SaglitzDesign MCP

An expert design & marketing brain for your AI coding agent.

A Model Context Protocol server that gives Claude, Cursor, and any MCP client expert‑level guidance on web, iOS, Android and macOS design — plus the UX, copywriting, SEO, GEO and marketing knowledge that makes a product actually convert.

96 curated knowledge documents · 36 tools · 9 build/review/port workflows · MCP resources with id autocomplete · a one‑call design‑system builder · real token/color/type/elevation/motion/a11y generators · design & UX‑copy linters · production component recipes · phased roadmaps · real‑world visual examples

Why · What's inside · Tools · Resources · Install · Usage · Your own rules · Changelog · License

---

Why

LLMs are confidently wrong about design. They reach for defaults, invent outdated specs, and give generic "make it clean and modern" advice. SaglitzDesign replaces that with a curated, sourced, prescriptive knowledge base your agent can query on demand — the kind of guidance you'd get from a senior product designer, a conversion copywriter, and a technical SEO all at once.

Ask your agent to design a paywall, audit a landing page, or plan an iOS app, and it pulls concrete rules ("44pt minimum touch target", "one primary button per screen", "LCP ≤ 2.5s"), real patterns from top apps, and a phased roadmap — instead of guessing.

• Runtime‑independent. The server reads only local files. No external API, no account, nothing to configure. It just works, offline.
• Local by design. It speaks MCP over stdio only and runs as a child process of your MCP client. There is no HTTP or SSE transport, so it cannot be hosted remotely or reached over a network — which is also why your prompts and code never leave your machine.
• Prescriptive, not vague. Every doc is written as rules an agent applies verbatim — numbers, thresholds, do/don't lists, anti‑patterns.
• Grounded. Design‑language specs from official sources; patterns studied from real top apps and sites; classics distilled from the actual books.

What's inside

96 knowledge documents across 11 categories:

Category	Coverage
🎨 Design languages	Material 3 & M3 Expressive · Apple HIG + Liquid Glass (iOS 26) · deep iOS, Android (Android 16 / M3 Expressive) and macOS app‑design guides · Apple accessibility (Dynamic Type, VoiceOver, hit regions) and Apple shipping readiness (Info.plist purpose strings, entitlements, app icon sets, orientation/multitasking restrictions) · Apple Intelligence design (AI features: Writing Tools, App Intents, on‑device Foundation Models) · visionOS / spatial design (Vision Pro) · Fluent 2 · 2026 web trends · design tokens & theming (W3C DTCG) · Apple WWDC design principles (fluid interfaces)
🧩 Components	Buttons (hierarchy, sizing, states, labels) · forms & inputs · navigation · cards / lists / modals / sheets / empty states
🧠 UX	Nielsen heuristics & behavioral laws · accessibility (WCAG 2.2) · typography · color & dark mode · spacing & grids · motion · mobile UX · conversion / CRO · data visualization · information architecture · i18n / localization (RTL) · AI product UX (chat, streaming, agentic) · onboarding & permission priming
✨ Craft	Expert polish standards · typographic craft · animation craft (easing, springs, interruptibility) · UX writing & cognitive load · 0–40 critique rubric · clean/minimal app design · design‑engineering (semantic HTML, CSS architecture, tokens‑in‑code) · ethical design (avoiding dark patterns) · iconography (choosing & using an icon system) · the AI‑default aesthetic (the stock gradient, font, card chrome and copy generated interfaces reach for, cited to each system's own docs)
📚 Books	Distilled classics — design: Norman, Krug, Refactoring UI, psychology of design, grid/typography, interaction design (Cooper/Tidwell), emotional design (Walter/Norman) · marketing: Cialdini, Positioning, StoryBrand + Ogilvy, Hooked
🗺️ Process	Product‑design & marketing‑website roadmaps · design‑systems methodology (Atomic Design, component API, governance) · design handoff (Figma Dev Mode, Code Connect, design↔dev)
📣 Marketing	Branding & identity · email marketing · HTML email development (Outlook, dark mode, bulletproof) · ad creative · paywall benchmarks (RevenueCat 2026) · growth frameworks (loops/AARRR/PLG) · pricing strategy · analytics & experimentation · value proposition & JTBD · content & distribution (topic clusters, community, referral) · App Store Optimization (ASO)
🔎 SEO	Technical SEO (Core Web Vitals) · on‑page & E‑E‑A‑T · SEO for designers
🤖 GEO	Generative Engine Optimization — visibility in ChatGPT / Perplexity / AI Overviews, llms.txt, citation tactics
🔐 Security	Web security headers & CSP — strict nonce/hash policies, strict-dynamic, Trusted Types, HSTS, SRI, cross-origin isolation, and the superseded headers auditors still ask for. Every claim sourced to a spec or vendor doc, re-verified on a 90-day clock.
🖼️ Patterns & examples	Real‑world patterns studied from top apps & sites (incl. e‑commerce & checkout and fintech / trust flows), plus a curated library of real‑world example screens

