Forecasting MCP

This is a Model Context Protocol (MCP) implementation for interacting with a forecasting website. It enables LLMs to make, update, and delete forecasts, as well as retrieve forecast data and relevant news articles.

Features

• Make forecasts on an existing forecasting website
• Create new forecasts
• Update existing forecasts
• Delete forecasts
• Access news articles to inform forecasts

Setup

1. Install the required dependencies:

pip install -r requirements.txt

2. Create a .env file with the following variables:

API_URL=<your_forecasting_api_url>
BOT_USERNAME=<your_bot_username>
BOT_PASSWORD=<your_bot_password>
NEWS_API_KEY=<your_news_api_key>  # Optional, for accessing news articles

Usage

Run the server with:

uvicorn forecasting:app --reload

Available MCP Tools

Forecasts Management

• get_forecasts: List forecasts by category and status
• get_forecast_data: Get details for a specific forecast
• create_forecast: Create a new forecast
• update_forecast: Modify an existing forecast
• delete_forecast: Remove a forecast

Forecast Points

• get_forecast_points: Get all forecast points for a forecast
• create_point: Make a new forecast point with reasoning

News Integration

• get_news_articles: Fetch news articles related to a query to inform forecasts

Development

This server uses FastAPI and the MCP library to provide a standardized interface for LLMs to interact with forecasting data.