Creative Agents provide AI-powered creative generation following AdCP PR #23. This integration enables conversational creative development and dynamic optimization.

What are Creative Agents?

Creative Agents are AI-powered tools that extend our creative system with intelligent automation:

Manifest Mode

Static Asset Generation
  • Generate asset manifests conversationally
  • Create optimized images, text, and layouts
  • Follow brand guidelines automatically
  • Output feeds into existing creative/create tools

Code Mode

Dynamic Creative Generation
  • Generate executable creative code
  • Real-time personalization logic
  • Interactive creative elements
  • Deploy to AdCP publisher network

Integration Architecture

Creative Agent Capabilities

Conversational Creative Generation

User: "Build a creative for our coffee brand targeting morning commuters"
Agent Response:
πŸ€– I'll create a morning-focused coffee ad. What format do you prefer?
Options:
- Banner ads for mobile and desktop
- Video creative for social media
- Native ad for content feeds
- Multi-format package

Advanced AI Features

Creative agents automatically follow brand guidelines:
Brand Guidelines Integration:
  Logo Usage: βœ“ Proper placement and sizing
  Color Palette: βœ“ Brand colors (#FF6B6B, #4ECDC4)
  Typography: βœ“ Brand fonts (Montserrat, Open Sans)
  Messaging: βœ“ Tone of voice consistency
  Restrictions: βœ“ No competitor mentions
Example:
"Create a creative for Starbucks following our brand guidelines"
Agent ensures:
  • Green color scheme adherence
  • Approved font usage
  • Logo placement standards
  • Brand voice consistency
AI optimizes creatives based on performance data:
# Performance-Driven Optimization
if creative.ctr < campaign.average_ctr:
    agent.suggest_improvements([
        "Try brighter colors for better attention",
        "Shorten headline for mobile readability", 
        "Test video format for higher engagement"
    ])

if creative.conversion_rate > campaign.top_performers:
    agent.create_variants([
        "Similar style for other products",
        "Seasonal variations of successful design",
        "Cross-format adaptations"
    ])
Code mode enables real-time creative adaptation:
// Dynamic Creative Logic
const personalizeCreative = (user, context) => {
  const creative = baseTemplate.clone();
  
  // Location-based personalization
  if (user.location.weather === 'cold') {
    creative.headline = "Warm Up with Hot Coffee";
    creative.image = warmCoffeeImage;
  }
  
  // Time-based personalization  
  if (context.hour < 10) {
    creative.offer = "Morning Special: 20% Off";
  }
  
  // Behavioral personalization
  if (user.previousPurchases.includes('premium')) {
    creative.product = premiumBlend;
  }
  
  return creative;
};

Integration Workflows

Workflow 1: Manifest Mode Integration

1

Natural Language Request

"Create display creatives for our holiday sale"
2

Creative Agent Processing

πŸ€– Creative Agent (Manifest Mode):
β€’ Analyzing brand guidelines βœ“
β€’ Generating asset manifest...
  - Hero image: Holiday-themed banner (1200x628)
  - Headline: "Holiday Sale - 50% Off"
  - CTA: "Shop Now"  
  - Logo: Brand logo overlay
  - Colors: Brand color palette
3

System Integration

πŸ”„ Integration Process:
1. Manifest β†’ CreateCreativeInput conversion
2. Call existing creative/create tool
3. Assets generated via AdCP publishers
4. Creative ready for campaign assignment
4

Result

βœ… Creative "Holiday Sale 2024" created with 4 AI-generated assets
πŸ“¦ Assets optimized for display advertising
🎯 Ready for campaign assignment

Workflow 2: Code Mode Integration

1

Dynamic Creative Request

"Create a dynamic video creative that personalizes based on user location"
2

Code Generation

πŸ€– Creative Agent (Code Mode):
// Generated executable creative code
const DynamicVideoCreative = {
  template: 'video_personalization_v1',
  
  personalization: {
    location: true,
    weather: true,
    timeOfDay: true
  },
  
  render: (user, context) => {
    const video = baseVideo.clone();
    video.overlayText = `${user.city} Special Offer`;
    
    if (context.weather.temp < 50) {
      video.product = 'winter_blend';
    }
    
    return video;
  }
};
3

Deployment

πŸš€ Deployment Process:
1. Code registered with AdCP publishers
2. Real-time inference endpoints created
3. Creative record created with code reference
4. Dynamic creative live on publisher network
4

Real-Time Operation

🌍 Dynamic Creative Active:
β€’ Personalizes for each viewer location
β€’ Weather-based content switching  
β€’ Real-time performance optimization
β€’ Campaign-ready with dynamic capabilities

Enhanced Creative Tools (Future Integration)

