Model Context Protocol (MCP) – Access For AI Agents

The Model Context Protocol (MCP) is an open standard developed by Anthropic that enables AI agents to securely connect with external data sources and tools. Think of MCP as a universal “plugin system” for AI – it provides a standardized way for AI models like Claude to interact with your business data, APIs, and applications without requiring direct database access.

Why MCP Matters for AI Agents

Traditional AI assistants operate in isolation, limited to their training data and unable to access real-time information. MCP changes this paradigm by:

Providing real-time data access to your business systems
Enabling action execution within your applications
Maintaining security through controlled, predefined interfaces
Ensuring data freshness without manual data exports

How to Expose Data to AI Agents Using MCP

The Traditional Problem

Without MCP, AI agents face significant limitations:

# What happens WITHOUT MCP
user_question = "What's the status of my recent order?"
ai_response = "I can help you with general questions about order status, but I don't have access to your actual order data. You'll need to check your account portal directly."

# What happens WITHOUT MCP
user_question = "What's the status of my recent order?"
ai_response = "I can help you with general questions about order status, but I don't have access to your actual order data. You'll need to check your account portal directly."

The MCP Solution

With MCP, you create secure “tools” that AI agents can use:

# What happens WITH MCP
user_question = "What's the status of my recent order?"
ai_thinking = "The user is asking about order status. I have an MCP tool 'orders.get_by_customer' that can retrieve this information. I'll call it with their email."
ai_action = mcp_tool.execute("orders.get_by_customer", {"email": "user@company.com"})
ai_response = f"Your most recent order #{
  ai_action.order_id} is currently {
  ai_action.status} and expected to ship on {
  ai_action.estimated_ship_date}."

# What happens WITH MCP
user_question = "What's the status of my recent order?"
ai_thinking = "The user is asking about order status. I have an MCP tool 'orders.get_by_customer' that can retrieve this information. I'll call it with their email."
ai_action = mcp_tool.execute("orders.get_by_customer", {"email": "user@company.com"})
ai_response = f"Your most recent order #{
  ai_action.order_id} is currently {
  ai_action.status} and expected to ship on {
  ai_action.estimated_ship_date}."

Practical Implementation Example

Let’s create a real-world scenario for an e-commerce support agent:

Step 1: Define Your Data Access Tools

JSON

{
  "mcp_tools": [
    {
      "name": "customer_lookup",
      "description": "Retrieve customer information by email",
      "parameters": {
        "email": "string"
      }
    },
    {
      "name": "order_status",
      "description": "Get current status and details of customer orders",
      "parameters": {
        "customer_id": "string"
      }
    },
    {
      "name": "create_support_ticket",
      "description": "Create a new customer support ticket",
      "parameters": {
        "customer_id": "string",
        "issue": "string",
        "priority": "string"
      }
    }
  ]
}

{
  "mcp_tools": [
    {
      "name": "customer_lookup",
      "description": "Retrieve customer information by email",
      "parameters": {
        "email": "string"
      }
    },
    {
      "name": "order_status",
      "description": "Get current status and details of customer orders",
      "parameters": {
        "customer_id": "string"
      }
    },
    {
      "name": "create_support_ticket",
      "description": "Create a new customer support ticket",
      "parameters": {
        "customer_id": "string",
        "issue": "string",
        "priority": "string"
      }
    }
  ]
}

Step 2: Sample Conversation Flow

User: "Can you check why my order #12345 is delayed?"

AI Agent Process:
1. Identifies need for customer data
2. Uses customer_lookup tool with user's email
3. Retrieves customer ID and order history
4. Uses order_status tool with customer ID
5. Analyzes order data and shipping information
6. Provides specific, accurate response about delay

Response: "I can see your order #12345 is delayed due to inventory issues. 
We've expedited the restocking and it should ship by tomorrow. 
I've created a support ticket (REF: ST-789) to follow up personally."

User: "Can you check why my order #12345 is delayed?"

AI Agent Process:
1. Identifies need for customer data
2. Uses customer_lookup tool with user's email
3. Retrieves customer ID and order history
4. Uses order_status tool with customer ID
5. Analyzes order data and shipping information
6. Provides specific, accurate response about delay

Response: "I can see your order #12345 is delayed due to inventory issues. 
We've expedited the restocking and it should ship by tomorrow. 
I've created a support ticket (REF: ST-789) to follow up personally."

Xano MCP Builder: The Bridge Between Your Data and AI

I already wrote about the no-code platform Xano in another article: Xano – No-Code Platform , so I won’t elaborate on it here, but it is a platform for building backends.

What is Xano MCP Builder?

Xano MCP Builder is a powerful feature that transforms your existing backend functions into MCP-compatible tools for AI agents. It implements the Model Context Protocol while maintaining the MCP (Model-Controller-Presenter) architectural pattern for clean code organization.

