STAGING re-do in progress

AI Agent Editor Guide

AI Agents are powerful automation tools that can create pages, manage blocks, answer visitor questions, and perform complex tasks using artificial intelligence. This guide will help you understand, configure, and use AI agents effectively.

What are AI Agents?

AI Agents are configured AI assistants that can perform specific tasks on your website. Each agent has:

  • Instructions: Detailed guidance on what the agent should do and how
  • Tools: Specific capabilities the agent can use (create blocks, search content, etc.)
  • Model: The AI model powering the agent (GPT-4o, GPT-4o-mini)
  • Temperature: Controls creativity vs. consistency in responses

Common Use Cases:

  • SEO Page Agent: Creates optimized pages with proper block structure
  • Visitor Chatbot: Answers visitor questions and captures leads
  • Content Editor: Updates and improves existing content
  • Block Manager: Reorganizes and optimizes page layouts

Agent Configuration

Every agent requires basic configuration to function properly.

Name and Slug

Name: A human-readable name for the agent

  • Example: "Create New SEO Page"
  • Appears in the agent list and run history
  • Should clearly describe what the agent does

Slug: A unique identifier for the agent

  • Example: "seo-page-agent"
  • Automatically generated from the name
  • Used in code and API calls
  • Must be unique across all agents
  • Cannot be changed after creation (affects integrations)

Instructions

Instructions are the most important part of agent configuration. They tell the agent:

  • What its purpose is
  • How to approach tasks
  • What tools to use and when
  • What to avoid
  • How to format responses

Writing Good Instructions:

  1. Be Specific: Clearly define the agent's role and responsibilities
  2. Provide Context: Explain the business context and goals
  3. Give Examples: Show examples of good outputs
  4. Set Boundaries: Explain what the agent should NOT do
  5. Use Markdown: Format instructions with headers, lists, and code blocks for clarity

AI Model Selection

The AI model determines the agent's capabilities, speed, and cost.

Available Models:

GPT-4o

  • ✅ Most capable and intelligent
  • ✅ Best for complex tasks (SEO page creation, content strategy)
  • ✅ Better at following detailed instructions
  • ✅ More creative and nuanced responses
  • ❌ Slower response time
  • ❌ Higher cost per request

GPT-4o-mini

  • ✅ Fast and efficient
  • ✅ Lower cost per request
  • ✅ Good for simple, repetitive tasks
  • ✅ Ideal for chatbots and quick responses
  • ❌ Less capable with complex reasoning
  • ❌ May struggle with very detailed instructions

Recommendations:

  • Use GPT-4o for: SEO page creation, content editing, complex block management
  • Use GPT-4o-mini for: Visitor chatbot, simple queries, high-volume tasks

Temperature Setting

Temperature controls how creative or consistent the agent's responses are. Range: 0.0 to 2.0

Low Temperature (0.0 - 0.5):

  • More deterministic and focused
  • Consistent, predictable responses
  • Follows instructions very literally
  • Best for: Data extraction, structured tasks, factual responses
  • Example: 0.3 for a chatbot that needs to provide accurate service information

Medium Temperature (0.6 - 1.0):

  • Balanced creativity and consistency
  • Natural, varied responses
  • Good for most use cases
  • Best for: Content creation, general chatbots, page building
  • Example: 0.9 for a visitor chatbot that should sound natural

High Temperature (1.1 - 2.0):

  • Very creative and diverse
  • More unexpected responses
  • Can be less focused
  • Best for: Creative writing, brainstorming, varied content
  • Example: 1.5 for generating unique marketing copy

Recommended Settings:

  • SEO Page Agent: 1.0-1.2 (creative but structured)
  • Visitor Chatbot: 0.7-0.9 (natural but consistent)
  • Data Extraction: 0.2-0.4 (precise and deterministic)

Tools Configuration

Tools give agents specific capabilities. Select only the tools your agent needs.

Why Limit Tools?

  • Prevents agents from performing unintended actions
  • Improves response speed (fewer options to consider)
  • Makes agent behavior more predictable
  • Reduces errors and confusion

Available Tools

The system provides three categories of tools for different purposes.

Block Management Tools

These tools allow agents to manipulate blocks on pages and folders.

Add Block to Blockable:

  • Creates new blocks on pages or folders
  • Specify block type, content, and position
  • Used by: SEO Page Agent, Content Editor

Update Block in Blockable:

  • Modifies existing block content
  • Can update any block field
  • Used by: Content Editor, SEO Page Agent

Remove Block from Blockable:

  • Deletes blocks from pages or folders
  • Permanent action - use with caution
  • Used by: Content Cleanup Agent

Move Block Position:

  • Reorders blocks on a page
  • Specify new position number
  • Used by: Layout Optimizer Agent

Get Blocks in Blockable:

  • Retrieves all blocks from a page or folder
  • Returns block content and metadata
  • Used by: All content agents (to understand current state)

List of Type of Blocks:

  • Returns available block types and their schemas
  • Helps agent understand what blocks it can create
  • Used by: SEO Page Agent, Content Creator

Visitor Chatbot Tools

These tools are specifically designed for the visitor chatbot agent.

