MCP server providing semantic CSS patterns optimized for AI code generation - supports Tailwind, Bootstrap, UnoCSS, Tachyons with 77% token savings
classmcp defines 10 tools with complete JSON Schema input specifications and descriptions. Naming follows verb_noun convention well (get_*, list_*, set_*, search_*, generate_, reload_). Descriptions are generally adequate (60-140 chars), exceeding the 10-char minimum. However, several descriptions lack actionable context for LLM selection: get_class and get_component descriptions do not explain when to prefer one over the other; generate_css lacks guidance on typical use cases. Parameters are well-typed (string, boolean, array) with enums for constrained inputs (framework, component, category). Output schemas are not explicitly documented in the source, the tool implementations return structured data, but the response format is not declared in parameter definitions or descriptions. Error handling is minimal: no guidance on recovery paths, no actionable error messages visible in the tool definitions. Descriptions avoid example values (good), but several lack dependency hints (e.g., 'Call list_classes first to discover available names'). Overall, the tools are competent for a CSS utility server, but fall short of production-grade quality due to missing output documentation and weak error recovery guidance.
Generate CSS that defines all semantic classes. Add this to your project's CSS file to use the semantic class names.
Get the full utility classes for a semantic class name. Returns the CSS classes you should use in your HTML. Supports SSR-safe filtering.
Get a complete HTML component example using semantic classes. Great for seeing how classes work together.
Get SSR/hydration safety information for a class pattern. Use this when building SSR/Next.js/Nuxt/Remix applications to avoid hydration mismatches.
List all available semantic class names, optionally filtered by category. Use this to discover available patterns.
List all custom (user-defined) patterns loaded from your config file.
Output schemas not documented. Tool descriptions state what data is returned (e.g., 'Returns the CSS classes'), but do not document the structure, field names, or types of the response object. LLMs cannot plan downstream tool chaining without knowing what fields will be available.
Descriptions lack actionable selection context. 'Get the full utility classes for a semantic class name' is functional but does not explain when to call get_class vs get_component vs list_classes. LLMs waste reasoning cycles deciding between overlapping tools.
No error recovery guidance. Tool descriptions do not include 'If X fails, try Y' hints. Descriptions like 'Reload custom patterns from the config file' do not explain what happens if the config is invalid, malformed, or missing. LLMs have no recovery strategy.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-21 | C | 61 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 21 | - | v1 |
List all available CSS frameworks and their statistics.
Reload custom patterns from the config file (.classmcp.json). Use this after modifying your config file.
Search for classes by name, description, or category. Use when you're not sure of the exact class name.
Set the CSS framework to use for all subsequent operations. Choose based on your project's CSS framework. DEFAULT: tailwind
Weak parameter descriptions for arrays. 'categories' in generate_css is typed as array of strings but lacks enumeration or example values. Should document valid category names (buttons, cards, forms, etc.) directly in the parameter description, not just in list_classes documentation.
Missing dependency hints. get_component description does not mention 'Call list_classes or search_classes first to discover available semantic class names.' Discovery-generation chains are not documented.