The AI brand stack has three layers: Design Tokens (technical values for product design ), Brand.md (semantic layer with brand identity and context), and MCP (delivery mechanism for real-time queries). Brand.md is the core source of truth. Design tokens are optional for brand-only use cases. MCP makes brand data queryable by AI tools in real-time.

How Do You Build the AI Brand Stack: From Tokens to Architecture?

Machine-readable brand systems are not a single format. They are a stack of three complementary layers, each serving a different purpose. And each layer has internal architecture that determines whether your system scales or breaks.

Understanding the stack matters because each layer answers a different question. Understanding the architecture matters because it determines whether your brand system works at scale or fails under pressure.

What Are the Three Layers of the AI Brand Stack?

Layer 1: Design Tokens — Technical values for product design (W3C-DTCG compliant).

Layer 2: Brand.md — Brand identity + semantic layer for AI execution.

Layer 3: MCP — Protocol that makes brand.md queryable by AI tools in real-time.

What Are Design Tokens and Why Do They Matter?

Design tokens are key-value pairs that define design decisions: colors, typography, spacing, shadows, animations.

{
  "color": {
    "primary": {
      "$value": "#FF6B35",
      "$type": "color",
      "$description": "Primary action color"
    }
  },
  "typography": {
    "heading": {
      "$value": {
        "fontFamily": "Inter",
        "fontSize": "32px",
        "fontWeight": "bold"
      },
      "$type": "typography"
    }
  }
}

{
  "color": {
    "primary": {
      "$value": "#FF6B35",
      "$type": "color",
      "$description": "Primary action color"
    }
  },
  "typography": {
    "heading": {
      "$value": {
        "fontFamily": "Inter",
        "fontSize": "32px",
        "fontWeight": "bold"
      },
      "$type": "typography"
    }
  }
}

{
  "color": {
    "primary": {
      "$value": "#FF6B35",
      "$type": "color",
      "$description": "Primary action color"
    }
  },
  "typography": {
    "heading": {
      "$value": {
        "fontFamily": "Inter",
        "fontSize": "32px",
        "fontWeight": "bold"
      },
      "$type": "typography"
    }
  }
}

What design tokens do:

1. Pass technical values between design and development tools

2. Maintain consistency across product design systems

3. Integrate with Figma, code repositories, and design systems platforms

What design tokens don't do:

1. Store semantic meaning or context

2. Define when and where to use each element

3. Provide AI with enough information to make intelligent decisions

4. Scale beyond product design to brand-wide consistency

Design tokens are necessary. They are not sufficient for AI-native brand systems.

What Is Brand.md and What Does It Include?

Brand.md is the core layer. It captures brand identity, strategic positioning, voice, and the semantic layer that allows AI to execute on brand guidelines.

Learn what a semantic layer is and why it matters ->

Brand.md has three internal architectural layers:

Identity Architecture (Static, Source of Truth)

The identity layer establishes the brand's core attributes that everything else flows from.

{
  "brand": {
    "name": "Sameness",
    "mission": "Make AI understand brand identity",
    "positioning": "The semantic layer for AI-native brands",
    "personality": {
      "traits": ["forward-thinking", "technical", "direct"],
      "voice": "expert but accessible",
      "tone_spectrum": {
        "formal_to_casual": 0.6,
        "technical_to_simple": 0.4,
        "optimistic_to_realistic": 0.7
      }
    },
    "values": [
      "Clarity over complexity",
      "Precision over approximation",
      "Future-ready over legacy-bound"
    ]
  }
}

{
  "brand": {
    "name": "Sameness",
    "mission": "Make AI understand brand identity",
    "positioning": "The semantic layer for AI-native brands",
    "personality": {
      "traits": ["forward-thinking", "technical", "direct"],
      "voice": "expert but accessible",
      "tone_spectrum": {
        "formal_to_casual": 0.6,
        "technical_to_simple": 0.4,
        "optimistic_to_realistic": 0.7
      }
    },
    "values": [
      "Clarity over complexity",
      "Precision over approximation",
      "Future-ready over legacy-bound"
    ]
  }
}

{
  "brand": {
    "name": "Sameness",
    "mission": "Make AI understand brand identity",
    "positioning": "The semantic layer for AI-native brands",
    "personality": {
      "traits": ["forward-thinking", "technical", "direct"],
      "voice": "expert but accessible",
      "tone_spectrum": {
        "formal_to_casual": 0.6,
        "technical_to_simple": 0.4,
        "optimistic_to_realistic": 0.7
      }
    },
    "values": [
      "Clarity over complexity",
      "Precision over approximation",
      "Future-ready over legacy-bound"
    ]
  }
}

This layer is relatively static. It changes rarely. It is the source of truth for everything downstream.

Execution Architecture (Dynamic, Queryable, Semantic)

The execution layer translates identity into executable rules. It is where semantic data lives.

