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

# Agent Guide

> Create, configure, and manage AI agents

## What is an Agent?

An agent is an AI chatbot that can:

* Answer questions based on your knowledge base
* Have conversations with users
* Learn from documents and URLs
* Integrate with external services
* Provide 24/7 customer support

## Creating an Agent

### Basic Creation

1. Go to your dashboard

2. Click **Create Agent** or **New Agent**

3. Fill in the required information:

   * **Name**: A descriptive name (e.g., "Customer Support Bot")
   * **Description**: What your agent does (optional)
   * **System Prompt**: How your agent should behave (optional)

4. Click **Create**

### Using Templates

1. Click **Create Agent**

2. Select **Use Template**

3. Choose from available templates:

   * Customer Support
   * Sales Assistant
   * FAQ Bot
   * Technical Support
   * Custom

4. Customize the template

5. Click **Create**

## Agent Configuration

### System Prompt

The system prompt defines your agent's personality and behavior. Examples:

**Customer Support Agent**:

```
You are a helpful customer support agent. Be friendly, professional, and concise.
Always try to solve the customer's problem. If you can't help, offer to escalate
to a human agent.
```

**Sales Assistant**:

```
You are a sales assistant. Help customers find the right products. Ask qualifying
questions. Be enthusiastic but not pushy. Always mention special offers when relevant.
```

### Model Selection

Choose the AI model for your agent:

* **GPT-4**: Most capable, best for complex tasks
* **GPT-4 Turbo**: Fast and capable
* **GPT-3.5 Turbo**: Fast and cost-effective
* **Claude 3 Opus**: Excellent for long conversations
* **Claude 3 Sonnet**: Balanced performance
* **Claude 3 Haiku**: Fast and efficient

### Model Parameters

* **Temperature** (0-2): Controls randomness. Lower = more focused, Higher = more creative
* **Max Tokens**: Maximum response length
* **Top P**: Nucleus sampling parameter

### Training Mode

Choose how your agent learns:

* **None**: No training, uses only system prompt
* **RAG** (Retrieval Augmented Generation): Uses knowledge base for context
* **Fine-tuned**: Uses a custom fine-tuned model
* **Hybrid**: Combines RAG and fine-tuning

## Knowledge Base

### Adding Documents

1. Go to your agent dashboard
2. Navigate to **Knowledge Base** or **Sources**
3. Click **Upload Document**
4. Select files (PDF, Word, Text)
5. Wait for processing to complete

### Adding URLs

1. Go to **Knowledge Base**

2. Click **Add URL**

3. Enter the URL

4. Configure scraping options:

   * **Single Page**: Scrape just this page
   * **Site Crawl**: Crawl the entire website
   * **Depth**: How many levels to crawl
   * **Max Pages**: Maximum pages to scrape

5. Click **Add**

### Connecting Notion

1. Go to **Knowledge Base**
2. Click **Connect Notion**
3. Authorize the connection
4. Select pages and databases to sync
5. Click **Sync**

See [Notion Integration](/integrations/notion) for detailed setup.

## Agent Settings

### General Settings

* **Name**: Agent name
* **Description**: Agent description
* **Status**: Active, Inactive, Training, Archived
* **Version**: Agent version number

### Advanced Settings

* **Embed Code**: Unique code for widget embedding
* **API Key**: Agent-specific API key
* **Organization**: Link to organization (if applicable)

## Widget Customization

Customize how your agent appears to users:

### Appearance

* **Primary Color**: Main brand color
* **Launcher Icon**: Icon style
* **Launcher Shape**: Circle or square
* **Widget Style**: Glass, solid, or minimal

### Behavior

* **Auto-open**: Automatically open widget
* **Auto-open Delay**: Delay before auto-opening
* **File Uploads**: Allow file uploads
* **Suggested Replies**: Show suggested responses

### Content

* **Title**: Widget header title
* **Subtitle**: Widget header subtitle
* **Initial Message**: First message shown
* **Suggested Questions**: Pre-defined questions

See [Widget Embed Guide](/guides/widget-embed) for details.

## Agent Analytics

View performance metrics:

* **Total Conversations**: Number of conversations
* **Total Messages**: Number of messages
* **Average Response Time**: Average time to respond
* **User Satisfaction**: Feedback ratings
* **Popular Questions**: Most asked questions
* **Conversation Flow**: How conversations progress

## Managing Agents

### Editing an Agent

1. Go to your dashboard
2. Click on the agent you want to edit
3. Make changes
4. Click **Save**

### Duplicating an Agent

1. Go to agent settings
2. Click **Duplicate**
3. Enter a new name
4. Click **Duplicate**

### Archiving an Agent

1. Go to agent settings
2. Click **Archive**
3. Confirm archiving

Archived agents are hidden but not deleted. You can restore them later.

### Deleting an Agent

1. Go to agent settings
2. Click **Delete**
3. Type the agent name to confirm
4. Click **Delete**

<Warning>This action cannot be undone.</Warning>

## Agent Limits

Plan-based limits on agents:

* **Free**: 1 agent
* **Starter**: 3 agents
* **Business**: 10 agents
* **Pro**: Unlimited agents

## Best Practices

### System Prompts

1. **Be Specific**: Clearly define the agent's role
2. **Set Boundaries**: Define what the agent should and shouldn't do
3. **Provide Examples**: Include example interactions
4. **Update Regularly**: Refine prompts based on performance

### Knowledge Base

1. **Quality Over Quantity**: Focus on relevant, accurate content
2. **Keep Updated**: Regularly update documents
3. **Organize Content**: Use clear file names and structure
4. **Test Coverage**: Ensure knowledge base covers common questions

### Testing

1. **Test Regularly**: Test agent responses before going live
2. **Monitor Performance**: Review analytics regularly
3. **Gather Feedback**: Collect user feedback
4. **Iterate**: Continuously improve based on feedback

## Troubleshooting

### Agent Not Responding

* Check agent status (should be "Active")
* Verify model is selected
* Check system prompt is set
* Review error logs

### Poor Responses

* Improve system prompt
* Add more knowledge base content
* Adjust temperature parameter
* Review conversation logs

### Knowledge Base Not Working

* Verify documents are processed
* Check knowledge base status
* Ensure training mode is set to "RAG"
* Review document quality

## Next Steps

<CardGroup cols={2}>
  <Card title="Knowledge Base Guide" icon="database" href="/guides/knowledge-base">
    Learn about knowledge base management
  </Card>

  <Card title="Widget Embed" icon="code" href="/guides/widget-embed">
    Customize your widget
  </Card>

  <Card title="API Reference" icon="terminal" href="/api-reference/introduction">
    Programmatic access
  </Card>

  <Card title="Integrations" icon="plug" href="/integrations/overview">
    Connect external services
  </Card>
</CardGroup>