Model Context Protocol (MCP)

Bridging AI Systems with the Digital World

The Model Context Protocol is Anthropic’s open framework that revolutionizes how AI systems connect with external resources. Think of it as a universal translator that allows language models to communicate effortlessly with various tools, databases, and services without custom coding for each connection.

Before MCP came along, connecting AI to different services meant writing different codes for each integration—it was like having a different charger for every gadget you own. This made it hard to scale and be efficient. MCP creates a standardized way for AI to talk to different services, just like how modern ports make it easy for different devices to connect.

Let’s explore how to use MCP servers, using Claude AI as an example. Claude AI is Anthropic’s AI chatbot that incorporates various AI models. If you ask Claude AI a question about current events, it won’t be able to provide you with the latest data. Instead, it will display a message indicating that it doesn’t have the most up-to-date information. 

If we want to give people access to Claude AI online, we can use an MCP server. For web access, we could use the Brave Search API since it lets us use 2000 credits a month for free. But there’s a catch: we’d have to give them our credit card details, which most of us don’t like. I’d recommend using the Brave Search API for business. But since we don’t want to give out our credit card details, I found a better option: the Tavily Search API which gives us 1000 credits a month for free, and it doesn’t ask for any credit card details. 

But before MCP server, we need to download Claude ai for desktop which is available for both macOS and windows. Download it from a browser and then install it which should look like the below. 

If I ask for the latest news about Trump as on 8th April 2025, it fails to respond as shown below

For using Tavily search MCP server, sign up with Tavily in https://app.tavily.com/ and then for go to https://docs.tavily.com/documentation/mcp and then run the npx command which is necessary to run MCP in their server. Ensure that node is installed in your system. To check the node installation, type node –version in your terminal as given below. 

If not installed, then for macOS:

brew install node

For windows:

a. Go to https://nodejs.org/en 
b. Download the LTS version.
c. Run the installer and follow the steps.

Then Follow the below steps:

1. Run the below command

npx -y tavily-mcp@0.1.3

2. Go to your desktop app, go to settings, you will find 2 options General and Developer. Click on Developer and click on Edit Config button as shown below

3. Once Edit Config is clicked it would take you to the claude_desktop_config.json file where it is saved in your computer.

4. Open the config file in vscode (I assume you are familiar with vscode)

5. Under tvly-YOUR_API_KEY-here, paste your api key copied from your tavily account as shown below where you click on the copy icon and then paste into your config file like “TAVILY_API_KEY”: “abcdxyz” and then save the config file.

6. Then close the Claude ai app and again open it so that it gets restarted with the MCP server as given below

Now you may ask the same question to Claude and see the response.

can you help me with the latest news about trump on 8th April 2025?

Allow for this chat and then you will see the magic below

Here’s the magic of the MCP server. If you need any help, just let us know by email. 

MCP is a game-changer in bridging the gap between AI and digital resources. It’s like having a super-efficient way to connect all these different pieces together. As more and more people use it, we’re going to see an increasingly interconnected AI world, with smarter and more helpful apps popping up everywhere.