Agent One LogoHelp Center
← Back to Homepage

Creating an Agent

This comprehensive guide will walk you through the process of creating, configuring, and deploying an AI agent on Agent One. Whether you're building a customer support chatbot, a knowledge assistant, or an interactive tool, this guide will help you set up your agent effectively.

Getting Started

Before you create your agent, it's helpful to:

  • Define a clear purpose for your agent
  • Gather relevant documentation and knowledge base materials
  • Identify any specific APIs or tools your agent will need
  • Determine whether your agent will be public or private

Step-by-Step Guide

1. Navigate to the Agent Creation Page

  1. Log in to your Agent One dashboard
  2. Click the "+ New Agent" button in the top right corner
  3. You'll see two options:
    • Public Agent: Accessible via a web URL, can be embedded on websites
    • Private Agent: For personal or team use only

Select the option that best suits your needs.

2. Configure Basic Settings

In the first setup page, you'll configure the fundamental details of your agent:

  • Name: Choose a descriptive name for your agent (e.g., "Support Assistant", "Product Expert")
  • Description: Write a brief description of what your agent does
  • Initial Message: This is the first message users will see when they start a conversation (e.g., "Hello! I'm your AI assistant. How can I help you today?")
  • Icon: Upload a custom image to represent your agent (recommended size: 192x192px)

Tips:

  • Keep the name concise but descriptive
  • Make the initial message friendly and clear about the agent's purpose
  • Consider your brand voice when writing these elements

3. Define Agent Behavior with Instructions

Instructions are crucial as they guide how your agent behaves and responds. This is where you define:

  • The agent's personality and tone: formal, friendly, technical, etc.
  • Specific knowledge areas: what the agent should know about
  • Restrictions: topics to avoid or actions not to take
  • Response format preferences: how information should be presented

Example instructions:

You are a customer support agent for Acme Tech, specializing in our cloud services.
- Maintain a helpful, friendly tone
- Answer questions about our cloud products, pricing, and technical specifications
- If asked about competitors, focus on our strengths without criticizing others
- When users report technical issues, collect specific details about their system
- For billing issues, direct users to our billing portal or suggest contacting billing@acmetech.com
- Format technical instructions as numbered steps
- Use emojis sparingly to maintain a professional tone

Tips:

  • Be as specific as possible in your instructions
  • Include example conversations if helpful
  • Update instructions periodically based on actual user interactions

3.5 Selecting the Right Model

Agent One offers different AI models to power your assistant. The model you choose affects your agent's capabilities, performance, and tool compatibility:

Available Models

  • Standard Models: These provide a balance of performance and capabilities, supporting all available tools.
  • o3-mini: A lightweight model that's faster but has some limitations, most notably:
    • Does not support the File Search capability
    • Automatically disables File Search if selected

Model Selection Considerations

When choosing a model, consider:

  1. Complexity of Tasks: More complex tasks may benefit from more capable models
  2. Response Time: Lighter models like o3-mini may provide faster responses
  3. Tool Requirements: If you need File Search capability, you cannot use o3-mini
  4. Content Length: Some models handle longer conversations better than others

If you change your model after configuration:

  • The system will automatically adjust tool availability
  • You may need to revise your instructions to align with the new model's capabilities
  • Always test your agent after changing models to ensure it performs as expected

4. Enable and Configure Tools

Agent One provides several powerful tools you can enable:

Code Interpreter

Useful for:

  • Generating code examples
  • Explaining programming concepts
  • Formatting data or performing calculations

To enable:

  1. Toggle the "Code Interpreter" switch to ON
  2. The agent will now be able to write, explain, and execute code when relevant

File Search

Allows your agent to search through your knowledge base:

  1. Toggle the "File Search" switch to ON
  2. Upload relevant documents in the Knowledge Base section
  3. The agent will automatically reference these documents when answering questions

Note: File Search is not available when using the o3-mini model.

Email Collection

For lead generation and follow-up:

  1. Toggle the "Email Collection" switch to ON
  2. Choose when to collect emails:
    • Beginning: The agent will ask for an email at the start of the conversation
    • End: The agent will ask before concluding the conversation
  3. Customize the prompt message if desired

Example prompt: "To better assist you, could you share your email address? This helps us follow up with additional information."

