10 Minutes News for Hoteliers 10 Minutes News for Hoteliers
  • Top News
    • Hotel Tech
    • Hotel Chains
    • Topics
    • Airbnb news
    • AI in Hospitality News
    • Marriott news
    • Booking.com news
    • OTA News
    • UCP news
    • PMS news
  • The Columns
  • Posts
    • Hotel Marketing
    • Revenue Management
    • CSR and Sustainability
    • Events
    • Hotel Openings
    • Hotel Operations
    • Human Resources
    • Innovation
    • Market Trends
    • Mergers & Acquisitions
    • Regulatory and Legal Affairs
  • 👉 Sign-up
  • 🌎 Languages
    • 🇫🇷 French
    • 🇮🇹 Italian
    • 🇪🇸 Spain
  • 📰 More
    • Largest Hotel Brands by Traffic
    • Hotel Brands of the World
    • OTAs of the World
    • Most read Articles
  • About Us
    • Editorial Policy
    • Contact Us
    • As Seen on 10 Minutes
10 Minutes News for Hoteliers 10 Minutes News for Hoteliers
  • Top News
    • Hotel Tech
    • Hotel Chains
    • Topics
    • Airbnb news
    • AI in Hospitality News
    • Marriott news
    • Booking.com news
    • OTA News
    • UCP news
    • PMS news
  • The Columns
  • Posts
    • Hotel Marketing
    • Revenue Management
    • CSR and Sustainability
    • Events
    • Hotel Openings
    • Hotel Operations
    • Human Resources
    • Innovation
    • Market Trends
    • Mergers & Acquisitions
    • Regulatory and Legal Affairs
  • 👉 Sign-up
  • 🌎 Languages
    • 🇫🇷 French
    • 🇮🇹 Italian
    • 🇪🇸 Spain
  • 📰 More
    • Largest Hotel Brands by Traffic
    • Hotel Brands of the World
    • OTAs of the World
    • Most read Articles
  • About Us
    • Editorial Policy
    • Contact Us
    • As Seen on 10 Minutes

A Beginner’s Guide to the Model Context Protocol (MCP):

  • Tony Loeb
  • 27 January 2026
  • 7 minute read
Total
0
Shares
0
0
0

source: https://x.com/Meer_AIIT/status/2015091824851836979

This isn’t a clickbait article that skims the surface. I’ve personally learned MCP from the ground up.

Worked through multiple courses.

Studied the official documentation.

Built servers myself.

I’ll link to all my sources so you can dive deeper but my goal here is to save you time by distilling everything into one comprehensive guide.

By the end of this article you’ll understand the complete fundamentals.

  • What MCP servers actually are.
  • How to use them.
  • Why they matter.

And how to build your own… both with code and without it.

What is MCP?

The Model Context Protocol (MCP) is an open standard that enables AI applications to connect with external tools and data sources in a standardized way.

Think of it as a universal language that lets AI models talk to the outside world.

Before MCP if you wanted Claude or another AI to access your Google Drive you’d need custom code.

Want it to also read from Slack?

More custom code.

Each integration required its own unique solution.

MCP changes this by providing a single consistent protocol.

Build once. Connect to everything.

image source: MCP: Build Rich-Context AI Apps with Anthropic

The USB-C Analogy:

Here’s a helpful way to think about it.

Worlds of Flavor, pastrami with horseradish, and Cambodian food
Trending
Worlds of Flavor, pastrami with horseradish, and Cambodian food

MCP is like USB-C for AI applications.

Remember when every device had its own charger?

Phones. Tablets. Laptops. All different cables.

USB-C standardized this mess.

Now one cable works for almost everything.

MCP does the same thing for AI integrations.

Instead of building custom connectors for every tool or database… developers implement the MCP standard once.

Their AI can then connect to any MCP-compatible service through a single unified interface.

file systems databases APIs enterprise tools

All through one protocol.

AI generated image

Who Introduced MCP:?

MCP was announced by Anthropic in November 2024 as an open standard for connecting AI assistants to data systems.