Search Published Content:

  • Searches the current brand's published pages
  • Helps visitors find relevant information
  • Returns page titles, URLs, and excerpts

Search All Brands Content:

  • Searches across all brands in the system
  • Useful for multi-brand organizations
  • Returns brand-specific results

Search Brand Services:

  • Finds services offered by the brand
  • Filters by job type (roofing, siding, etc.)
  • Returns service details and descriptions

Search Office Locations:

  • Finds brand office locations
  • Returns addresses, phone numbers, hours
  • Helps visitors find nearby offices

Check Service Area:

  • Determines if brand serves a specific location
  • Uses ZIP code or city/state
  • Returns coverage information

Create Lead:

  • Captures visitor information as a lead
  • Collects name, email, phone, message
  • Integrates with Salesforce
  • Links lead to conversation for tracking

Email Conversation:

  • Sends conversation transcript to visitor
  • Requires visitor's email address
  • Provides record of interaction

General Tools

Get Website Page:

  • Fetches content from any public URL
  • Useful for research and competitive analysis
  • Returns page text content
  • Used by: SEO Page Agent (to research topics)

Running an Agent

Once configured, agents can be executed to perform their tasks.

How to Run an Agent:

  1. Navigate to a page or folder (for content agents)
  2. Click the Run Agent button
  3. Select the agent you want to run
  4. Optionally add additional instructions
  5. Click Execute

What Happens:

  1. An Agent Run is created with status "not_started"
  2. A background job is queued
  3. Status changes to "in_progress"
  4. The agent executes using its tools
  5. Status changes to "completed" or "failed"
  6. Results are stored in the Agent Run

Agent Runs

Agent Runs track each execution of an agent.

Agent Run Information:

  • Agent: Which agent was executed
  • Context: What page/folder it ran on (if applicable)
  • Status: not_started, in_progress, completed, failed
  • Additional Instructions: Any extra guidance provided
  • Responses: The agent's output and tool calls
  • Timestamps: When started and completed

Viewing Agent Runs:

  • Go to Admin → Agent Runs
  • Click on any run to see details
  • View the full conversation and tool usage
  • See what changes the agent made

Best Practices

Writing Effective Instructions:

  • Start with a clear role definition ("You are an SEO content specialist...")
  • Provide specific examples of desired output
  • Include brand voice and tone guidelines
  • List what the agent should NOT do
  • Use structured formatting (headers, lists, code blocks)
  • Test and iterate - refine based on actual results

Choosing the Right Model:

  • Start with GPT-4o-mini for testing (faster, cheaper)
  • Upgrade to GPT-4o if results aren't good enough
  • Use GPT-4o for complex, high-value tasks
  • Use GPT-4o-mini for high-volume, simple tasks

Setting Temperature:

  • Start with 0.7-0.9 for most tasks
  • Lower for factual, consistent responses
  • Raise for creative, varied content
  • Test different values to find the sweet spot

Tool Selection:

  • Only enable tools the agent actually needs
  • More tools = slower responses and more confusion
  • Review agent runs to see which tools are used
  • Remove unused tools to improve performance

Testing Agents:

  • Test on draft pages first, not published content
  • Review agent run results before publishing
  • Start with simple tasks, then add complexity
  • Keep a log of what works and what doesn't

Troubleshooting

Agent produces poor results:

  • Review and refine instructions - be more specific
  • Try a higher-capability model (GPT-4o instead of GPT-4o-mini)
  • Adjust temperature (lower for more consistency)
  • Provide more examples in instructions
  • Check that the agent has the right tools enabled

Agent run fails:

  • Check the error message in the agent run details
  • Verify the agent has necessary tools enabled
  • Ensure the context (page/folder) exists and is accessible
  • Check for API rate limits or service issues

Agent doesn't use tools:

  • Make sure tools are enabled in configuration
  • Instructions should mention when to use tools
  • Provide examples of tool usage in instructions
  • Check that tool parameters are valid

Agent is too slow:

  • Switch to GPT-4o-mini for faster responses
  • Reduce the number of enabled tools
  • Simplify instructions (shorter = faster)
  • Break complex tasks into smaller agent runs

Agent makes unwanted changes:

  • Add explicit boundaries in instructions
  • Remove tools that enable unwanted actions
  • Test on draft content first
  • Review agent runs before publishing
  • Consider lowering temperature for more predictable behavior

Example Agent Configurations

SEO Page Agent:

  • Model: GPT-4o
  • Temperature: 1.0-1.2
  • Tools: All block management tools, Get Website Page
  • Purpose: Create comprehensive, SEO-optimized pages

Visitor Chatbot:

  • Model: GPT-4o-mini
  • Temperature: 0.7-0.9
  • Tools: All visitor chatbot tools
  • Purpose: Answer visitor questions and capture leads

Content Editor:

  • Model: GPT-4o
  • Temperature: 0.5-0.7
  • Tools: Update Block, Get Blocks
  • Purpose: Make precise updates to existing content

Next Steps

Now that you understand AI agents, explore these related features:

  • Block System: Learn about all available block types agents can create
  • Conversations: Understand how the visitor chatbot tracks conversations
  • Agent Runs: Review past agent executions to improve configurations
Chat with us
200 remaining