# BrandSweets

> The brand memory and iteration layer for AI agents. Extract a brand from a domain or shape one on a canvas, then serve it as structured, agent-readable data — brand.md, tokens.json, brand.css, and a per-brand MCP endpoint every future session already knows.

BrandSweets turns a brand into a structured, machine-readable system any AI agent can READ to stay on-brand, and (with human approval) WRITE to. Every brand serializes losslessly to the formats agents actually consume — brand.md, tokens.json, brand.css — plus a per-brand Model Context Protocol (MCP) endpoint. No bespoke SDK.

## What the structured brand data is

Each brand is a validated BrandDoc covering identity (name, story, mission, voice, dos & don'ts, sample copy), design tokens (role-based color, typography, spacing, radius, gradients — W3C DTCG compatible), a logo/asset inventory, and usage rules. It serializes to:

- [brand.md](https://brandsweets.com/b/{slug}/brand.md): narrative brief an agent grounds on before it writes (text/markdown)
- [tokens.json](https://brandsweets.com/b/{slug}/tokens.json): role-based design tokens, W3C Design Tokens format (application/json)
- [brand.css](https://brandsweets.com/b/{slug}/brand.css): CSS custom properties — one import and a prototype is on-brand (text/css)

## Connect an agent over MCP

Two endpoints, both Streamable HTTP:

- Public read (one brand, no auth): `https://mcp.brandsweets.com/{slug}/mcp`
- Workspace loop (read + write, all brands): `https://mcp.brandsweets.com/w/{workspace}/mcp` — send a workspace API key as `Authorization: Bearer <token>`

Claude Code, one line:

    claude mcp add --transport http brandsweets https://mcp.brandsweets.com/{slug}/mcp

## Install the skills plugin

    claude plugin marketplace add ethanplusai/brandsweets
    claude plugin install brandsweets-skills

The bundle wires the workspace MCP server and the BrandSweets design skills (brand-design, logo-design, on-brand, imagery, design-taste).

## Fonts — an open, agent-queryable font library

~1,900 free, open-licensed (OFL/Apache/UFL) typefaces, self-hosted with WOFF2 + license files, with hand-vetted pairings and mood tags.

- [Font library](https://brandsweets.com/fonts): browse, filter, preview, and self-host (HTML)
- Per font: [/fonts/{slug}](https://brandsweets.com/fonts/{slug}) with a specimen, pairings, embed CSS, and a machine-readable [/fonts/{slug}/font.md](https://brandsweets.com/fonts/{slug}/font.md)
- Collections: [/fonts/serif](https://brandsweets.com/fonts/serif), [/fonts/mood/editorial](https://brandsweets.com/fonts/mood/editorial), [/fonts/pairs-with/{slug}](https://brandsweets.com/fonts/pairs-with/inter)
- Free alternatives to commercial fonts: [/fonts/alternatives](https://brandsweets.com/fonts/alternatives) hub; per font [/fonts/alternatives/{commercial-slug}](https://brandsweets.com/fonts/alternatives/helvetica) with a machine-readable [alternatives.md](https://brandsweets.com/fonts/alternatives/helvetica/alternatives.md); the published, deterministic scoring formula at [/fonts/alternatives/methodology](https://brandsweets.com/fonts/alternatives/methodology)

Connect an agent to the public fonts MCP (read-only, no auth) to search the catalog and get pairings:

    claude mcp add --transport http brandsweets-fonts https://mcp.brandsweets.com/fonts/mcp

Tools: `list_fonts` (filter by category/mood/superfamily/weight/language/license), `suggest_fonts` (ranked pairings with rationale), `find_alternatives` (free alternatives to a commercial font with per-axis scores), `list_commercial_fonts`.

## Colors — palette tools for agents and humans

A deterministic color engine (no AI): analysis, WCAG contrast with fix suggestions, patterned palette generation with locks, ~1,700 named colors (incl. real brand colors), and ~60 curated palettes with rationale.

- [Color tools](https://brandsweets.com/colors): hub — any color → analysis page (HTML)
- [Palette generator](https://brandsweets.com/palettes/new): Coolors-style — space to generate, lock what you like; the URL IS the palette state
- [Contrast checker](https://brandsweets.com/contrast) · [Curated palettes](https://brandsweets.com/palettes) · per-brand colors at /colors/brands/{slug}
- Machine-readable: /colors/{name-or-hex}/color.md, /palettes/{mood}/palette.md

Connect an agent to the public colors MCP (read-only, no auth):

    claude mcp add --transport http brandsweets-colors https://mcp.brandsweets.com/colors/mcp

Tools: `analyze_color`, `check_contrast` (with fixes), `generate_palette` (N candidates with rationale + a shareable URL each), `suggest_palette` (curated-first by mood/industry), `list_palettes`.

