inactive
    * Model configuration errors
    * Network connectivity issues

    **Solutions**:

    * Check agent status is "Active"
    * Verify model configuration
    * Test with simple queries first
  </Accordion>

  <Accordion title="Tool Not Working">
    **Possible Causes**:

    * Tool not properly configured
    * Authentication issues
    * Tool not added to agent

    **Solutions**:

    * Test tool independently
    * Check tool configuration
    * Verify tool is assigned to agent
  </Accordion>

  <Accordion title="Poor Response Quality">
    **Possible Causes**:

    * Unclear system instructions
    * Wrong model selection
    * Inappropriate temperature setting

    **Solutions**:

    * Refine system prompt with specific examples
    * Try different model (GPT-4 for complexity)
    * Adjust temperature (lower for more focused)
  </Accordion>

  <Accordion title="High Costs">
    **Possible Causes**:

    * Using expensive models unnecessarily
    * Long conversations without limits
    * Excessive tool usage

    **Solutions**:

    * Use GPT-3.5 for simple tasks
    * Set conversation length limits
    * Monitor and optimize tool usage
  </Accordion>
</AccordionGroup>

## Next Steps

Congratulations! You've created your first AI agent. Here's what to do next:

<CardGroup cols={2}>
  <Card title="Deploy to Production" icon="rocket" href="/guides/production-deployment">
    Learn how to deploy your agent for real users
  </Card>

  <Card title="Add More Tools" icon="plus" href="/guides/multi-tool-setup">
    Enhance your agent with additional capabilities
  </Card>

  <Card title="Monitor Performance" icon="chart-line" href="/monitoring/overview">
    Set up analytics and monitoring
  </Card>

  <Card title="Advanced Configurations" icon="settings" href="/agents/analytics/overview">
    Explore advanced agent features
  </Card>
</CardGroup>

## Best Practices Summary

<Warning>
  **Production Readiness**: Always test thoroughly before deploying agents to handle real customer interactions.
</Warning>

✅ **Do's**:

* Test with realistic scenarios
* Set clear system instructions
* Monitor performance regularly
* Collect user feedback
* Start simple and iterate

❌ **Don'ts**:

* Deploy untested agents
* Use overly complex prompts initially
* Ignore cost monitoring
* Forget to set guardrails
* Skip regular updates

## Getting Help

If you encounter issues:

<CardGroup cols={3}>
  <Card title="Documentation" icon="book" href="/concepts">
    Review core concepts and features
  </Card>

  <Card title="Community" icon="discord" href="https://discord.gg/plaisolutions">
    Join our Discord community
  </Card>

  <Card title="Support" icon="life-ring" href="mailto:support@plaisolutions.com">
    Contact our support team
  </Card>
</CardGroup>

You've successfully created your first AI agent! This foundation will serve you well as you build more sophisticated agents and explore advanced features of PLai Framework.