I Built an API That Gives LLMs Brand Context — Here's How It Works

The Problem

Design systems were built for humans reading documentation. Figma files, PDFs, Notion pages — all designed for visual consumption.

LLMs can't read your Figma file. They can't reliably parse your 47-page brand PDF. And even if they could, you'd need to paste it into every single prompt.

The result: every AI-generated asset — email copy, landing page code, component markup — comes out looking like it was built by someone who's never seen your brand.

The Solution: A Machine-Readable Brand System

BrandAPI lets you register your design tokens once and serves them as a structured JSON endpoint. Any tool that can make an HTTP request gets your complete brand context.

curl https://brandapi-ogk6.polsia.app/v1/brand/your-brand \
  -H "Authorization: Bearer YOUR_API_KEY"

Response:

{
  "brand": "Acme Corp",
  "tokens": {
    "colors": {
      "primary": { "value": "#7c3aed", "usage": "CTAs, links, active states" },
      "background": { "value": "#0a0a0c", "usage": "Primary surface" }
    },
    "typography": {
      "heading": { "family": "Space Grotesk", "weight": "700" },
      "body": { "family": "Inter", "size": "1rem" }
    },
    "voice": {
      "tone": "Direct, confident, no buzzwords",
      "avoid": ["passive voice", "exclamation marks", "marketing fluff"]
    }
  }
}

Integration: 3 Lines

// Fetch brand context
const brand = await fetch('https://brandapi-ogk6.polsia.app/v1/brand/context', {
  headers: { 'Authorization': `Bearer YOUR_API_KEY` }
}).then(r => r.json());

// Inject into any AI prompt
const prompt = `Use these brand guidelines:\n${JSON.stringify(brand.tokens, null, 2)}`;

// Every output is on-brand. Done.

Works with any LLM — OpenAI, Anthropic, local models. Works with any AI tool that lets you inject system prompts — Cursor, v0, Bolt, custom agents.

What You Can Store

• Colors — primary, secondary, accent, backgrounds, with usage context
• Typography — font families, weights, sizes, line heights
• Spacing — scale values with where to use them
• Voice & Tone — how the brand sounds, what to avoid
• Logos — URLs to assets in different formats
• Custom tokens — anything else your brand needs

Each brand also gets a public style guide page — a live, interactive reference that stays in sync with your API. No more outdated brand PDFs.

The Architecture

Simple Express.js API backed by PostgreSQL. Each brand is a set of token groups, each token group contains individual tokens with values and metadata.

The API is designed around one principle: the response should be directly injectable into an LLM prompt without any transformation. No nested structures that need flattening, no references that need resolving. Fetch and inject.

Why This Beats Pasting Guidelines

Copy-pasting brand guidelines into prompts breaks down at scale in three ways:

• Stale — your pasted guidelines drift as the real brand evolves
• Inconsistent — every developer pastes a different version or excerpt
• Fragile — long pasted text hits context limits and gets truncated

A brand API endpoint gives you a single live source. Update your primary color in BrandAPI — every prompt, every AI tool, every integration picks it up on the next call. Zero coordination overhead.

Try It

Free tier: 1 brand, 10,000 API calls/month, no credit card required.

You can see a live brand guide (BrandAPI's own brand, eating our own dogfood): Live demo →

Or just sign up and have your brand API running in 2 minutes: brandapi-ogk6.polsia.app →