Which documents' sources are enforced, and which are not. Of the 96 documents, 11 have their sources: checked by the test suite against a tiered allowlist — the six Apple design‑language guides (apple-hig-liquid-glass, ios-app-design, macos-app-design, apple-accessibility, apple-shipping-readiness, wwdc-design-principles) and the five security documents. The other 85 are curated but not yet checked, and extending the assertion to them is its own piece of work rather than a formality: measured today, 66 of the 85 would fail on 282 sources between them — 227 URL citations across 133 distinct hosts, plus 55 that are not URLs at all and so have no host to check. (The 227 and the 55 are the two halves of the 282, not two figures to add.) Note that apple-intelligence-design and visionos-spatial-design are Apple‑topic documents that sit outside the enforced set. get_design_doc prints which side of this line a document is on, next to its sources, so you never have to come back here to find out.

Workflows (/ prompts) — "build me a…"

Beyond answering questions, SaglitzDesign ships prompts that orchestrate an entire build end‑to‑end. In Claude Code they appear in the / menu under a name that depends on how you installed it. Installed as the Claude Code plugin they are /saglitzdesign:build_landing_page and so on — plugin commands, one generated file per workflow in commands/. With the server installed on its own they are /mcp__saglitzdesign__build_landing_page, which the menu labels saglitzdesign:build_landing_page (MCP). That label is typeable too, (MCP) suffix and all — it is just a clumsier way to reach the same prompt. What does not work is that short form with the suffix dropped: Claude Code hands the bare server:prompt alias to first‑party Anthropic connectors only — the URL has to be https, on api.anthropic.com, under /v1/design/ — and no other server gets one, stdio or remote. (The plugin's /saglitzdesign:build_landing_page above only looks like that alias — it is a plugin command namespace, a different mechanism.)

Invoke one and the agent runs the full method — roadmap → positioning & copy → generates the design system (color, type, layout, elevation, tokens) → real examples → writes the actual code from the component recipes → runs the deterministic verify gate (design_lint, audit_accessibility, audit_design_system, audit_ux_copy) → opens it in a browser, screenshots, scores it against the critique rubric, and iterates until it passes.

Workflow	What it does
build_landing_page	Designs & builds a conversion‑focused landing page, copy‑first, with a visual critique loop.
build_website	Builds a multi‑page marketing site — positioning, IA, SEO/GEO, shared design system.
build_mobile_app_ui	Builds iOS or Android screens on the correct platform baseline (HIG/Liquid Glass or Material 3).
build_dashboard	Builds a dense SaaS dashboard / app shell — navigation, tables, empty states — not a marketing page with KPI cards glued on.
critique_screenshot	Measures the screenshot, then critiques it against the fixed 0–40 rubric — cites real ratios and colour counts, specific elements, no padding.
review_paywall	Scores a paywall / subscription onboarding against real RevenueCat 2026 conversion benchmarks.
design_review	Audits an existing site/app — runs the deterministic auditors first, so findings lead with measured numbers, then the checklists and the 0–40 rubric, ranked by severity.
redesign	Improves an existing UI (bolder / quieter / higher‑converting) using the craft standards, with a measured before→after (consistency score, critique score, lint findings, contrast failures).
port_to_platform	Takes an existing UI to another platform (iOS ↔ Android ↔ macOS ↔ web) surface by surface — porting the intent and IA, never the components.

