AI & Developer Experience
Urbicon UI is built for AI-assisted development. Every component is discoverable, documented, and scaffold-ready for Claude, Cursor, Copilot, and any MCP-compatible tool.
MCP Server
The Model Context Protocol server exposes 10 read-only tools, 10 design prompts and 7 guide
resources. Connect it to your IDE or AI assistant for real-time component discovery, code
generation, and a closed design loop: generate UI, lint it with validate_design, judge it against a scoring rubric, and keep design intent across sessions in a design.manifest.md, maintained locally by the urbicon CLI.
Setup
Available Tools
10 Tools
Discover: find_components, get_component, find_icons · Generate: get_recipe, suggest_implementation, get_implementation_checklist, get_css_reference · Design loop: get_design_principles, get_pattern, validate_design
10 Prompts · 7 Guide Resources
Prompts: the design-verb table — onboard, adopt, compose, redesign, polish, critique, fix, retheme, audit, migrate (each a recipe over the design loop) · Guides: API grammar, component families, design tokens, design quality, customization, style presets, auth setup
llms.txt
Every component has its own llm.txt file with structured documentation. The MCP server can query individual sections to minimize token
usage.
LLM Documentation
.cursorrules
IDE-specific guidance for AI code generation. Ensures generated code follows project conventions: barrel imports, semantic tokens, Svelte 5 runes, focus-visible, and more.
AI Coding Conventions
How It Works
JSDoc annotations in component source files are the single source of truth. One edit automatically updates the documentation site, llms.txt files, and MCP catalog.