Content repositories. Business management tools. Development environments.

The protocol emerged from a practical problem Anthropic’s team kept encountering.

The “N×M integration problem.”

As the number of AI applications (N) grows and the variety of tools and data sources (M) increases… the complexity of creating custom integrations for each combination becomes unmanageable.

The protocol was released with SDKs in Python, TypeScript, C# and Java.

Anthropic open-sourced everything from day one.

Since launching in November 2024 adoption has been rapid.

The community has built thousands of MCP servers.

SDKs are available for all major programming languages.

The industry has adopted MCP as the de-facto standard for connecting agents to tools and data.

Following its announcement the protocol was adopted by major AI providers including OpenAI and Google DeepMind.

Core Concepts & Architecture

MCP uses a straightforward client-server architecture.

image source: MCP: Build Rich-Context AI Apps with Anthropic

Understanding this structure is key to working with the protocol effectively.

The Three Main Components:

Hosts: are the AI applications you interact with.

Claude Desktop, An IDE with AI features, A custom chatbot.

The host manages the overall experience and coordinates between the AI model and external services.

Clients: live inside hosts and handle communication with MCP servers.

Think of them as translators that speak the MCP language on behalf of the host application.

Servers: are lightweight programs that expose specific capabilities.

A server might provide access to your file system.

Connect to a database.

Or integrate with a business tool like Slack or GitHub.

The Three Core Primitives:

MCP servers can expose three types of capabilities.

Tools: are functions the AI can call to perform actions.

Searching files. Querying a database. Sending a message. Making an API request.

When Claude needs to “do something” it uses a tool.

Resources: provide data and content the AI can read as context.

Documents. Database records. Configuration files.

Resources give the AI information to work with.

Prompts: are reusable templates and workflows.

They help standardize how the AI approaches certain tasks ensuring consistent and high-quality outputs.

How Communication Works

When you ask Claude to “find all my project files modified this week” here’s what happens behind the scenes:

  1. Claude connects to your file system MCP server via the host and client
  2. The client requests the available tools from the server
  3. Claude identifies the right tool and calls it with appropriate parameters
  4. The server executes the request and returns results
  5. Claude processes the results and responds to you

All of this happens through a standardized message format.

Typically using JSON over stdio for local servers or HTTP for remote servers.

Why MCP Matters

Problems It Solves

Eliminates integration fragmentation.

Without MCP every combination of AI app and external tool requires custom code.

For 10 AI apps connecting to 20 tools that’s potentially 200 unique integrations.

MCP reduces this to 10 + 20 = 30 implementations.

Enables true AI agents.

AI becomes dramatically more useful when it can take action… not just discuss things.

MCP provides the secure standardized way for AI to interact with real-world systems.

Maintains security and control.

Every action requires your approval.

The host controls which servers connect.

You approve each tool invocation.

Your data stays protected while still being accessible to AI.

Reduces vendor lock-in.

Because MCP is an open standard your integrations work across different AI providers.

Switch from one model to another without rebuilding your tool connections.

Real Benefits

For developers MCP means building an integration once and having it work everywhere.

For organizations it means AI assistants that can actually access internal systems and data.

For users it means AI that’s genuinely helpful.

Not just conversational but capable.

Getting Started

No-Code Approach: Desktop Extensions

The easiest way to experience MCP is through Claude Desktop’s extension system.

Desktop extensions provide a streamlined way to install and manage local MCP servers through single-click installable packages.

Instead of manually configuring JSON files and managing dependencies… you can now install local MCP servers as easily as browser extensions.

Here’s how to get started:

  1. Download and install Claude Desktop from claude.ai
  2. Open Claude Desktop and go to Settings > Extensions
  3. Browse the extension directory for Anthropic-reviewed tools
  4. Click “Install” on any extension that interests you
  5. Configure any required settings like API keys through the interface

That’s it.

The extension handles all the technical setup automatically.

Low-Code Approach: Pre-Built Servers

If you’re comfortable with a terminal you can install pre-built MCP servers manually.

