Skip to main content
Glama
README.md
# πŸš€ VibeCraft MCP (`vibecraft-mcp`)

> **Deterministic UI/UX Intelligence & Systems Performance Engine for AI Vibe-Coding.**

[![GitHub Repo](https://img.shields.io/badge/GitHub-originsatyam%2Fvibecraft--mcp-indigo.svg)](https://github.com/originsatyam/vibecraft-mcp)
[![License: MIT](https://img.shields.io/badge/License-MIT-emerald.svg)](LICENSE)
[![Build Status](https://img.shields.io/badge/build-passing-brightgreen.svg)]()

**VibeCraft MCP** is a 100% free, open-source Model Context Protocol (MCP) server that embeds real-world **UI Engineering**, **UX Product Psychology**, and **Backend Performance Auditing** directly into your AI coding agent sessions (Google Antigravity, Cursor, Windsurf, Claude Code).

It permanently eliminates **AI Slop**β€”generic, un-styled, un-accessible, N+1 query-heavy codeβ€”by enforcing proven design principles and component engineering automatically.

---

## ⚑ 1-Click Quickstart Setup (100% Free)

Add `vibecraft-mcp` to your IDE's `mcp_config.json` (Cursor, Antigravity, Windsurf, Claude Code).

### Method 1: Direct from GitHub (Instant & 100% Free)

```json
{
  "mcpServers": {
    "vibecraft-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "github:originsatyam/vibecraft-mcp"
      ],
      "env": {
        "UI_UX_MCP_API_KEY": "your_api_key_here"
      }
    }
  }
}
```

### Method 2: Local Node Execution

```json
{
  "mcpServers": {
    "vibecraft-mcp": {
      "command": "node",
      "args": [
        "/path/to/vibecraft-mcp/dist/index.js"
      ],
      "env": {
        "UI_UX_MCP_API_KEY": "your_api_key_here"
      }
    }
  }
}
```

---

## πŸ›  The 7 Core MCP Tools

### 1. `get_layout_blueprint`
Retrieves Nielsen Norman Group (NN/g) Information Architecture macro layout scaffolds:
* `saas-dashboard` β€” Collapsible sidebar, top utility header, stat grid, filterable data table.
* `saas-settings` β€” Task-oriented vertical tabs with sticky unsaved changes bar.
* `ai-generation-pattern` β€” AI co-pilot streaming prompt container and action bar.
* `destructive-confirm-pattern` β€” Double-confirmation safety dialog with undo ledges.

### 2. `get_ux_guidelines`
Queries embedded design principles & product psychology rules:
* **Frank Chimero's TSOD**: 3 Levers (Message, Tone, Format), How vs Why Balance.
* **Nir Eyal's Hook Model**: Habit loops ($B=MAT$ action friction reduction, variable rewards, investment).
* **Interaction Design Foundation (IxDF)**: Morville 7 Factors Honeycomb, Quesenberry 5 Cs.
* **Laws of UX & Growth.design**: Hick's Law, Fitts's Law (44px touch targets), Doherty Threshold (<400ms skeleton UI).

### 3. `audit_ux_compliance`
Audits React & Tailwind code for UX anti-patterns:
* Competing primary CTA buttons (`bg-primary` on 3+ adjacent controls).
* Hardcoded hex colors (`bg-[#121212]`) missing HSL semantic design tokens.
* Touch targets smaller than 44x44px.
* Icon-only buttons lacking `aria-label`.

### 4. `audit_system_performance`
Audits backend code and database queries:
* **N+1 DB Query Elimination**: Refactors loop queries to ORM `include` / `JOIN FETCH`.
* **Sequential Promise Waterfalls**: Parallelizes independent calls with `Promise.all()`.
* **Transaction Safety**: Wraps multi-step mutations in `db.$transaction()` with automatic rollback.
* **OWASP IDOR Security**: Validates tenant/user ownership checks (`WHERE id = params.id AND tenantId = user.tenantId`).

### 5. `search_ui_components`
Searches 21st.dev and curated component registry by keyword, category, or tag (Glowing cards, Floating magnetic docks, Bento grids).

### 6. `get_component_code`
Retrieves copy-paste ready React + Tailwind CSS + Framer Motion code with `shadcn` CLI install commands. Supports live HTTP fallback fetching.

### 7. `get_design_tokens`
Provides ready-to-copy `globals.css` light & dark mode HSL CSS variables, typography scales, spatial grid, icon stroke rules, and spring animation parameters.

---

## πŸ“š Embedded Design & Engineering Frameworks

| Framework | Core Concept | Purpose in Vibe-Coding |
| :--- | :--- | :--- |
| **Frank Chimero's TSOD** | 3 Levers (Message, Tone, Format) | Ensures design choices match user context & utility. |
| **Nir Eyal's Hook Model** | Trigger $\rightarrow$ Action $\rightarrow$ Reward $\rightarrow$ Investment | Builds habit-forming SaaS features and delight loops. |
| **Nielsen Norman Group** | Information Architecture (IA) | Prevents broken, chaotic SaaS page layouts. |
| **Apple HIG & WCAG AAA** | Touch targets & Access | Mandates 44x44px touch targets and screen-reader accessibility. |
| **transitions.dev** | Shared Element Motion | Physics spring timing (`stiffness: 300, damping: 30`). |
| **21st.dev & shadcn** | Component Registry | Production-grade UI code primitives. |

---

## πŸ‘€ Author & Contribution

Developed by **Satyam** ([@originsatyam](https://github.com/originsatyam)).

Contributions, issues, and feature requests are welcome!

```bash
git clone https://github.com/originsatyam/vibecraft-mcp.git
cd vibecraft-mcp
npm install
npm run build
```

---

## πŸ“„ License

MIT Β© [originsatyam](https://github.com/originsatyam)

TDQS

A3.6/5.0

Scored across 7 tools

Disambiguation3/5

Most tools have clear boundaries, but audit_ux_compliance and get_ux_guidelines reference nearly the same set of UX rules and heuristics, making their purposes easy to confuse. get_layout_blueprint also overlaps with get_ux_guidelines via NN/g IA content, though the descriptions help differentiate them.

Naming Consistency5/5

All tools follow a consistent verb_noun snake_case pattern: audit_*, get_*, search_*. This makes the toolset predictable and easy to navigate.

Tool Count5/5

Seven tools is well-scoped for a design/UX assistant server. Each tool covers a distinct part of the workflow: auditing, guidelines, layout, components, and design tokens, without feeling bloated or thin.

Completeness4/5

The toolset covers the core design/UX loop well: retrieve guidelines and tokens, find and fetch components, and audit both system performance and UX compliance. Minor gaps exist, such as no explicit tool for applying fixes or generating a consolidated audit report, but agents can work around these.

Maintenance

ActivityMaintained
ResponsivenessNo issues