Custom API Functions

Connect your agent to external data and services:

  1. Click "Add API Function"
  2. Configure each API with:
    • Function Name: A descriptive identifier (e.g., "get_product_price")
    • Description: What the function does (e.g., "Retrieves current pricing for a specific product")
    • Endpoint URL: The API endpoint (e.g., "https://api.example.com/products/{product_id}/price")
    • Parameters: Define the input parameters in JSON schema format

Example parameter configuration:

{
  "product_id": {
    "type": "string",
    "description": "The unique identifier for the product",
    "required": true
  },
  "currency": {
    "type": "string",
    "description": "The currency code (USD, EUR, etc.)",
    "required": false,
    "default": "USD"
  }
}

Important notes:

  • APIs must be publicly accessible (no authentication required)
  • Only GET requests are currently supported
  • Test your APIs thoroughly before deployment

5. Knowledge Base Setup

Your knowledge base provides the information your agent can reference:

  1. In the Knowledge Base section, click "Upload Files"
  2. Select relevant documents (PDF, TXT, DOCX, etc.)
  3. Organize documents with descriptive names
  4. You can add context or descriptions to files after upload

Best practices:

  • Break large documents into logical sections
  • Use clear, descriptive filenames
  • Update documents when information changes
  • Remove outdated materials to prevent confusion

Your agent automatically indexes and uses this information when responding to queries.

6. Testing Your Agent

Before deploying your agent, thoroughly test it:

  1. Click the "Preview" button in the top right
  2. A test chat interface will appear
  3. Try different types of questions, including:
    • Common queries you expect
    • Edge cases or unusual questions
    • Questions that should trigger your custom APIs
    • Scenarios where email collection should activate

Testing checklist:

  • Does the agent maintain the correct tone and personality?
  • Are responses accurate and helpful?
  • Do tools like Code Interpreter work correctly?
  • Are API calls functioning as expected?
  • Is email collection triggered appropriately?

Make adjustments to your configuration based on test results.

7. Deployment Options

For Public Agents:

  1. Configure your site settings:

    • If you haven't created a site yet, you'll be prompted to do so
    • Choose between a subdomain (yourname.agnt.one) or custom domain
    • Complete site configuration (logo, colors, etc.)

  2. Get your embed code:

    • In your agent settings, find the "Embed" section
    • Copy the provided JavaScript snippet
    • Paste it into your website's HTML before the closing </body> tag

  3. Deployment methods:

    • Chat Widget: Appears as a chat bubble on your website
    • Full Page: A dedicated page for your agent
    • iFrame: Embed directly within a specific section of your site

For Private Agents:

  1. Your agent is immediately available in your dashboard
  2. Share access with team members if needed:
    • Go to Agent Settings > Access
    • Add team member emails
    • Set appropriate permission levels

Best Practices for Effective Agents

Instructions and Behavior

  • Be specific: Give clear examples of desired behavior
  • Define boundaries: Specify what the agent should and shouldn't do
  • Update regularly: Refine instructions based on actual usage
  • Test edge cases: Try unusual queries to ensure proper handling

Knowledge Base

  • Keep organized: Use clear naming conventions
  • Update regularly: Remove outdated information
  • Cover common questions: Include FAQs and frequent topics
  • Use quality sources: Ensure information is accurate and authoritative

API Integration

  • Simplify endpoints: Keep APIs focused on specific functions
  • Handle errors gracefully: Test edge cases and parameter variations
  • Provide clear descriptions: Help the agent understand when to use each API
  • Test thoroughly: Verify all parameter combinations

Continuous Improvement

Your agent gets better over time with these practices:

  1. Review conversations: Regularly check actual user interactions
  2. Update knowledge: Add new information as products or services evolve
  3. Refine instructions: Adjust based on how users actually interact
  4. Expand capabilities: Add new APIs or tools as needed
  5. Collect feedback: Consider adding a feedback mechanism

Troubleshooting Common Issues

Agent Doesn't Use Knowledge Base

  • Check file formats (PDF, DOCX, TXT are best)
  • Verify files are properly uploaded
  • Ensure questions relate to the content

Custom APIs Not Working

  • Confirm the endpoint is publicly accessible
  • Check parameter formatting
  • Verify URL structure and variable placement
  • Test the API independently before integration

Agent Tone Issues

  • Review and refine your instructions
  • Add specific examples of desired tone
  • Include sample responses in instructions

Next Steps

After creating your agent:

  1. Explore the Customizing Assistants guide for advanced configuration
  2. Learn about Managing Sites for public agent deployment
  3. Check the FAQs for quick answers to common questions

Remember that agent creation is an iterative process. Monitor performance, gather feedback, and make regular improvements to ensure your agent effectively serves its purpose.