This gives you more options than the extension directory.

First ensure Node.js is installed on your system.

Open a terminal and run:

bash

node -v

If you see a version number you’re ready.

If not download Node.js from nodejs. org.

Next configure Claude Desktop to use a server.

Open the Claude Desktop configuration file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.jsonWindows: %APPDATA%\Claude\claude_desktop_config.json

Add a server configuration.

Here’s an example for the filesystem server:

json

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-filesystem",
        "/Users/yourname/Desktop",
        "/Users/yourname/Documents"
      ]
    }
  }
}

Replace the paths with directories you want Claude to access.

Save the file and restart Claude Desktop.

Developer Approach: Building Your Own

For full control you can build custom MCP servers using the official SDKs.

This requires programming knowledge but offers unlimited flexibility.

We’ll cover a basic example in the next section.

Simple Implementation Example

Let’s build a minimal MCP server in Python.

This server will expose one tool: a calculator that evaluates mathematical expressions.

Setup

First create a project directory and install the MCP SDK:

bash

# Create and enter project directory
mkdir my-mcp-server
cd my-mcp-server

# Create virtual environment (recommended)
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install the MCP SDK
pip install mcp

The Server Code

Create a file called calculator_server.py:

python

from mcp.server.fastmcp import FastMCP

# Initialize the MCP server with a name
mcp = FastMCP("Calculator Server")

# Define a tool using the @mcp.tool() decorator
# The docstring becomes the tool's description for the AI
@mcp.tool()
def calculate(expression: str) -> str:
    """
    Evaluate a mathematical expression and return the result.
    
    Args:
        expression: A math expression like "2 + 2" or "10 * 5 / 2"
    
    Returns:
        The calculated result as a string
    """
    try:
        # eval() is used here for simplicity - in production,
        # use a safer math parser like numexpr or simpleeval
        result = eval(expression)
        return f"Result: {result}"
    except Exception as e:
        return f"Error: {str(e)}"

# Run the server using stdio transport (for local use)
if __name__ == "__main__":
    mcp.run(transport="stdio")

What’s Happening Here

The FastMCP class handles all the protocol details for you.

When you decorate a function with

@mcp

.tool() it automatically becomes available as an MCP tool.

The function’s type hints (expression: str) and docstring are parsed to create the tool’s schema.

This tells AI models what parameters the tool expects and what it does.

Testing Your Server

You can test your server using the MCP CLI:

bash

mcp dev calculator_server.py

This opens a local dashboard where you can interact with your tools directly.

Connecting to Claude Desktop

Add your server to Claude Desktop’s configuration:

json

{
  "mcpServers": {
    "calculator": {
      "command": "python",
      "args": ["/full/path/to/calculator_server.py"]
    }
  }
}

Restart Claude Desktop and you can now ask Claude to calculate things.

It will use your server to perform the actual math.

Common Use Cases

MCP opens up a wide range of practical applications.

File and document management

Let AI organize your downloads. Search through documents. Create and edit files on your system.

Database access

Connect AI to PostgreSQL, MySQL or other databases.

Ask questions about your data in plain English and get actual query results.

Development workflows

Integrate with GitHub for managing issues and pull requests.

Connect to your IDE for AI-assisted coding.

Business tool integration

Connect to Slack. Google Drive. Notion. Salesforce.

Build AI assistants that understand your organization’s context.

Custom automation

Build servers that wrap your own APIs and internal tools.

Create AI-powered workflows specific to your needs.

Research and analysis

Give AI access to data sources and web scraping tools.

Let it gather and process information for you.

Key Takeaways

MCP represents a fundamental shift in how AI applications connect to the world.

Instead of isolated chatbots we now have AI that can actually participate in our digital workflows.

The protocol is open. Standardized. Already widely adopted.

Whether you’re a casual user installing extensions or a developer building custom integrations… MCP provides a path forward.

Security remains in your hands.

Every connection is controlled.

Every action requires approval.

