What it does

Raven is a design knowledge base MCP server that gives Claude access to curated design intelligence across eight layers: principles (Nielsen's Heuristics, Laws of UX, Gestalt, WCAG), UI patterns (signup flows, pricing, forms, navigation, dashboards), content systems (voice and tone from Mailchimp, GOV.UK, Shopify Polaris), research methods, service design (blueprinting, human handoff patterns), brand and visual guidelines, business strategy (monetization, retention, onboarding), and design tokens from production systems like Stripe and Linear. The server offers 27 tools that evaluate designs, generate design systems, audit HTML and layouts, and extract relevant guidance for a given UI context.

Who it's for

Product designers and engineers integrating Claude into their design workflow, teams building or scaling design systems, product managers optimizing onboarding and conversion, and researchers evaluating user experience decisions against established principles. Most useful when generating UI with Claude and wanting design guardrails, consistency guidance, and decision frameworks baked in.

Common use cases

• Query design principles and UI patterns when generating signup flows, forms, or dashboards
• Evaluate a design description against Nielsen's Heuristics or WCAG accessibility guidelines
• Extract voice and tone guidance from brand systems (Mailchimp, GOV.UK, Shopify Polaris, Atlassian)
• Generate a custom design system from a brand color or token composition
• Audit HTML and CSS against visual-design standards for rhythm, alignment, and contrast
• Get service-design patterns for omnichannel experiences and moments of truth

Setup pitfalls

• Early-stage project (1 GitHub star, first commit 1 day ago) — expect rapid iteration and possible breaking changes
• Requires Node.js and npm; installation via npx -y raven-mcp or from source with npm install && npm run build
• Not officially endorsed by Intuit despite sourcing Intuit design patterns — review the NOTICE file for attribution
• Knowledge base reflects opinionated design choices — use as a starting point, not as absolute guidance