Creating Agents in Ax
Learn how to create and configure agents in Ax to handle specific tasks and workflows.
Getting Started
- 1. Navigate to the Dashboard
- 2. Click on Agents in the left sidebar
- 3. Click "Create Agent" or select an existing agent to modify
Basic Configuration
Name (Required)
- • Enter a unique, descriptive name for your agent
- • Example: "RefundsAgent"
Description (Required)
- • Provide a clear description of the agent's purpose
- • Explain what tasks the agent will handle
- • Example: "Help the user with a refund. If the reason is that it was too expensive, offer the user a refund code. If they insist, then process the refund."
System Prompt (Required)
- • Define the agent's role and behavior
- • Set the context for how the agent should operate
- • Example: "You are a refund specialist"
✨
AI-Assisted Prompt Enhancement
Click the circular button (🔄) next to the system prompt field to:
- • Automatically improve your prompt
- • Get AI-suggested enhancements
- • Make your agent's instructions more precise and effective
- • Optimize the prompt for better performance
Adding Routines
Agents can use multiple routines to perform their tasks:
- • Click on "Select routines" dropdown
- • Choose relevant routines from the list
Example routines:
- • applyDiscount
- • processRefund
Selected routines will appear as icons below
💡
Pro Tip:
Choose routines that work together to achieve the agent's goal. For example, a RefundsAgent might need both discount and refund processing capabilities.
Organizing with Tags
- • Add relevant tags to categorize your agent
- • Click "Add a tag..." field
- • Press enter to add each tag
- • Use tags for better organization and searchability
Connected Swarms
The Connected Swarms panel shows:
- • Which swarms are using this agent
- • Current connection status
- • "No Swarms Yet" if the agent isn't part of any swarms
- • Helps track where the agent is being utilized
Best Practices
ℹ️
Key Points to Remember:
- • Give agents clear, specific names and descriptions
- • Configure system prompts to define agent behavior
- • Add all necessary routines before saving
- • Use tags for better organization
- • Check swarm connections before making changes