Your data stays protected. Here are links of each sources from I’ve learned entire MCP Servers: 1.

https://www.deeplearning.ai/short-courses/mcp-build-rich-context-ai-apps-with-anthropic

2.

https://cloud.google.com/discover/what-is-model-context-protocol

3.

https://modelcontextprotocol.io/docs/getting-started/intro

If you want to stay updated with the latest AI news and tools and research paper breakdowns… subscribe to my free newsletter:

https://www.theainight.com
author avatar
Tony Loeb Co-Founder
Co-fondateur de 10 Minutes et de Pragmatik, Tony Loeb cumule plus de vingt ans d'expérience dans la hotel tech, du web au marketing et à la distribution hôtelière.
See Full Bio
social network icon
Total
0
Shares
Share 0
Tweet 0
Pin it 0
Tony Loeb

Tony Loeb est un entrepreneur de la hotel tech qui compte plus de vingt ans d'expérience dans l'industrie hôtelière. Webmaster à ses débuts, il est devenu directeur technique et a piloté une équipe d'une trentaine de personnes dédiée à la création de sites, au SEO et au SEA pour les hôtels. Il a ensuite accompagné des hôtels de luxe et des chaînes à l'international, avant de cofonder un CRM et une solution d'email marketing dédiés à l'hôtellerie, utilisés par des milliers d'hôtels indépendants dans le monde. Aujourd'hui, il est co-fondateur de 10 Minutes (10 minutes pour un hôtelier), le média de référence de l'hôtellerie européenne, qui réunit newsletter, podcasts et vidéos pour atteindre plus de 350 000 décideurs hôteliers. Il y anime les podcasts « 10 min pour un hôtelier » et « Inno'Tourism ». Il est également co-fondateur de Pragmatik, une agence marketing qui aide les entreprises de la hotel tech à transformer des contenus clairs, sans blabla, en leads qualifiés. Pragmatique par nature, Tony décrypte ce qui compte en marketing hôtelier, distribution, expérience client et technologie pour aider les hôteliers à décider mieux et plus vite. Basé en région parisienne.

Previous Article

Aubamar Palma Resort 5* To Open Its Doors In March 2026

  • Automatic
  • 27 January 2026
View Post
Next Article

Ian Schrager Company And Highgate Join Forces And Enter Into A New Partnership To Expand The PUBLIC Brand Globally

  • Automatic
  • 27 January 2026
View Post
You should like too
LinkedIn logo centered above the text 'Posted on LinkedIn' on a blue background, indicating social post status.
View Post
  • Innovation

GitHub Copilot moves to usage-based billing on June 1, with $39 plan now metered by token

  • 21 November 2026
View Post
  • Innovation

Martin Soler questions whether hotels should block AI agents from accessing their inventory

  • 21 November 2026
View Post
  • Innovation

Beta rollout on Sabre’s MCP Server introduces hotel rate monitoring, automated disruption alerts and conversational booking

  • hospitalitytech.com
  • 21 November 2026
View Post
  • Innovation

Saudi hotel operator to deploy dailypoint’s Central Profile architecture to unify guest data across PMS, POS and digital touchpoints

  • hospitalitytech.com
  • 21 November 2026
View Post
  • Innovation

BirchStreet’s generally available AI inventory tool measures liquor bottle fill levels, cutting cycle times by up to 50%

  • hospitalitytech.com
  • 21 November 2026
View Post
  • Innovation

Founding family member succeeds David Duncan, plans third-party management expansion and AI deployment across property and corporate workflows

  • hospitalitytech.com
  • 21 November 2026
View Post
  • Innovation

Automated tools scan decompiled code in seconds, exposing annual hospitality audits’ limitations in detecting hardcoded secrets

  • hospitalitytech.com
  • 21 November 2026
Rosa, Sizzle's Synth Human, greeting a guest in the Sheraton Universal Hotel lobby
View Post
  • Hotel Operations
  • Innovation

Sizzle brings Rosa, a Synth Human AI concierge, to Sheraton Universal Hotel guests

  • Press Releases
  • 20 November 2026
Downloads
  • The AI Trip Planning Market 2026

