Saglitzdesign MCP

What HalidSaglam/saglitzdesign-mcp MCP server does

HalidSaglam/saglitzdesign-mcp MCP server supplies design and marketing knowledge for web, iOS, Android, and macOS projects. Its knowledge base covers platform design languages, components, UX, accessibility, typography, motion, copywriting, SEO, GEO, marketing, security, process, and real-world interface patterns. Documents include metadata, prescriptive guidance, and cited sources.

The server supports both research and production-oriented work. An agent can search for design guidance, retrieve a document by ID, compare platform conventions, or follow a phased roadmap for a website, landing page, SaaS product, or native application. It can also produce code and structured outputs rather than only recommendations.

How it works

HalidSaglam/saglitzdesign-mcp runs locally as an MCP child process and communicates through stdio. It does not expose HTTP or SSE endpoints. The implementation reads bundled and local files, makes no external API calls, and does not require an account or credential. Prompts and project files therefore remain on the local machine when the server is used as described.

A typical workflow begins by calling get_design_roadmap or create_design_system. The agent can then retrieve relevant design-language or component documents, generate tokens and layout foundations, request reference code, and run deterministic checks. Knowledge discovery starts with list_design_knowledge, search_design_knowledge, or a direct document lookup with get_design_doc.

Setup and configuration

Install the package with npm:

npx -y saglitzdesign-mcp

Configure the command as a local stdio MCP server in a compatible client. The supplied material explicitly identifies Claude and Cursor as clients, and the server is intended for MCP clients generally. No environment variables, API keys, remote URL, or additional service configuration are required.

Tools and capabilities

HalidSaglam/saglitzdesign-mcp MCP server includes capabilities in four broad areas:

• Knowledge and workflows: search and browse design documents, retrieve platform references, create project roadmaps, assemble review checklists, compare design languages, inspect knowledge freshness, and consult SEO/GEO guidance.
• Design generation: create complete starter systems; generate colors, type scales, elevation, motion, layouts, and design tokens for CSS, Tailwind, SwiftUI, Jetpack Compose, and W3C DTCG formats.
• Implementation support: request component guidance, production-oriented recipes for supported stacks, icon-library recommendations, and curated examples from real applications and websites. Example recipes cover elements such as buttons, inputs, modals, tabs, cards, toasts, and list rows.
• Auditing and measurement: check WCAG 2.2 contrast and platform target sizes, repair failing color pairs, lint HTML/CSS/JSX/Tailwind, assess UX copy, measure design-system consistency, audit project directories, import named tokens, and measure PNG screenshots. Screenshot and project audits read the paths supplied by the caller and do not make network calls.

The generators are described as deterministic. For example, color generation returns a tonal palette and semantic light/dark tokens, while token generation emits artifacts for selected platforms. The component recipes provide reference code with states and accessibility details, not just design rationale.

Limitations and notes

The package is local and stdio-only, so it cannot be hosted as a remote HTTP or SSE MCP endpoint. Screenshot measurement accepts PNG files, and the example library returns notes and source links rather than bundled third-party screenshot images. The server’s audits are scoped: design lint is a static snippet check, while project auditing examines the directory provided; neither should be treated as a complete runtime or usability evaluation. JavaScript configurations are not evaluated by token import, which reads named token structures instead.

The knowledge base is curated and sourced, but the repository distinguishes documents whose sources are enforced by tests from documents that are not yet covered by that check. Developers should use the returned source information and freshness report when source recency matters.

For teams adopting HalidSaglam/saglitzdesign-mcp MCP server, the main fit is an agent-assisted design and implementation workflow that benefits from local, repeatable guidance and generated artifacts rather than live SaaS data or external integrations.