Just type, e.g., /saglitzdesign:build_landing_page a SaaS invoicing tool for freelancers — everything after the command name becomes the brief, and the workflow asks for anything missing, then builds it. The /mcp__saglitzdesign__… form takes only the first whitespace‑separated word as its brief, so use the plugin command whenever the brief is a phrase. These are yours to type: every command carries disable-model-invocation, so the agent never starts one on its own. Asked in prose it reaches for the skills instead, and they cover much of the same ground — design review, landing‑page conversion, Apple platforms — as guidance rather than as this orchestrated build. The saglitzdesign skill's own trigger vocabulary names paywalls, so prose about one does reach a skill; what no skill carries is review_paywall's RevenueCat benchmark scoring, which runs only when you type it.

The visual critique loop uses whatever browser tool is connected (Claude in Chrome, Playwright, or chrome‑devtools MCP) to see and refine its own output. Without one, it reviews the code directly.

Tools

Tool	What it does
create_design_system	The one‑call foundation. Brand color + vibe + platform → a direction card (named defaults to leave, the type pairing as a decision, one signature move, and a stock‑region fact if the seed sits in indigo/violet/purple) plus WCAG‑verified color (light+dark), matched fonts, an icon library, a modular type scale, an elevation ramp, ready‑to‑paste tokens, the components to build, and a checklist — all generated to work together.
get_design_roadmap	Start here for process. A phased, expert process for a project type (website, landing page, iOS / Android / macOS app, SaaS web app) — each phase has a goal, exit criteria, and the exact docs to read.
search_design_knowledge	Natural‑language search across everything, returning the most relevant section of the best‑matching docs.
get_design_doc	Fetch any document in full by id.
get_component_guidance	Deep dive on a component or screen (button, form, paywall, hero, pricing…) — specs + real‑world patterns.
get_design_language	Full platform / design‑system references (Material 3, Liquid Glass, iOS/Android/macOS, Fluent 2, web trends, tokens).
compare_design_languages	Building on more than one platform? Side‑by‑side iOS / Android / macOS / web conventions for one surface (navigation, buttons, sheets, motion, forms…), plus the porting rules — and an explicit do NOT port list.
get_design_examples	Curated real‑world examples of a pattern from top apps/sites, with notes on what each does well. (Screenshots are a local‑only asset — see Visual examples.)
design_review_checklist	An assembled audit checklist per project type and focus (UI, UX, accessibility, SEO, GEO, conversion, copywriting).
seo_geo_guide	SEO and GEO guides, optionally narrowed to a topic.
import_design_tokens	Already have a design system? Paste your CSS custom properties, shadcn :root block, DTCG file or theme JSON and get back the roles it names, the roles it leaves undefined, a WCAG check on the pairs it defines, and the whole set re‑emitted as CSS / Tailwind / SwiftUI / Compose / DTCG. The inverse of generate_design_tokens — take a web theme to iOS, or audit an inherited one. Only named tokens are read; JS configs are never evaluated.
generate_design_tokens	Real artifacts, not advice — turns a color/spacing/type spec into CSS variables, Tailwind v4, SwiftUI, Jetpack Compose, and W3C DTCG JSON.
audit_accessibility	Deterministic WCAG 2.2 checks — exact contrast ratios for color pairs + tap‑target sizes per platform, with fixes.
get_component_recipe	Production‑ready, accessible reference code for a component (button, input, modal, toast, card, switch, tabs, empty‑state, list‑row, navigation, search, select, table, tooltip, form, pagination, skeleton, badge, breadcrumb) in react‑tailwind, html‑css, SwiftUI, or Compose — all states, ARIA, keyboard, correct motion. Pass your scales — the neutral, primary and danger ramps generate_color_system returns — and the code comes back in your colours, dark‑mode shades included.
generate_color_system	One brand color → a full palette. A 50–950 tonal scale, a cohesive neutral ramp, status colors (danger / success / warning) harmonised with your brand, and light + dark semantic tokens — every text/UI pair WCAG‑verified and auto‑corrected. Feeds straight into generate_design_tokens.
