mcp-lens-ui 0.1.3

Swagger UI & Developer Tools for the Model Context Protocol

MCP Lens

MCP Lens is the Swagger UI equivalent for the Model Context Protocol (MCP). It provides a beautiful, interactive, and dark-themed web interface to explore, execute, and debug your MCP servers locally.

MCP Lens automatically intercepts calls to your MCP Server (using fastmcp), aggregates the schema of all available tools, resources, and prompts, and instantly spins up a local web UI.

---

🚀 Features

• 🛠 Tools Explorer: View all available tools, their JSON schemas, and test them with a live form. See execution traces and server responses.
• 📚 Resources Explorer: Browse all standard and dynamically paginated resources exposed by your MCP servers and inspect their values instantly.
• 💬 Prompts Explorer: View predefined prompt templates and invoke them dynamically to see exactly how they will format for LLMs.
• 📜 Request History: Keep a log of all executions along with performance metrics (execution time, status codes) and effortlessly replay them.
• 🔐 Multi-Server & Auth: Supports instrumenting multiple MCP servers concurrently and provides an interface for Bearer token authorizations per server.

---

📦 Installation

To use MCP Lens in your project, simply install it via pip:

pip install mcp-lens-ui

(Note: Ensure your MCP server is built using the fastmcp package.)

---

⚡ How to Use

Using MCP Lens is incredibly easy. Just import instrument from mcp_lens and wrap your FastMCP app instances.

Minimal Example (example.py)

Here is a full example showing how to create a basic weather MCP server and visualize it with MCP Lens.

import httpx
from fastmcp import FastMCP
from mcp_lens import instrument

app = FastMCP("OpenMeteo Weather")

@app.tool()
async def get_current_weather(latitude: float, longitude: float) -> str:
    """Get the current weather for a specific latitude and longitude. Uses Open-Meteo free API."""
    url = f"https://api.open-meteo.com/v1/forecast?latitude={latitude}&longitude={longitude}&current_weather=true"
    
    async with httpx.AsyncClient() as client:
        response = await client.get(url)
        data = response.json()
        
    if "current_weather" in data:
        weather = data["current_weather"]
        return f"Current temperature is {weather['temperature']}°C with wind speed {weather['windspeed']} km/h."
    return "Weather data not available."

@app.resource("weather://about")
def about_weather_api() -> str:
    """Information about the weather data source."""
    return "This weather data is provided by Open-Meteo, a free open-source weather API!"

@app.prompt()
def pack_for_trip(location_name: str) -> str:
    """Ask an AI to help pack for a trip based on the weather."""
    return f"I am taking a trip to {location_name}. Can you check the weather forecast for that location and tell me what clothes I should pack?"

# The magic line that turns this into a Swagger UI!
instrument(app, ui=True, ui_port=8000)

if __name__ == "__main__":
    print("Starting Open-Meteo Weather MCP Server...")
    app.run()

When you run this script (python example.py), it will start your standard MCP server on stdio while launching the MCP Lens UI in the background on http://localhost:8000/mcp. Open that URL in your browser to start exploring!

---

🧠 Architecture

MCP Lens operates uniquely to coexist with your production stdio MCP server without interfering with how LLM clients communicate with it.

1. Instrumentation layer: The instrument(app) function reaches into the internal FastAPI router of the FastMCP app.
2. Metadata Aggregation: It queries /tools/list, /resources/list, and /prompts/list endpoints to build an internal state (AppState) of your MCP endpoints.
3. Background UI Server: It spawns an independent FastAPI application running a Uvicorn ASGI server in a separate daemon thread on a specified port (default 8000).
4. Vite React Frontend: The web UI (built with React + Tailwind + Vite) is bundled and served statically by the background FastAPI server. It communicates with the UI backend to dispatch simulated MCP requests dynamically.

Because the UI server runs on a completely distinct port/thread, your LLM client continues communicating natively over stdio undisturbed.

---

🤝 Contributing

Contributions are heavily welcomed! Whether it's fixing bugs, adding new features, or improving the frontend aesthetic.

1. Fork the repository
2. Create a feature branch
3. Submit a Pull Request

📄 License

MIT License