{
  "colors": {
    "primary": {
      "name": "Deep Blue",
      "hex": "#0052CC",
      "semantic_associations": ["trust", "stability", "professionalism"],
      "usage": {
        "primary": "CTAs, headers, primary actions",
        "secondary": "backgrounds, accents",
        "avoid": "body text, disabled states"
      },
      "accessibility": {
        "wcag_aa": true,
        "wcag_aaa": false,
        "contrast_on_white": 8.5
      }
    }
  },
  "voice": {
    "rules": {
      "do": ["Use contractions", "Use active voice", "Use second person"],
      "avoid": ["Passive voice", "Jargon", "Formal register"]
    },
    "scored_examples": [
      {
        "text": "You can save hours every week.",
        "score": 1.0,
        "reason": "Active voice, second person, contractions"
      }
    ]
  }
}

{
  "colors": {
    "primary": {
      "name": "Deep Blue",
      "hex": "#0052CC",
      "semantic_associations": ["trust", "stability", "professionalism"],
      "usage": {
        "primary": "CTAs, headers, primary actions",
        "secondary": "backgrounds, accents",
        "avoid": "body text, disabled states"
      },
      "accessibility": {
        "wcag_aa": true,
        "wcag_aaa": false,
        "contrast_on_white": 8.5
      }
    }
  },
  "voice": {
    "rules": {
      "do": ["Use contractions", "Use active voice", "Use second person"],
      "avoid": ["Passive voice", "Jargon", "Formal register"]
    },
    "scored_examples": [
      {
        "text": "You can save hours every week.",
        "score": 1.0,
        "reason": "Active voice, second person, contractions"
      }
    ]
  }
}

{
  "colors": {
    "primary": {
      "name": "Deep Blue",
      "hex": "#0052CC",
      "semantic_associations": ["trust", "stability", "professionalism"],
      "usage": {
        "primary": "CTAs, headers, primary actions",
        "secondary": "backgrounds, accents",
        "avoid": "body text, disabled states"
      },
      "accessibility": {
        "wcag_aa": true,
        "wcag_aaa": false,
        "contrast_on_white": 8.5
      }
    }
  },
  "voice": {
    "rules": {
      "do": ["Use contractions", "Use active voice", "Use second person"],
      "avoid": ["Passive voice", "Jargon", "Formal register"]
    },
    "scored_examples": [
      {
        "text": "You can save hours every week.",
        "score": 1.0,
        "reason": "Active voice, second person, contractions"
      }
    ]
  }
}

This layer is dynamic. It changes as the brand evolves. It is queryable by AI tools.

Query Architecture (Interface Layer)

The query layer is how AI tools access the execution layer. It defines the interface and constraints.

When an AI tool needs to generate an image, it queries:

The system responds with the semantic data the AI needs to execute correctly.

What Is MCP and How Does It Work?

MCP (Model Context Protocol) is the delivery mechanism. It takes brand.md and makes it queryable by AI tools in real-time.

When an AI tool needs to know "what color should I use?" or "how should I write this?", it queries the MCP endpoint. Brand.md responds with semantic data, not just hex codes or adjectives.

MCP enables:

1. Real-time queries from AI tools

2. Consistent brand execution across hundreds of AI calls per day

3. Scalability without manual intervention

4. Version control and brand evolution

How Do These Three Brand Layers Stack Together?

Brand.md is the core. It contains everything: technical values (colors, fonts, spacing), semantic context (mood, personality, usage rules), and behavioral rules (tone, voice, negative constraints).

Design tokens are optional. If you are also managing product design (web apps, component libraries), you can extract design tokens from brand.md for your design system. But if you are only executing brand (packaging, marketing, brand assets), you don't need a separate token layer. Brand.md is sufficient.

MCP is the delivery mechanism. It makes brand.md queryable by AI tools at scale.

For brand-only execution:

For product + brand execution:

In both cases, brand.md is the source of truth. Design tokens are optional. MCP is the delivery mechanism.

Why Does This Architecture Matter?

A poorly structured semantic layer has these problems:

1. Redundancy: The same data is stored in multiple places, causing inconsistency

2. Ambiguity: The same concept is named differently in different sections

3. Brittleness: Changing one element breaks relationships elsewhere

4. Scalability: Adding new brand elements requires restructuring the entire system

A well-structured semantic layer has these properties:

1. Single source of truth: Each data point exists once, referenced everywhere

2. Clear relationships: Elements are explicitly connected, not implicitly assumed

3. Composability: New elements can be added without restructuring

4. Scalability: The system grows without breaking

The difference between a brand system that works and one that fails is architecture.

AI-native brand systems require all three layers working together: design tokens for product design, brand.md for brand identity and semantic context, and MCP for AI delivery at scale. The internal architecture of brand.md determines whether your system scales or breaks under pressure.

How AI-native teams put this stack into practice ->