sujathaynweh/sampleMCPToolServer

Sample MCP Tool Server

This repository contains a simple Flask-based server implementing an MCP (Model Context Protocol) tool, which exposes a /run-tool API endpoint and serves its OpenAPI specification. This tool can be connected to Open WebUI and used with a locally hosted LLaMA 3 model.

Features

• Flask API with a single POST endpoint: /run-tool
• Serves an OpenAPI spec at /.well-known/openapi.json
• CORS enabled
• Ready to integrate with Open WebUI

Prerequisites

• Python 3.8+
• Open WebUI installed and running
• LLaMA 3 model downloaded and configured

Installation

1. Clone this repository:

   git clone https://github.com/your-username/sampleMCPToolServer.git
   cd sampleMCPToolServer
   

2. Create and activate a virtual environment (optional but recommended):

   python3 -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate
   

3. Install dependencies:

   pip install flask flask-cors
   

Folder Structure

sampleMCPToolServer/
│
├── app.py
└── static/
    └── .well-known/
        └── openapi.json

Running the Server

python app.py

This starts the server on http://localhost:5050.

• OpenAPI spec is available at: http://localhost:5050/.well-known/openapi.json
• Tool endpoint: http://localhost:5050/run-tool

Connecting to Open WebUI

Downloading a Model in Open WebUI

1. Go to the Models tab in Open WebUI.

2. Search for and download a supported LLM model by specifying the model name (e.g., llama3).

3. Once the model is downloaded and loaded, it can be used to call the tool.

4. Install and start Open WebUI:

   Follow the instructions at https://github.com/open-webui/open-webui

5. Add your tool to Open WebUI:

   • Navigate to Open WebUI
   • Go to Settings > Tools > Add Tool
   • Enter the URL to your OpenAPI JSON:

     http://localhost:5050/.well-known/openapi.json
     

   • Save and verify that the tool is successfully added.

6. Use the tool in chats:

   Type messages that reference or trigger the tool (e.g., "Use the test tool with name John").

Example Request

Endpoint:

POST /run-tool

Request Body:

{
  "name": "Alice"
}

Response:

{
  "message": "Hello, Alice! The tool was successfully called."
}

License

MIT License

Author

• sujathay.nweh