suggest_font_pairing	Curated, production font pairings for a vibe (SaaS, editorial, bold, native…) — heading + body (+ mono) with paste‑ready CSS stacks, weights, rationale, and a type scale.
fix_contrast	Repairs a failing color pair: computes the nearest accessible color (hue/saturation preserved) that meets your WCAG target — the corrected value, not just a fail report.
suggest_icon_library	Recommends the right icon system for a vibe/platform (Lucide, Phosphor, Solar, SF Symbols, Material Symbols…) — with license, install command, coverage, fit rationale, and usage rules.
generate_type_scale	A modular type scale (base × ratio) → named steps with line‑heights, tracking, and fluid clamp(), as CSS variables + Tailwind.
generate_elevation_system	A cohesive layered box‑shadow ramp (flat→modal) as CSS variables + Tailwind, with dark‑mode guidance.
generate_layout_system	Breakpoints (with what changes at each), container widths, a column grid, an intrinsic auto‑fit grid, container queries, and a fluid section‑rhythm scale — CSS variables + Tailwind v4.
generate_motion	Easing + duration tokens and ready‑to‑paste keyframe animations (fade/slide/scale/spring/shimmer) in CSS, Framer Motion, or SwiftUI — reduced‑motion included.
design_lint	Lints an HTML/CSS/JSX/Tailwind snippet for design & a11y anti‑patterns (hardcoded values, killed focus, missing alt/labels, clickable divs, unlabelled inputs…) with line numbers and fixes. Tag‑aware, so formatting never changes the verdict. Returns markdown plus structured output — findings, a severity summary, and a machine‑readable notVisible list of what it could not check.
measure_screenshot	Measures your actual screen. Give it a PNG and it reports the real palette and colour count, true WCAG contrast ratios for the pairs on screen, density, structural detections (alignment, rhythm, off‑grid gaps) each with a confidence level, and a stock‑region fact when a significant cluster sits in Tailwind indigo/violet/purple — plus a self‑contained HTML report you can open and share. Pure‑Node PNG decoding, no network, no dependencies.
audit_project	Audits a real codebase, not a snippet. Point it at a directory: it walks your design source, lints every file, and scores the whole project for consistency — cross‑file drift being exactly what a single‑file lint cannot see. Findings ranked worst‑file‑first with file:line, plus an explicit list of what it did not look at. Returns markdown plus structured output — findings, a severity summary, a machine‑readable notVisible list, and a scan block reporting how many files and bytes were actually read, which were skipped for size or could not be opened, and whether a cap was hit. A missing or non‑directory path comes back as an error result, not an empty audit.
audit_security	Audits a web project or snippet for the defects that actually ship — missing or weak Content‑Security‑Policy, absent HSTS, unpinned cross‑origin scripts, mixed content, credentials in localStorage, secret‑named NEXT_PUBLIC_/VITE_ variables, unsandboxed third‑party iframes, wildcard postMessage, raw‑HTML sinks with no sanitiser, production source maps and un‑ignored .env files. Header state is inferred from wherever your stack declares it — next.config, vercel.json, netlify.toml, _headers, staticwebapp.config.json, firebase.json, Nuxt routeRules, a Remix headers export, SvelteKit hooks.server.ts / kit.csp, Astro middleware and Next middleware or its Next‑16 rename proxy.ts, new Response(…, { headers }) and new Headers({…}) on Workers/Deno/Bun, a quoted header‑name property in any JSON or object literal, any call whose method is set, setHeader, append or header whatever the object is named (res.set / res.setHeader, headers.set / headers.append, Fastify reply.header, Hono c.header, Koa ctx.set), and <meta http-equiv> — read as text and never evaluated; this makes no network request, and says what it could not see. Returns markdown plus structured output — findings, a severity summary, and a machine‑readable notVisible list of what it could not check. A missing or non‑directory path comes back as an error result, not an empty audit.
