mcp-server

vivekprajapat305/mcp-server

3.1

If you are the rightful owner of mcp-server and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.

This repository contains a simple Model Context Protocol (MCP) server implemented with FastAPI, designed to interface with ChatGPT connectors.

MCP Server

This repository contains a simple Model Context Protocol (MCP) server implemented with FastAPI. The server exposes two endpoints that are compatible with the ChatGPT connectors feature.

Endpoints

POST /search

Accepts a JSON body with a query string and returns search results from your data source. Currently the implementation returns dummy results but you can customise it to return real content.

POST /fetch

Accepts a JSON body with a list of document ids and returns the document content.

Running locally

  1. Clone this repository:

    git clone https://github.com/vivekprajapat305/mcp-server.git
cd mcp-server

  2. Create a virtual environment and install dependencies:

    python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

  3. Start the server:

    uvicorn main:app --host 0.0.0.0 --port 8000

The server will run on http://localhost:8000.

Connecting to ChatGPT

To connect this MCP server to ChatGPT, follow these steps:

  1. Deploy the server to a publicly accessible URL (e.g., using a cloud service like Render, Heroku, or your own server).
  2. In ChatGPT's "Actions" or "Connectors" settings, choose to add a new MCP connector and enter the base URL of your running server (e.g., https://your-domain.com).
  3. ChatGPT will automatically discover the /search and /fetch endpoints and allow you to ask questions against your data.

Customising

The current implementation in main.py includes placeholder logic. You should replace the search and fetch functions with your own logic to retrieve data from your actual sources.