When creative agents are integrated with our MCP orchestration + REST architecture, these tools will be enhanced:

creative/build_creative

Conversational Generation
  • Iterative creative development
  • Natural language refinement
  • Brand guideline integration
  • Multi-format output
"Build a creative with HTML snippet for our coffee brand"
Integration: Works with existing MCP orchestration layer

creative/manage_library

AI-Powered Management
  • Auto-tag creatives by content
  • Performance pattern analysis
  • Optimization recommendations
  • Creative lifecycle management
"Analyze my creative library performance"

Technical Implementation

API Extensions

// Enhanced with creative agent generation (MCP orchestration)
interface CreateCreativeInput {
  creativeName: string;
  
  // Current MCP orchestration approach
  format: {
    type: 'adcp' | 'publisher' | 'creative_agent';
    formatId: string;
  };
  content?: {
    htmlSnippet?: string;
    javascriptTag?: string;
    vastTag?: string;
    assetIds?: string[];  // References to REST-uploaded assets
    productUrl?: string;
  };
  
  // NEW: Creative agent generation via MCP
  generateWithAgent?: {
    mode: 'manifest' | 'code';
    prompt: string;
    brandGuidelines?: BrandGuidelines;
    targetFormats?: CreativeFormat[];
    optimizationGoals?: OptimizationGoal[];
  };
  
  advertiserDomains: string[];
  assemblyMethod: 'publisher' | 'creative_agent' | 'pre_assembled';
}

Usage Examples

Example 1: E-commerce Creative Generation

User: "Create product showcase creatives for our electronics store"

Agent: "I'll create a product showcase creative package. What format do you need?"

User: "HTML snippets for display banners targeting tech enthusiasts"

Agent: "Perfect! Generating tech-focused HTML creatives..."

Result:
🎨 Creative Package: "Tech Showcase Q4"
πŸ“± HTML Creative with smartphone hero imagery
🎧 HTML Creative with headphones product focus
πŸ“Š Format: publisher/display_banner with pre_assembled method
🎯 Content: HTML snippets ready for ad server deployment

Example 2: Seasonal Campaign Automation

User: "Set up automated seasonal creatives for our coffee shop"

Agent: "I'll create dynamic seasonal creatives. What format and assembly method?"

User: "Creative agent format with dynamic code for weather personalization"

Agent: "Deploying seasonal automation with weather triggers..."

Result:
β˜€οΈ Summer: Dynamic HTML with iced coffee focus, bright colors
πŸ‚ Fall: Dynamic content with warm drinks, cozy imagery
❄️ Winter: Dynamic messaging for hot beverages, comfort themes
🌸 Spring: Dynamic creative with fresh start, renewal messaging
🌀️ Format: creative_agent/dynamic_seasonal with real-time personalization

Example 3: A/B Testing Automation

User: "Generate A/B test variants for our best-performing creative"

Agent: "I'll analyze your top creative and generate variants. Which elements to test?"

User: "Headlines, colors, and call-to-action buttons"

Agent: "Creating systematic A/B test variants..."

Result:
🎨 Original Creative (Control)
πŸ“ Variant A: Different headline approach
🎨 Variant B: Alternative color scheme  
πŸ”² Variant C: New CTA button style
πŸ“Š A/B test tracking configured
⚑ Automatic winner selection enabled

Benefits for Users

Advertisers

Faster Creative Development
  • AI generates assets in minutes
  • Brand consistency automatically maintained
  • Performance optimization built-in
  • Scale creative production efficiently
Cost Reduction
  • Less manual creative production
  • Automated variant generation
  • Performance-driven improvements
  • Reduced agency dependency

Campaign Managers

Natural Language Interface
  • β€œCreate banner ads for spring sale”
  • Conversational creative refinement
  • Automatic campaign assignment
  • Performance insights and recommendations
A/B Testing Automation
  • Automatic variant generation
  • Performance comparison tracking
  • Winner selection and scaling
  • Continuous optimization loops

Developers

Seamless Integration
  • Works with existing creative tools
  • Backward compatibility maintained
  • AdCP publisher compatibility
  • Extensible architecture
Dynamic Capabilities
  • Real-time personalization code
  • Interactive creative elements
  • Performance monitoring built-in
  • Multi-publisher deployment

Migration Timeline

1

Phase 1: Foundation (Available Now)

  • Manual creative management fully functional
  • AdCP-aligned Creative/Asset hierarchy
  • Human-readable APIs ready for AI integration
  • Pass-through architecture prepared
2

Phase 2: Manifest Mode (Coming Soon)

  • AI-generated static assets
  • Brand guideline integration
  • Conversational creative development
  • Enhanced creative/create tool
3

Phase 3: Code Mode (Future)

  • Dynamic creative generation
  • Real-time personalization
  • Interactive creative elements
  • Full AdCP publisher integration
4

Phase 4: Advanced AI (Future)

  • Automated A/B testing
  • Performance optimization
  • Creative library management
  • Predictive creative recommendations
Current Status: Creative agents integration is in development. All existing creative management tools work fully without creative agents.
Ready to use creative management today? Start with the main quickstart guide to create and manage creatives using our current tools.