audit_seo_geo	Audits the SEO and GEO signals that are actually in your source — a missing <title> and one outside the width a result gives it, the same for the meta description, multiple H1s, skipped heading levels, a missing canonical, one written relative in a self‑contained document, one left pointing at localhost or staging, an hreflang set that never lists the page itself, JSON‑LD that does not parse or declares no @context/@type or declares a type whose rich result Google has retired, missing alt text, robots.txt crawl rules (including the AI crawlers behind ChatGPT, Claude, Perplexity and Google's AI surfaces), a robots.txt naming no sitemap, no llms.txt beside it, and content that only exists once a script has run. Absence is claimed only where it can be proven — a self‑contained HTML document, or a whole directory — and a capped scan downgrades every absence claim to an unconfirmed note. It reads source and measures nothing: no request is made, so no finding is a Core Web Vitals result, an indexing status or a ranking outcome. Returns markdown plus structured output — findings, a severity summary, and a machine‑readable notVisible list of what it could not check.
audit_performance	Audits the performance signals that are actually in your source — a hero image held back by loading="lazy" or contradicting its own fetchpriority, the LCP‑candidate image declaring no fetch priority at all, a hero background declared in CSS or an inline style that the HTML preload scanner never sees, images with no width/height or aspect-ratio to reserve their box, a <script src> in the <head> carrying neither defer nor async nor type="module", @font-face without font-display, third‑party font hosts, and scripts loaded from more distinct remote domains than any reading of "minimise" defends. The hero rules are deliberately narrow — the first image inside <main>, with the header logo and the mid‑article diagram structurally excluded — so some pages get no hero finding at all, and that is returned explicitly rather than left to read as a clean result. Core Web Vitals are field data; this loads nothing and times nothing. Returns markdown plus structured output.
audit_apple_ui	Audits an iOS or macOS app against Apple's own documentation. Point it at the Xcode project directory: it reads the four surfaces a project declares configuration on — the information property list, INFOPLIST_KEY_* build settings in project.pbxproj, the entitlements plist, and each colorset's Contents.json — infers whether the target is iOS or macOS from those plus the Swift imports, and runs eight rules: a custom colorset with no luminosity: dark appearance, the deprecated UIRequiresFullScreen key (either spelling), a microphone entitlement declared under one capability and not its twin, no App Sandbox entitlement on macOS (reported as a fact about the Mac App Store channel, not as a defect), plus NavigationView, .font(.system(size:)) on iOS only, a colour written as numbers, and a Button whose whole label is one SF Symbol. Every platform‑scoped rule stays silent when the signals do not settle the question, and the report names the verdict and the signals behind it so a silence reads as the gate rather than as a result. It builds nothing, runs no simulator and takes no screenshot. Returns markdown plus structured output — findings, a severity summary, a scan block, and the longest notVisible list this server ships, every entry of it derived from a run. Directory only; a code argument, a missing path or a path that is a file comes back as an error result, not an empty audit.
audit_android_ui	Audits an Android app against Material 3 and android-app-design. Point it at the project or module directory: it reads AndroidManifest.xml, resource XML and Kotlin, infers whether the folder is an Android project from those plus Compose/Android imports, and runs six rules: the edge-to-edge opt-out attribute (deprecated and ignored on Android 16), the predictive-back opt-out, a day theme XML with no values-night/ file among the surfaces read, a Color(0x…) literal, fontSize = N.sp, and a Material 2 component import. Every configuration rule stays silent when no Android signal was found, and the report names the verdict and the signals behind it so a silence reads as the gate rather than as a result. It builds nothing, starts no emulator and takes no screenshot. Returns markdown plus structured output — findings, a severity summary, a scan block, and a machine-readable notVisible list, every entry of i