Google created DESIGN.md.
We generate it for any brand on the internet — automatically.
Visual brand intelligence for AI agents. Any URL. Any agent. One API call.
# Give your agent eyes for brand design
from visualmcp import client
brand = get_brand_context("apple.com")
brand
# →
{
"url": "https://apple.com",
"archetype": "The Creator",
"tone": "minimal · confident · premium",
"colors": {
"bg": "#ffffff", "fg": "#1d1d1f",
"accent": "#0071e3"
},
"typography": {
"display": "SF Pro Display",
"body": "SF Pro Text",
"scale": "1.25"
},
"radius": "12px", "density": "airy",
"confidence": 0.97
}// The problem
Your agent has no idea what your brand looks like.
Agents are design-blind
LLMs see text, not brand. Without visual context they default to generic, off-brand output.
Every call starts from zero
Re-explaining a brand to your agent on every run is wasted tokens, drift, and inconsistency.
Screenshots are slow and expensive
Headless browsers plus vision models cost seconds and cents per request. They don't scale.
// How it works
From URL to brand intelligence in one request.
Point to any URL
Pass a domain or page. No auth into the site, no scraping setup, no headless browser to maintain.
VisualMCP extracts & analyzes
We parse the rendered page — colors, type, spacing, density, archetype — into a structured model.
Your agent gets brand intelligence
Receive JSON or a valid DESIGN.md the agent can read on every call. Consistent, cacheable, cheap.
// Live demo
See it work. Right now.
Enter any URL. Get a DESIGN.md file your AI agents can read. Free, no account required.
// Output formats
Two formats. Same source of truth.
JSON output
For developers{
"brand": "stripe.com",
"archetype": "The Sage",
"tone": ["technical", "calm", "precise"],
"palette": {
"bg": "#ffffff",
"fg": "#0a2540",
"accent": "#635bff",
"muted": "#425466"
},
"type": {
"display": "Sohne",
"mono": "Sohne Mono",
"scale": 1.2
},
"radius": "8px",
"density": "compact",
"wcag": { "AA": "pass", "AAA": "partial" }
}DESIGN.md output
For DESIGN.md workflows---
brand: stripe.com
archetype: The Sage
tone: [technical, calm, precise]
palette:
bg: "#ffffff"
fg: "#0a2540"
accent: "#635bff"
type:
display: Sohne
scale: 1.2
---
# Brand voice
Stripe communicates with technical precision. Avoid
marketing fluff. Prefer concrete nouns and code examples
over adjectives. Lead with the API, not the promise.
## Layout principles
- Generous whitespace, dense within containers
- Sharp 8px radius, no glassmorphism
- Accent #635bff reserved for primary actions only// Use cases
Built for teams that ship with agents.
AI content agents
Generate on-brand copy, posts, and assets without re-prompting the brand on every run.
Marketing agency workflows
Onboard a client by URL. Ship campaigns that match their visual system from day one.
Competitive brand analysis
Diff palettes, type, and tone across competitors. Track brand evolution over time.
Design governance at scale
Enforce brand tokens across teams, agents, and pipelines. Audit drift with one call.
// Pricing
Pay for extractions. Not seats.
Starter
For solo devs and small agent projects.
- 1,000 extractions / mo
- JSON + DESIGN.md output
- MCP server + REST API
- Community support
Pro
For teams shipping agent products.
- 5,000 extractions / mo
- Brand comparison & diffs
- WCAG validation
- Priority support
Agency
For agencies and platforms.
- 20,000 extractions / mo
- Multi-client workspaces
- White-label reports
- Dedicated Slack channel