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

# Creating and Managing Agents

> Step-by-step guide to creating, configuring, and managing your AI agents

## Accessing the Agents Page

Navigate to the **Agents** page from the sidebar to view and manage all your agents. This page displays:

* All agents organized by workspace
* Quick actions for creating and importing agents
* Search and filtering tools
* Agent cards with key information

## Creating a New Agent

There are three ways to create an agent in the platform:

### Method 1: Create from Scratch

<Steps>
  <Step title="Open the New Agent Dialog">
    Click the **New Agent** button in the top-right corner of the Agents page.
  </Step>

  <Step title="Configure Basic Information">
    Provide the following details:

    * **Agent Name**: A clear, descriptive name for your agent
    * **Description**: Explain what this agent does (optional but recommended)
    * **Icon**: Choose an emoji or icon to represent your agent
  </Step>

  <Step title="Set Up Agent Behavior">
    Define how your agent should operate:

    * **System Instructions**: Provide detailed instructions about the agent's role, behavior, and goals
    * **Model Selection**: Choose the AI model that will power your agent
  </Step>

  <Step title="Add Tools and Integrations">
    Connect the tools and integrations your agent needs to perform its tasks. You can add these in the Tools and Integrations tabs.
  </Step>

  <Step title="Configure Model Settings">
    Select the AI model, adjust temperature, and enable agent abilities like Python code execution, web search, or computer use.
  </Step>

  <Step title="Save and Deploy">
    Click **Save** to create your agent. It will be immediately available for use.
  </Step>
</Steps>

### Method 2: Browse Templates

Start from pre-built templates designed for common use cases:

<Steps>
  <Step title="Open Templates">
    Click the **Browse Templates** button on the Agents page.
  </Step>

  <Step title="Select a Template">
    Browse available templates and choose one that matches your needs.
  </Step>

  <Step title="Customize the Template">
    Modify the template's configuration to fit your specific requirements.
  </Step>

  <Step title="Save Your Agent">
    Save the customized agent to your workspace.
  </Step>
</Steps>

<Tip>
  Templates are a great way to get started quickly. You can always customize them further after creation.
</Tip>

### Method 3: Import an Agent

Import agents from files or other sources:

<Steps>
  <Step title="Click Import">
    Select the **Import** button on the Agents page.
  </Step>

  <Step title="Upload Configuration">
    Upload your agent configuration file (JSON format).
  </Step>

  <Step title="Review and Adjust">
    Review the imported configuration and make any necessary adjustments.
  </Step>

  <Step title="Complete Import">
    Save the imported agent to add it to your collection.
  </Step>
</Steps>

## Managing Existing Agents

### Viewing Agent Details

Each agent is displayed as a card showing:

* **Icon/Emoji**: Visual identifier for the agent (clickable to change)
* **Name**: The agent's display name
* **Type Badge**: Indicates the entity type (AGENT)
* **Description**: Brief explanation of the agent's purpose
* **Action Buttons**: Quick access to configure or chat with the agent

### Agent Header Information

When editing an agent, the header displays:

* **Agent Icon**: Click to change the emoji or icon
* **Agent Name**: Edit directly in the header
* **Description**: Add or modify the agent description
* **Security Groups**: Shows how many groups have access to this agent
* **Action Buttons**:
  * **Save**: Save all configuration changes
  * **Download spec**: Export the agent configuration as a file
  * **Chat**: Test the agent in chat mode
  * **Close (X)**: Exit the configuration panel

### Configuring an Agent

To modify an agent's settings, click the **Configure** button on the agent's card. This opens the agent configuration panel with the following tabs:

#### Prompt Engineering

Define how your agent should behave and respond to user inputs.

* **Instructions**: Create structured instructions organized into sections
* **Add Section**: Group related instructions for better clarity
* **Add Instruction**: Define specific behaviors and guidelines for your agent

<Tip>
  Organize your instructions into logical sections (e.g., "Role & Purpose", "Guidelines", "Constraints") for better maintainability.
</Tip>

#### Prompt Presets

Create predefined prompts that users can quickly select in chat.

* Prompt presets appear as clickable tiles in the chat interface
* Useful for common queries or tasks
* Helps users get started quickly with your agent

#### Integrations

Add and configure integrations that your agent can use to connect with external services:

<AccordionGroup>
  <Accordion title="Database Integrations">
    * **ClickHouse Database**: Connect to ClickHouse columnar databases
    * **SQL Database**: Integrate any standard SQL database with read-only access
  </Accordion>

  <Accordion title="Cloud Services">
    * **Azure**: Microsoft's MCP Server for Azure cloud services
    * **Amazon Web Services (AWS)**: AWS cloud services management and operations
  </Accordion>

  <Accordion title="Development Tools">
    * **GitHub**: Repository management, issues, pull requests, and more
  </Accordion>

  <Accordion title="Business Applications">
    * **MYOB**: Atlastix's MCP server for MYOB accountancy software
    * **Xero**: Payroll and bank management
    * **Xero OAuth2**: Xero with OAuth2 authentication
  </Accordion>

  <Accordion title="Documentation">
    * **AWS Knowledge Base**: Vector search for AWS documentation
  </Accordion>
</AccordionGroup>

Each integration can be added with the **Add** button and configured with specific credentials and settings.

#### Tools

Add and configure custom tools that extend your agent's capabilities to perform specific tasks. Tools allow your agent to:

* Execute custom actions
* Query databases
* Process files
* Call external APIs
* And more

#### Subagents

Configure subagents that can be called by this agent for specialized tasks. Subagents help break down complex workflows by:

* Delegating specific tasks to specialized agents
* Creating hierarchical agent structures
* Improving modularity and reusability

You can select from existing agents in your workspace to use as subagents.