    View Post
  • The OTA Market, Finally Mapped

    View Post
  • The Hotel Internet Is Controlled by a Handful of Brands

    View Post
  • Global hotel supply by region: hotels, rooms and the chain vs independent split (2027 edition) - chart by 10 Minutes

    Global Hotel Supply 2027: China Passes 370,000 Hotels, and Three in Four Worldwide Are Independent

    View Post
Join our 300,000+ Readers!
Most Read
  • Katanox logo, financial infrastructure for hospitality
    Katanox and Mobi.AI partner to power payments for agentic hotel commerce
    • 18 November 2026
  • Guest room with a winter view at The Josie Hotel, RED Mountain, Autograph Collection
    RED Mountain Resort owners to acquire The Josie, an Autograph Collection hotel in Rossland, B.C.
    • 18 November 2026
  • Portrait of Beau Kent, Chief Operating Officer of Tammen Hospitality Partners
    Tammen Hospitality Partners appoints Beau Kent as Chief Operating Officer
    • 18 November 2026
  • Design rendering of a voco hotel restaurant by Vertti & Co
    Vertti & Co to design the conversion of 11 European hotels to IHG’s voco, Holiday Inn and Garner brands
    • 19 November 2026
  • Should Hotels Let AI Agents Book?
    • 19 November 2026
Sponsors
  • Monitor your blind spots: Is AI getting your hotel right?
  • Monitor your blind spots: Is AI getting your hotel right? 
  • Shiji spent a year testing AI on its own people before offering it to a single hotel
Top News
  • LinkedIn logo centered above the text 'Posted on LinkedIn' on a blue background, indicating social post status.
    Thibault Catala | Catala Consulting | 59 likes · 24 comments | LinkedIn
    • 22 November 2026
  • AHLA’s First GM Community Ambassador Award Recipient Brad Everett Hosts Tree Lighting Drawing Up to 10,000 in Manhattan, Kansas
    • 22 November 2026
  • How to Get Your Hotel Data Ready for AI
    • 21 November 2026
  • image
    Gstaad: Dona Bertarelli has reportedly sold the five-star Park Gstaad (75 keys on reopening) to Greek shipowner George Procopiou for CHF 350-400m (ca. CHF 4.7-5.3m per key). (Gstaad Confidential)… | Eliott Camb
    • 21 November 2026
  • image
    AI can predict hotel room pricing Depending on the AI model Check this out: For context, Timestral is a time series prediction model – Claude and ChatGPT predict the next word. Timestral predicts… | Assem Chammah | 52 comments
    • 21 November 2026
Sponsored Posts
  • Monitor your blind spots: Is AI getting your hotel right?

    View Post
  • Monitor your blind spots: Is AI getting your hotel right? 

    View Post
  • Shiji spent a year testing AI on its own people before offering it to a single hotel

    View Post
Contact informations

[email protected]

Advertise with us
Contact Tony to learn more: [email protected]
Press release
[email protected]
10 Minutes News for Hoteliers 10 Minutes News for Hoteliers
  • Most important news in hospitality
  • Latest news about Booking.com
  • Latest news about Marriott
  • Latest news about Hilton
  • Latest news about Airbnb
  • Largest Hotel Brands by Traffic
  • AI in Hospitality News
  • Expedia News Hub
  • Revenue Management
  • Property Management Systems (PMS) News
  • Latest news about Siteminder
  • OTA News for Hotels
  • Hotel Marketing News
  • Most read Articles
  • The Complete OTAs of the World List
  • The Hotel Brands of the World
  • Hotel Openings
  • Human Resources in Hospitality
  • Mergers & Acquisitions
  • Regulatory and Legal Affairs
Discover the best of international hotel news. Categorized, and sign-up to the newsletter

Input your search keywords and press Enter.

New

Booking and Expedia are just 2 of 192.

See the full map of hotel distribution.

See the 192 channels→
The OTA Database - 192 hotel distribution channels Free preview
New

There are 192 OTAs.

The complete map of hotel distribution.

Explore the database→

Free