The agent ↔ human loop: `generate_palette` returns URLs like `/palettes/new/3b82f6-1e293b-f59e0b-fef3c7-0f172a?lock=0,2`. Hand one to a human; they lock/tweak in the tool and send the link back; the `?lock=` indices are their decisions — regenerate around them.

## Gradients — perceptual gradients with a contrast report

A deterministic gradient engine (no AI): interpolation in OKLab/OKLCH (no grey midpoint), eased stops, anti-banding grain, mesh/conic/duotone/glass, and a REPORT on every gradient — worst-case WCAG contrast for white and black text along the line, banding risk, gamut warnings. Exports CSS (`in oklch`), Tailwind v4, SVG, PNG, DTCG.

- [Gradients hub](https://brandsweets.com/gradients): by colour (/gradients/blue), by mood (/gradients/mood/sunset), colour-to-colour (/gradients/gold-to-cornflowerblue), techniques (/gradients/mesh, /grainy, /conic, /tailwind), brand gradients (/gradients/brands/{slug})
- [Gradient maker](https://brandsweets.com/gradients/new): sRGB | OKLab | OKLCH side by side, easing editor, live contrast strip with a one-click fix; the URL IS the gradient (`/gradients/new/v1.<hex>-<hex>?in=oklch&a=135`)
- Machine-readable: /gradients/{family|slug|a-to-b}/gradients.md, /gradients/mood/{mood}/gradients.md, /gradients/brands/{slug}/gradients.md
- Bundled corpus: uiGradients + WebGradients (MIT, notices at https://brandsweets.com/gradients/licenses), ~550 named gradients tagged by measured OKLCH hue family and mood

Connect an agent to the public gradients MCP (read-only, no auth):

    claude mcp add --transport http brandsweets-gradients https://mcp.brandsweets.com/gradients/mcp

Tools: `make_gradient` (2–5 colours → every export + the report in one call), `brand_gradients` (brand slug or palette → brand gradient, text-safe hero, button with states, variants). Resource: `gradients://corpus` (the whole corpus as JSON). Hand a `/gradients/new/…` URL to a human; the link they send back is the gradient they chose.

## Key URLs

- [Home](https://brandsweets.com/): what BrandSweets is
- [Font library](https://brandsweets.com/fonts): ~1,900 open-licensed fonts + pairings (HTML)
- [Color tools](https://brandsweets.com/colors): palette generator, contrast checker, named + brand colors (HTML)
- [Gradients](https://brandsweets.com/gradients): OKLCH gradient maker, corpus by colour/mood, pair pages, brand gradients (HTML)
- [Brand directory](https://brandsweets.com/b): the public library of brands
- [Agent setup guide](https://brandsweets.com/docs/agents): connect MCP + install the plugin (HTML)
- [Agent guide (markdown)](https://brandsweets.com/agents.md): the same, as clean markdown for agents
- [MCP discovery manifest](https://brandsweets.com/.well-known/mcp.json): machine-discoverable endpoints + capabilities
- [Full dump](https://brandsweets.com/llms-full.txt): everything in one fetch, incl. the full agent guide and every brand's brand.md URL
- [Pricing](https://brandsweets.com/pricing): free for one brand; Pro for unlimited + unthrottled MCP
- [Sitemap](https://brandsweets.com/sitemap.xml)

## Brand corpus

- [duolingo](https://brandsweets.com/b/duolingo): brand page — machine surfaces at /b/duolingo/brand.md, /b/duolingo/tokens.json, /b/duolingo/brand.css
- [figma](https://brandsweets.com/b/figma): brand page — machine surfaces at /b/figma/brand.md, /b/figma/tokens.json, /b/figma/brand.css
- [honda-com](https://brandsweets.com/b/honda-com): brand page — machine surfaces at /b/honda-com/brand.md, /b/honda-com/tokens.json, /b/honda-com/brand.css
- [fly-io](https://brandsweets.com/b/fly-io): brand page — machine surfaces at /b/fly-io/brand.md, /b/fly-io/tokens.json, /b/fly-io/brand.css
- [qualcomm-com](https://brandsweets.com/b/qualcomm-com): brand page — machine surfaces at /b/qualcomm-com/brand.md, /b/qualcomm-com/tokens.json, /b/qualcomm-com/brand.css
- [render-com](https://brandsweets.com/b/render-com): brand page — machine surfaces at /b/render-com/brand.md, /b/render-com/tokens.json, /b/render-com/brand.css
- [framer](https://brandsweets.com/b/framer): brand page — machine surfaces at /b/framer/brand.md, /b/framer/tokens.json, /b/framer/brand.css
- [github](https://brandsweets.com/b/github): brand page — machine surfaces at /b/github/brand.md, /b/github/tokens.json, /b/github/brand.css

Full index (406 public brands): https://brandsweets.com/b and https://brandsweets.com/llms-full.txt