🇮🇱 Israeli Weather MCP with Playwright

📌 Overview

This project implements an MCP (Model Context Protocol) Server that allows an LLM to retrieve Israeli weather forecasts using Playwright browser automation.

Instead of using a weather API, the MCP opens the Weather2day website in a real browser, searches for the requested city, selects it, extracts the weather information from the page, and provides the extracted content back to the LLM.

The project demonstrates how MCP + Playwright + LLM can work together to give an AI agent browser-based capabilities.

Related MCP server: MCP-with-Playwright

🎯 Project Goals

The project was developed to practice:

• Building an MCP Server independently.

• Creating custom MCP Tools.

• Using Playwright for browser automation.

• Allowing an LLM to control a browser through MCP Tools.

• Extracting and cleaning web page content.

• Providing extracted page content back to the LLM so it can answer the user's question.

🛠️ Technologies

• Python

• MCP SDK / FastMCP

• Playwright

• uv

• Weather2day

• LLM

🧩 How It Works

The user asks the LLM for the weather in an Israeli city.

The LLM then uses the MCP Tools in the following order:

User
  ↓
LLM
  ↓
open_weather_forecast_israel
  ↓
enter_weather_forecast_city_israel
  ↓
select_weather_forecast_city_israel
  ↓
get_weather_forecast_content_israel
  ↓
Playwright extracts the page content
  ↓
Content is cleaned and returned to the LLM
  ↓
LLM answers the user

🔧 MCP Tools

1. open_weather_forecast_israel

Opens the Weather2day Israeli forecast page using Playwright.

https://www.weather2day.co.il/forecast

2. enter_weather_forecast_city_israel

Receives a city name and enters it into the Weather2day search field.

Example:

city = "בני ברק"

The Tool also waits for the city's autocomplete suggestions.

3. select_weather_forecast_city_israel

Selects the requested city from the visible autocomplete suggestions.

This completes the browser interaction required to reach the city's forecast page.

4. get_weather_forecast_content_israel

Extracts the visible content from the currently selected city's page.

The Tool:

1. Reads the page content using Playwright.

2. Removes empty lines.

3. Removes duplicate lines.

4. Limits the returned content size.

5. Returns the cleaned information to the LLM.

This Tool implements the second stage of the project, where the LLM receives information extracted from the web page and uses it to formulate the final answer.

📂 Project Structure

mcp/
├── host.py
├── client.py
├── weather_Israel.py
├── pyproject.toml
└── README.md

weather_Israel.py

Contains the Israeli Weather MCP Server and its Playwright-based Tools.

host.py

Runs the terminal chat and connects the LLM to the MCP Tools.

client.py

Provides the MCP client functionality.

🚀 Installation

1. Install dependencies

From the project directory:

uv sync

2. Install the Playwright Chromium browser

uv run playwright install chromium

▶️ Running the Project

Start the Host:

uv run host.py

You should see:

Israeli Weather MCP is ready.
Type 'exit' to quit.

You can then ask questions about Israeli cities.

💬 Example

User

give me the weather in Bne Brak

MCP Tool calls

[CITY] Using city: בני ברק
[MCP] Calling open_weather_forecast_israel({})
[MCP] Calling enter_weather_forecast_city_israel({'city': 'בני ברק'})
[MCP] Calling select_weather_forecast_city_israel({})
[MCP] Calling get_weather_forecast_content_israel({})

Assistant

Current temperature in Bne Brak: 25°C.
Wind speed: 5 km/h from the southwest.
Humidity: 87%.
Last updated: 07:00.

The final answer is generated by the LLM using the weather information extracted from the browser page.

💡 Example Questions

The Agent can answer questions such as:

What is the weather in Jerusalem?

Give me the weather in Bne Brak.

What's the current temperature in Tel Aviv?

What is the humidity in Haifa?

🌐 Data Source

Weather information is retrieved from:

Weather2day – Israeli Weather Forecast

https://www.weather2day.co.il/forecast

The project intentionally uses browser automation with Playwright instead of a dedicated weather API.

🤖 Why Playwright?

Playwright allows the MCP Server to interact with a real browser programmatically.

In this project, Playwright is responsible for:

• Opening the website.

• Entering the city.

• Finding the city suggestion.

• Clicking the city.

• Reading the resulting forecast page.

This allows the LLM to interact with a website through MCP Tools without requiring manual browser interaction.

🧠 MCP + Browser Automation

This project demonstrates how MCP can extend an LLM with capabilities beyond generating text.

The LLM decides which Tool to call, while the MCP Server handles the actual browser interaction.

This creates a simple architecture where:

LLM = decides what needs to be done
        ↓
MCP = provides the available Tools
        ↓
Playwright = performs the browser actions
        ↓
Weather2day = provides the weather data
        ↓
LLM = interprets the data and answers

📚 Learning Outcomes

Through this project I practiced:

• Creating an MCP Server with FastMCP.

• Defining MCP Tools using decorators.

• Connecting an LLM to custom MCP Tools.

• Automating browser interactions with Playwright.

• Handling dynamic web pages and autocomplete suggestions.

• Extracting and cleaning web page content.

• Returning web content to an LLM for further reasoning.

👩‍💻 Project

Built as part of an MCP + Playwright learning project focused on giving LLMs browser automation capabilities.