How to Use Xano MCP Builder

Step 1: Create Your MCP Builder Functions

In your Xano workspace, create functions that will serve as your MCP tools:

Function: customer.get_profile
Stage: Controller
Input: email (string)
Logic: 
  - Validate email format
  - Query customer table
  - Join with recent orders
  - Return customer object with order history

Function: orders.get_by_customer
Stage: Controller  
Input: customer_id (string), status (optional)
Logic:
  - Verify customer exists
  - Query orders with optional status filter
  - Include shipping and product details
  - Return formatted order data

Function: support.create_ticket
Stage: Controller
Input: customer_id, subject, description, priority
Logic:
  - Create support ticket record
  - Notify support team via webhook
  - Return ticket reference number

Function: customer.get_profile
Stage: Controller
Input: email (string)
Logic: 
  - Validate email format
  - Query customer table
  - Join with recent orders
  - Return customer object with order history

Function: orders.get_by_customer
Stage: Controller  
Input: customer_id (string), status (optional)
Logic:
  - Verify customer exists
  - Query orders with optional status filter
  - Include shipping and product details
  - Return formatted order data

Function: support.create_ticket
Stage: Controller
Input: customer_id, subject, description, priority
Logic:
  - Create support ticket record
  - Notify support team via webhook
  - Return ticket reference number

Step 2: Configure MCP in Xano

Navigate to the MCP section in your Xano dashboard
Select which functions to expose as MCP tools
Configure authentication and rate limiting
Generate your MCP server endpoint

Step 3: Connect to Your AI Agent

TypeScript

// Example connection configuration
const mcpConfig = {
  server: "https://your-workspace.xano.io/api/mcp",
  tools: [
    "customer.get_profile",
    "orders.get_by_customer", 
    "support.create_ticket"
  ],
  authentication: {
    type: "bearer",
    token: "your-api-token"
  }
};

// Example connection configuration
const mcpConfig = {
  server: "https://your-workspace.xano.io/api/mcp",
  tools: [
    "customer.get_profile",
    "orders.get_by_customer", 
    "support.create_ticket"
  ],
  authentication: {
    type: "bearer",
    token: "your-api-token"
  }
};

Complete Real-World Example: E-commerce Support Agent

Scenario: A customer needs help with a delayed order and wants to understand their options.

AI Agent Conversation:

Customer: "My order #56789 hasn't arrived and it's 3 days late. 
           Can you check what's happening and help me cancel if it's not shipping soon?"

AI Agent Process:
1. Identifies need for order information
2. Calls Xano MCP Builder: orders.get_by_customer with order number
3. Receives real-time order data:
   {
     "order_id": "56789",
     "status": "processing",
     "customer_id": "cust_123",
     "items": [...],
     "delay_reason": "weather_delay",
     "estimated_ship_date": "2024-01-15"
   }

4. Calls Xano MCP Builder: customer.get_profile for context
5. Analyzes the situation and provides informed response

AI Response: "I can see your order #56789 is delayed due to severe weather conditions 
affecting our shipping carrier. It's now expected to ship on January 15th. 
Since this is a significant delay, I can offer you:
- 15% discount on this order
- Priority shipping once it dispatches
- Option to cancel with full refund

Would you like me to apply the discount, or would you prefer to cancel the order?"

Customer: "My order #56789 hasn't arrived and it's 3 days late. 
           Can you check what's happening and help me cancel if it's not shipping soon?"

AI Agent Process:
1. Identifies need for order information
2. Calls Xano MCP Builder: orders.get_by_customer with order number
3. Receives real-time order data:
   {
     "order_id": "56789",
     "status": "processing",
     "customer_id": "cust_123",
     "items": [...],
     "delay_reason": "weather_delay",
     "estimated_ship_date": "2024-01-15"
   }

4. Calls Xano MCP Builder: customer.get_profile for context
5. Analyzes the situation and provides informed response

AI Response: "I can see your order #56789 is delayed due to severe weather conditions 
affecting our shipping carrier. It's now expected to ship on January 15th. 
Since this is a significant delay, I can offer you:
- 15% discount on this order
- Priority shipping once it dispatches
- Option to cancel with full refund

Would you like me to apply the discount, or would you prefer to cancel the order?"

Security and Best Practices

Data Security:

MCP Builders act as a security layer – AI only sees what you expose
Implement authentication in your Xano functions
Use field-level security to exclude sensitive data
Add rate limiting and monitoring

Function Design Tips:

Keep MCP Builders focused and single-purpose
Implement proper error handling and validation
Use the MCP stages appropriately:
- Model Stage: Data validation and preparation
- Controller Stage: Business logic and data processing
- Presenter Stage: Response formatting and sanitization

Benefits Summary

For AI Agents:

Real-time access to current business data
Ability to perform actions within your systems
Context-aware responses based on actual data
Reduced hallucinations with factual grounding

For Developers:

Reuse existing backend logic
Maintain security and control
Standardized interface for AI integration
Scalable architecture with clear separation of concerns

For Business:

Enhanced customer experience with informed AI support
Reduced operational overhead
Faster resolution times
Consistent data access patterns

Conclusion

The combination of Model Context Protocol and Xano MCP Builder creates a powerful foundation for building intelligent, data-aware AI agents. By exposing your business data through secure, controlled interfaces, you enable AI to provide accurate, context-rich assistance while maintaining full control over your data and systems.

This approach transforms AI from a generic chat companion into a truly intelligent business assistant that can access, analyze, and act upon your real-time business data safely and effectively.