MCP Server

from flask import Flask, request, jsonify import os import requests import json from mcp_handler import MCPHandler from tools.weather import WeatherTool from tools.calculator import CalculatorTool app = Flask(__name__) # Initialize MCP handler with available tools handler = MCPHandler() handler.register_tool(WeatherTool()) handler.register_tool(CalculatorTool()) # LLM API configuration LLM_API_KEY = os.environ.get("LLM_API_KEY") LLM_API_URL = "https://api.llmprovider.com/v1/chat/completions" # Example URL @app.route("/chat", methods=["POST"]) def chat(): data = request.json messages = data.get("messages", []) # If this is the first message, inject tool definitions if len(messages) == 1: system_message = { "role": "system", "content": f"You are an AI assistant with access to the following tools:\n{handler.get_tool_definitions()}" } messages.insert(0, system_message) # Get response from LLM response = get_llm_response(messages) # Check if the response contains tool invocations processed_response = handler.process_response(response) # Add the processed response to messages for context messages.append({"role": "assistant", "content": processed_response}) return jsonify({ "response": processed_response, "messages": messages }) def get_llm_response(messages): """Get a response from the LLM API""" headers = { "Content-Type": "application/json", "Authorization": f"Bearer {LLM_API_KEY}" } payload = { "model": "llm-model-name", "messages": messages, "max_tokens": 1000 } response = requests.post(LLM_API_URL, headers=headers, json=payload) return response.json()["choices"][0]["message"]["content"] if __name__ == "__main__": app.run(debug=True, host="0.0.0.0", port=5000)