Content
# Universal AI Agent Router (Infinity MCP) Orchestrator
The **Universal AI Agent Router** (powered by **Infinity MCP**) is an autonomous orchestration layer designed to coordinate specialized agent capabilities, memory systems, and design frameworks to build premium, slop-free software and web experiences.
By using persistent memory, dynamic capability loading, and a sequenced execution pipeline, it abstracts the complexity of selecting individual design and coding tools from the client.
---
## 🚀 Core Features
### 1. Sequenced 5-Phase Web Construction Pipeline
When website tasks are detected, the orchestrator executes a strict sequence instead of raw parallel execution to prevent AI slop:
* **Phase 1: Initial Design Foundation**: Utilizes `open-design` and the `frontend-design` prompt guidelines to initialize CSS variables, HSL colors, layout grids, and typographic scales.
* **Phase 2: Component Assembly & Animations**: Integrates components using premium libraries (`shadcn`, `magicui`, `aceternity-ui`, etc.) and mandates animations (entrance, exit, hover, scroll-triggered, and ambient micro-animations) across all components.
* **Phase 3: Refinement & Validation**: Enhances the code structure using secondary layout tools and performs static syntax checks via `debug-skill`.
* **Phase 4: Final Premium Polish**: Employs `taste-skill` and `impeccable` at the end to adjust spacing, typography, and colors. Checks for styling constraints:
- **No rainbow gradients on dark backgrounds** (uses bold/cohesive colors).
- **No excessive glow effects** (prevents visual noise/clutter).
* **Phase 5: Visual Quality Check & Critique**: Launches `playwright-mcp` to capture screenshots and run visual sanity checks, performing a final self-critique pass to guarantee a premium look.
### 2. Integrated Design Knowledge Training
Pre-loaded with **74 top-tier design system guides** from the `awesome-design-md` repository (e.g., Stripe, Linear, Vercel, Apple). When a design task is received, the context is automatically retrieved from Graphify and injected into the execution pipeline, enabling the orchestrator to model its outputs after world-class design systems.
### 3. Graphify Memory Integration
Maintains session-scoped knowledge. Every completed task indexes its outcomes back into the graph, while new tasks begin by retrieving relevant context, past patterns, and developer preferences from Graphify.
---
## 🛠️ MCP Tools Exposed
Clients (IDE extension, AI agent, or custom UI) interface with the orchestrator using four primary MCP tools:
1. **`activate_agent`**: Activates a specialized agent configuration (e.g., `website_builder`, `motion_animation`, `coding`, `research`, `security`) and loads its specific capabilities.
2. **`execute_task`**: Executes complex multi-step workflows. Automatically resolves agent intents, fetches context from Graphify, runs the sequential pipeline, performs visual reviews, and index results.
3. **`delegate_subtask`**: Delegates specific sub-tasks to internal sub-agents.
4. **`get_status`**: Queries active orchestrator states, loaded skills, and Graphify tasks.
---
## 🚀 Installation & Setup
### Prerequisites
* [Node.js](https://nodejs.org/) (v18+)
* [npm](https://www.npmjs.com/)
### 1. Install Dependencies
Clone the repository and install packages:
```bash
npm install
```
### 2. Build the Server
Compile the TypeScript code:
```bash
npm run build
```
### 3. Add to MCP Config
Add the server definition to your MCP client config (e.g. Claude Desktop Config or Cursor MCP config):
**Stdio Transport (Default)**:
```json
{
"mcpServers": {
"universal-orchestrator": {
"command": "node",
"args": [
"c:/Users/Shriyans/Desktop/MCP/universal-mcp/dist/index.js"
]
}
}
}
```
**SSE Transport**:
To start the server using Server-Sent Events (SSE):
```bash
npm run start -- --sse
```
And add it to your configuration using HTTP:
```json
{
"mcpServers": {
"universal-orchestrator-sse": {
"command": "npx",
"args": [
"-y",
"@modelcontextprotocol/client",
"http://localhost:3000/sse"
]
}
}
}
```
---
## 📂 Project Structure
```
universal-mcp/
├── src/
│ ├── config/ # Agent capabilities configuration (agents.ts)
│ ├── loaders/ # Dynamic skill loader (skillsLoader.ts)
│ ├── memory/ # Graphify memory manager (graphify.ts)
│ ├── orchestration/ # OhMyPI framework delegation (ohmypi.ts)
│ ├── prompts/ # Core system and frontend prompt guidelines
│ └── index.ts # Server entry point and MCP tool handlers
├── knowledge/ # awesome-design-md training data
├── SKILL.md # Orchestrator orchestration guide (antigravity skill)
└── tsconfig.json # TypeScript configuration
```