#### Model Settings

Configure advanced model parameters and behavior settings:

<AccordionGroup>
  <Accordion title="Language Model">
    * **Model**: Select the LLM (e.g., Claude 4.5 Sonnet) that powers your agent
    * **Temperature**: Control randomness (0 = deterministic, higher = creative)
  </Accordion>

  <Accordion title="Agent Abilities">
    * **AI Python Code Execution**: Enable the agent to write and execute Python code
    * **Web Search**: Allow the agent to search the web for information
    * **Computer Use**: Enable interaction with web browsers and interfaces
    * **Artifact Generation**: Choose artifact generation style (Classic, Tool-based, or Disabled)
    * **Parallel Tool Calls**: Allow the agent to call multiple tools simultaneously
  </Accordion>

  <Accordion title="Execution Settings">
    * **Maximum Number of Tool Retries**: How many times the agent retries a failed tool before marking the task as failed
    * **Timeout Value**: How long an agent can work on a task before timing out
    * **Maximum Tokens**: Maximum number of tokens to generate in a response
  </Accordion>
</AccordionGroup>

#### Metadata

Add additional information and categorization for organization and discovery:

* **Tags**: Add tags to help categorize and find your agent
* **Category**: Assign your agent to a category (e.g., Documentation, Support, Analysis)
* **User Documentation**: Add links or notes about how to use this agent

#### Version History

View and track changes made to your agent over time. Version history is automatically created each time you save your agent, allowing you to:

* Review past configurations
* Restore previous versions if needed
* Track who made changes and when

<Note>
  Version history begins tracking after you save the agent for the first time.
</Note>

### Saving Your Configuration

After making changes in any tab:

1. Click the **Save** button in the top-right corner to apply your changes
2. Use **Download spec** to export your agent's configuration as a file
3. Click **Chat** to test your agent immediately after saving

### Chatting with an Agent

Test and interact with your agent directly:

1. Click the **Chat** button on the agent's card
2. Start a conversation to test the agent's responses
3. Verify that the agent has access to the correct tools and knowledge

<Note>
  Chatting with an agent is a great way to test its configuration before deploying it in production workflows.
</Note>

### Additional Agent Actions

Click the three-dot menu (⋯) on any agent card to access additional options:

* **Edit**: Modify the agent's configuration
* **Duplicate**: Create a copy of the agent
* **Export**: Download the agent's configuration as a file
* **Delete**: Remove the agent from your workspace

<Warning>
  Deleting an agent is permanent and cannot be undone. Scheduled tasks using this agent will also be affected.
</Warning>

## Organizing Agents

### Workspaces

Organize your agents using workspaces:

* **Create Workspaces**: Group related agents by team, project, or purpose
* **Move Agents**: Drag agents between workspaces or use the move option in the menu
* **No Workspace**: Agents without a workspace assignment appear in the "No Workspace" section

### Search and Filter

Quickly find agents using the search and filter tools:

<Accordion title="Search">
  Use the search bar to find agents by name or description. Search is real-time and updates results as you type.
</Accordion>

<Accordion title="Filter by Category">
  Use the category dropdown to filter agents by type or classification.
</Accordion>

<Accordion title="Filter by Tags">
  Click the **Tags** button to filter agents by assigned tags.
</Accordion>

<Accordion title="Sort Options">
  Change the sort order using the sort dropdown:

  * Alphabetically (A-Z)
  * Recently created
  * Recently modified
  * Most used
</Accordion>

### View Options

Toggle between different view modes:

* **Grid View**: See agents as cards with full details (default)
* **List View**: See agents in a compact list format

## Best Practices

<CardGroup cols={2}>
  <Card title="Clear Naming" icon="tag">
    Use descriptive names that clearly indicate each agent's purpose and scope.
  </Card>

  <Card title="Detailed Instructions" icon="file-lines">
    Provide comprehensive system instructions to ensure consistent agent behavior.
  </Card>

  <Card title="Test Thoroughly" icon="vial">
    Always test agents via chat before using them in production workflows.
  </Card>

  <Card title="Regular Updates" icon="arrows-rotate">
    Review and update agent configurations regularly to maintain optimal performance.
  </Card>
</CardGroup>

## Troubleshooting

<AccordionGroup>
  <Accordion title="Agent not responding as expected">
    * Review system instructions for clarity and completeness
    * Verify that required tools and integrations are properly connected
    * Check that knowledge bases contain relevant information
    * Test with simpler queries to isolate the issue
  </Accordion>

  <Accordion title="Agent can't access certain tools">
    * Confirm tools are added in the agent's configuration
    * Verify tool permissions and API credentials
    * Check integration status in the Integrations section
  </Accordion>

  <Accordion title="Agent missing knowledge">
    * Ensure knowledge bases are connected to the agent
    * Verify knowledge base content is up to date
    * Check that documents are properly indexed
  </Accordion>
</AccordionGroup>

## Next Steps

<CardGroup cols={2}>
  <Card title="Connect Tools" icon="wrench" href="/tools/overview">
    Learn how to add tools to extend your agent's capabilities
  </Card>

  <Card title="Add Knowledge" icon="book" href="/knowledge-bases/overview">
    Connect knowledge bases to make your agent smarter
  </Card>

  <Card title="Set Up Integrations" icon="plug" href="/integrations/overview">
    Integrate external services with your agents
  </Card>

  <Card title="Schedule Tasks" icon="clock" href="/scheduled-tasks/overview">
    Automate agent execution with scheduled tasks
  </Card>
</CardGroup>