mcpify

⚡ MCPify

Analyze any AI agent URL, detect framework capabilities, and generate ready-to-use Model Context Protocol (MCP) configurations. MCPify also exposes its own /mcp endpoint so it can itself be added as an MCP connector to any MCP client (Claude Desktop, Cursor, VS Code, etc.).

🌐 Live: mcpify.aikart.co  •  Built by Aikart  •  MIT Licensed

MCPify converts any AI agent, REST API, GraphQL API, or OpenAPI specification into a ready-to-use Model Context Protocol (MCP) server - with one-click configs for Claude Desktop, Cursor, Windsurf, Cline, VS Code, and the Claude Code CLI.

---

🚀 Features

• 🔍 Universal Agent URL Analysis: Probes common endpoints (/mcp, /sse, /health, /tools, /docs, /openapi.json) concurrently with HTTP heuristics.
• 🧠 Framework Detection: Identifies FastAPI, Flask, LangChain/LangGraph, Express, and Next.js agents with confidence scoring.
• ⚙️ Instant MCP Config Generator: Outputs ready-to-copy configuration JSON for:
  • Claude Desktop (mcpServers format)
  • Cursor & VS Code (SSE/Remote MCP format)
  • Remote MCP Endpoint
• 🔌 Self-Hosting MCP Server: Powered by fastapi-mcp, exposing its own /mcp endpoint and tools directly to AI clients.
• ⏱️ Zero-Downtime Free Tier Keep-Alive: Integrated APScheduler keep-alive worker pinging /health every 10 minutes to prevent Render/Koyeb sleeping.
• 🌐 Full CORS Support: Open for web frontends, browser extensions, and developer portals.

---

📁 Project Structure

mcpify/
├── main.py              # FastAPI app entry point with FastApiMCP & APScheduler
├── requirements.txt     # Python dependencies
├── render.yaml          # Render deployment blueprint
├── .env.example         # Sample environment configuration
├── README.md            # Project documentation & guides
└── app/
    ├── __init__.py      # Package initialization
    ├── analyzer.py      # URL probing & framework detection heuristics
    ├── generator.py     # MCP configuration JSON generator
    └── mcp_handler.py   # MCP tools definition & REST router

---

🛠️ MCP Tools Exposed

When connected via /mcp, MCPify exposes three primary tools:

Tool Name	Parameters	Description
analyze_agent	url (string)	Probes agent endpoints, detects framework, returns confidence score and accessible routes.
generate_mcp_config	url (string)	Analyzes agent and generates complete Claude Desktop and Cursor JSON configs.
get_integration_guide	url (string), platform (claude_desktop | cursor | web)	Step-by-step instructions for adding the agent to your client.

---

💻 Local Setup & Installation

1. Clone & Navigate

git clone <your-repo-url>
cd mcpify

2. Create Virtual Environment

# Windows
python -m venv venv
venv\Scripts\activate

# macOS / Linux
python3 -m venv venv
source venv/bin/activate

3. Install Dependencies

pip install -r requirements.txt

4. Configure Environment

cp .env.example .env

At minimum, set:

APP_URL=http://localhost:10000

Running without a database: DATABASE_URL, JWT_SECRET_KEY, etc. are all optional for local dev - /analyze, /generate, /guide, and the MCP tools work fine without them (no signup/login required historically; they now gate behind auth in production - see note below). Only signup/login/dashboard/admin-ads/GA endpoints need a real Postgres connection. The test suite (pytest) never needs any of this either.

Getting real values (teammates with Railway access): this project deploys from Railway, not the render.yaml below (that's legacy from before the move). There's no committed .env - env vars live in Railway's dashboard. If you need to run against the same DB/Redis/GA/Cloudinary the production deploy uses: Railway project → Variables tab → copy each value into your local .env. Don't commit .env or paste these values into chat/commits - .env is already gitignored.

Note: /analyze, /generate, and /guide currently require a logged-in user (Authorization: Bearer <jwt>) even in this local setup, so without DATABASE_URL + JWT_SECRET_KEY configured you can exercise the analyzer/generator logic directly via app/analyzer.py's functions or the MCP tools in app/mcp_handler.py, but not the REST endpoints end-to-end.

5. Run the Server

uvicorn main:app --host 0.0.0.0 --port 10000 --reload

• Interactive Swagger API Docs: http://localhost:10000/docs
• MCP Endpoint: http://localhost:10000/mcp
• Health Check: http://localhost:10000/health

---

🚢 Deployment

Current production: Railway. The repo has no railway.json/railway.toml - Railway auto-detects the Python app from requirements.txt and runs uvicorn main:app. To deploy your own instance: create a Railway project, connect this repo, add a Postgres database (or point DATABASE_URL at Neon/any provider), and set the env vars from .env.example in Railway's Variables tab.

This project includes a render.yaml for 1-click deployment on Render's free tier, kept for anyone who prefers Render over Railway - not what the live deployment actually runs on.

1. Push this repository to GitHub or GitLab.
2. Log into Render Dashboard.
3. Click New + > Blueprint.
4. Connect your repository.
5. Set APP_URL in the environment variables to your assigned Render URL (e.g. https://mcpify.onrender.com).
6. Deploy! The built-in APScheduler will automatically keep your service awake.

---

🔌 Connecting MCPify to Your Client

1. Claude Desktop

Add MCPify to claude_desktop_config.json:

{
  "mcpServers": {
    "mcpify": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://your-mcpify-url.onrender.com/mcp"
      ]
    }
  }
}

2. Cursor / VS Code

In Cursor Settings > Features > MCP > Add New MCP Server:

• Name: mcpify
• Type: sse
• URL: https://your-mcpify-url.onrender.com/mcp

---

📡 API Endpoints

1. Health Check

GET /health

{
  "status": "ok"
}

2. Analyze Agent URL

POST /analyze

{
  "url": "https://sample-agent.onrender.com"
}

Response:

{
  "url": "https://sample-agent.onrender.com",
  "detected_framework": "FastAPI",
  "confidence_score": 0.95,
  "available_endpoints": [
    {
      "path": "/health",
      "status_code": 200,
      "accessible": true,
      "content_type": "application/json"
    },
    {
      "path": "/docs",
      "status_code": 200,
      "accessible": true,
      "content_type": "text/html; charset=utf-8"
    }
  ],
  "recommended_mcp_endpoint": "https://sample-agent.onrender.com/mcp",
  "details": {
    "signals": [
      "Server header contains 'uvicorn'",
      "OpenAPI specification available at /openapi.json",
      "Swagger UI documentation available at /docs"
    ],
    "framework_scores": {
      "FastAPI": 0.95,
      "Flask": 0.0,
      "LangChain": 0.0,
      "Express": 0.0,
      "Next.js": 0.0
    }
  }
}

3. Generate MCP Config

POST /generate

{
  "url": "https://sample-agent.onrender.com"
}

4. Integration Guide

POST /guide

{
  "url": "https://sample-agent.onrender.com",
  "platform": "claude_desktop"
}

---

📜 License

MIT License. Built for seamless AI agent interoperability with Model Context Protocol.