ragx-aipm

RAGX AI-Native Package Manager — An MCP server that analyzes any workspace and generates context-aware, zero-dependency implementations.

What is this?

ragx-aipm is not a traditional package manager. It does not download static code.

Instead, it is an MCP (Model Context Protocol) server that:

1. Scans your workspace — detects your language, framework, dependencies, database, and existing agent configs.
2. Pulls a Canonical Prompt — retrieves a highly refined set of instructions for the requested feature.
3. Returns the prompt to your LLM — your AI assistant (Claude, Cursor, Gemini, etc.) then generates a bespoke implementation perfectly matched to your project's exact architecture.

No two installations are the same. Every generated implementation is tailored to your stack.

Available Packages

| Package | Description | |---|---| | ragx-core | Zero-dependency in-memory vector database with cosine similarity search | | ragx-auth | Context-aware authentication — adapts to Firebase, Supabase, or implements JWT from scratch | | ragx-search | Semantic search UI component powered by local vector similarity | | ragx-embed | Zero-dependency text embedding generator (TF-IDF to float vectors) |

Quick Start

Option 1: Run directly with npx (after publishing)

npx ragx-aipm

Option 2: Add to your MCP client config

Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "ragx-aipm": {
      "command": "npx",
      "args": ["-y", "ragx-aipm"]
    }
  }
}

Cursor (.cursor/mcp.json):

{
  "mcpServers": {
    "ragx-aipm": {
      "command": "npx",
      "args": ["-y", "ragx-aipm"]
    }
  }
}

VS Code / Gemini (.vscode/settings.json):

{
  "mcpServers": {
    "ragx-aipm": {
      "command": "npx",
      "args": ["-y", "ragx-aipm"]
    }
  }
}

Option 3: Install locally and run

cd ragx-aipm
npm install
npm start

MCP Tools

analyze_workspace

Scans a directory and returns a detailed JSON report of the detected tech stack.

list_packages

Returns all available RAGX packages with descriptions.

install_package

The core AIPM action. Analyzes the workspace, then returns a contextualized implementation prompt. The LLM then executes those instructions to generate files.

How it works

┌─────────────────────┐
│  LLM Client         │  (Claude, Cursor, Gemini, etc.)
│  "Install ragx-auth"│
└────────┬────────────┘
         │ stdio (JSON-RPC)
         ▼
┌─────────────────────┐
│  ragx-aipm MCP      │
│  ├─ analyze_workspace│  → Scans package.json, tsconfig, etc.
│  ├─ list_packages    │  → Shows available RAGX packages
│  └─ install_package  │  → Generates contextualized prompt
└────────┬────────────┘
         │
         ▼
┌─────────────────────┐
│  Canonical Prompt    │  "You are integrating ragx-auth into a
│  (Context-Aware)     │   Next.js 16 TypeScript project with
│                     │   Firebase Auth already installed..."
└────────┬────────────┘
         │
         ▼
┌─────────────────────┐
│  LLM Generates      │  Bespoke, zero-dependency code
│  Tailored Code      │  matched to YOUR exact project
└─────────────────────┘

License

MIT