Day 6 — Add a Live Weather Tool

MCP & AI Agents — Agent Bootcamp Part 2 — RohithBuilds

Every tool your agent has so far works with data you created — exam dates, your notes. Today you'll give it a tool that pulls real, live data from the internet: current weather for any city in the world, using a free API that needs no signup and no API key.

Step 1 — Recap & Today's Plan

The plan is short and familiar — you've done this shape of work four times already:

• Write a logic function that calls a public weather API
• Wrap it with @mcp.tool()
• Test it in Inspector
• Then — the fun part — ask your Day 5 chat UI a weather question and watch it pick up the new tool with zero changes to app.py

Step 2 — Install requests

This is a new tool for this course's venv:

pip install requests

Expected Output

Step 3 — Build the Weather Lookup

We'll use Open-Meteo — a free weather API that needs no key. It works in two steps: turn a city name into coordinates (geocoding), then ask for the weather at those coordinates.

Open server.py and add import requests to your imports. Then add this above if __name__ == "__main__"::

WEATHER_CODES = {
    0: "clear sky",
    1: "mainly clear",
    2: "partly cloudy",
    3: "overcast",
    45: "fog",
    48: "freezing fog",
    51: "light drizzle",
    53: "moderate drizzle",
    55: "dense drizzle",
    61: "slight rain",
    63: "moderate rain",
    65: "heavy rain",
    71: "slight snow",
    73: "moderate snow",
    75: "heavy snow",
    80: "rain showers",
    81: "moderate rain showers",
    82: "violent rain showers",
    95: "thunderstorm",
}


def get_weather_logic(city: str) -> str:
    geo = requests.get(
        "https://geocoding-api.open-meteo.com/v1/search",
        params={"name": city, "count": 1}
    ).json()

    if not geo.get("results"):
        return f"Could not find a location called '{city}'."

    location = geo["results"][0]
    lat = location["latitude"]
    lon = location["longitude"]
    name = location["name"]
    country = location.get("country", "")

    weather = requests.get(
        "https://api.open-meteo.com/v1/forecast",
        params={"latitude": lat, "longitude": lon, "current_weather": True}
    ).json()

    current = weather["current_weather"]
    condition = WEATHER_CODES.get(
        current["weathercode"], f"weather code {current['weathercode']}"
    )

    return (
        f"{name}, {country}: {current['temperature']}°C, "
        f"{condition}, wind {current['windspeed']} km/h"
    )

WEATHER_CODES covers the most common conditions. Anything not in the dictionary still works — it just shows the raw code as a fallback.

Step 4 — Add get_weather as an MCP Tool

Add the wrapper, same pattern as every tool this week:

@mcp.tool()
def get_weather(city: str) -> str:
    """Get the current weather for a city. Works for cities worldwide."""
    return get_weather_logic(city)

Your server now has 5 tools: exam_countdown, add_note, list_notes, delete_note, and get_weather — plus the resources and prompts from Day 3.

Step 5 — Restart Inspector & Test get_weather

mcp dev server.py

In the Tools tab, click get_weather, enter city: Bengaluru, and run it.

Expected Output

Try a few more cities — Mumbai, Delhi, Tokyo, anything. No restarts, no new keys — Open-Meteo covers the whole world.

Step 6 — The Big Test: Ask Your Chat UI

Here's the payoff. Make sure app.py from Day 5 is running:

python app.py

Open http://127.0.0.1:5000 and ask:

• What's the weather in Bengaluru right now?

Expected Output

You didn't touch app.py, agent.py, or templates/index.html — yet your chat UI can now answer weather questions. This is the entire point of MCP: add a tool to the server, and every client that uses it gets the new capability automatically.

Now try something that needs reasoning on top of the data:

• Should I carry an umbrella in Mumbai today?

Expected Output

Groq fetched the real weather data via your tool, then reasoned about it to answer a question the tool itself was never designed to answer. That combination — real data plus reasoning — is what makes MCP-powered agents useful.

Step 7 — Fallback Test Without Inspector

Add to test_local.py:

from server import get_weather_logic

print(get_weather_logic("Bengaluru"))
print(get_weather_logic("Mumbai"))
print(get_weather_logic("Nonexistentcityxyz"))

Run it:

python test_local.py

Expected Output

Notice the third line handles a city that doesn't exist gracefully — no crash, just a clear message. Good tools fail politely.

Step 8 — Save Your Progress

pip freeze > requirements.txt

Your server now has 5 tools, 2 resources, and 2 prompts — all built across 6 days, all running through the same chat UI. One day left.

✅ Day 6 Complete

Here is what you accomplished today:

What Is Coming Tomorrow

On Day 7 — the final day — you will:

• Push your complete project to GitHub
• Deploy it live on Render, just like Part 1, Day 7
• Write a README explaining your MCP architecture
• Get a public URL anyone can use to chat with your MCP-powered agent

See you there — last day! 🚀