Model Context Protocol (MCP) Servers: A Complete Guide to Building AI-Powered Integrations
Author: Gupta Abhishek Premkumar
Published: September 2026
Reading Time: 13 minutes
Tags: AI, MCP, Model Context Protocol, LLM, Integration, Developer Tools
As Large Language Models (LLMs) become integral to modern software development, the need for standardized communication protocols between AI assistants and external tools has never been greater. The Model Context Protocol (MCP) emerges as a groundbreaking open standard that enables seamless, secure, and scalable integrations between AI models and external data sources, APIs, and services.
This article provides a comprehensive guide to understanding, building, and deploying MCP servers, empowering developers to extend AI capabilities beyond their inherent limitations.
The evolution of AI assistants has reached an inflection point. While Large Language Models possess remarkable reasoning and generation capabilities, they remain fundamentally limited by their training data cutoff and inability to interact with real-time systems. Enter the Model Context Protocol (MCP) — an open standard designed to bridge this gap by providing a universal interface for AI models to communicate with external tools, databases, and services.
Think of MCP as the "USB standard" for AI integrations. Just as USB standardized how peripherals connect to computers, MCP standardizes how AI assistants connect to the digital world.
The Model Context Protocol is an open, JSON-RPC-based protocol that defines how AI applications (clients) communicate with external services (servers) to access tools, resources, and contextual information. Developed with the goal of creating a universal standard for AI integrations, MCP enables:
| Benefit | Description |
|---|---|
| Interoperability | Works across different AI platforms and providers |
| Security | Built-in authentication and authorization mechanisms |
| Scalability | Designed for enterprise-grade deployments |
| Extensibility | Easy to add new tools and capabilities |
| Developer Experience | Simple APIs with comprehensive documentation |
The MCP architecture follows a client-server model with clear separation of concerns:
┌─────────────────┐ ┌─────────────────┐ ┌─────────────────┐
│ │ │ │ │ │
│ AI Client │◄───────►│ MCP Server │◄───────►│ External │
│ (LLM Host) │ JSON │ (Your Code) │ │ Services │
│ │ RPC │ │ │ (APIs, DBs) │
└─────────────────┘ └─────────────────┘ └─────────────────┘
Tools are the primary mechanism for AI models to perform actions. Each tool has:
// Example Tool Definition
{
name: "get_weather",
description: "Retrieves current weather information for a specified city",
inputSchema: {
type: "object",
properties: {
city: {
type: "string",
description: "The city name to get weather for"
},
units: {
type: "string",
enum: ["celsius", "fahrenheit"],
default: "celsius"
}
},
required: ["city"]
}
}
Resources provide read-only access to data sources. They are ideal for:
// Example Resource Definition
{
uri: "file:///config/settings.json",
name: "Application Settings",
description: "Current application configuration",
mimeType: "application/json"
}
Prompts are reusable templates that help AI models understand how to interact with specific domains or workflows.
// Example Prompt Definition
{
name: "code_review",
description: "Template for performing code reviews",
arguments: [
{
name: "language",
description: "Programming language of the code",
required: true
}
]
}
Let's build a practical MCP server that provides database query capabilities. We'll use TypeScript with the official MCP SDK.
mkdir mcp-database-server
cd mcp-database-server
npm init -y
npm install @modelcontextprotocol/sdk zod
npm install -D typescript @types/node ts-node
Create tsconfig.json:
{
"compilerOptions": {
"target": "ES2022",
"module": "Node16",
"moduleResolution": "Node16",
"outDir": "./dist",
"rootDir": "./src",
"strict": true,
"esModuleInterop": true,
"skipLibCheck": true,
"forceConsistentCasingInFileNames": true
},
"include": ["src/**/*"],
"exclude": ["node_modules", "dist"]
}
Create src/index.ts:
import { Server } from "@modelcontextprotocol/sdk/server/index.js";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
import {
CallToolRequestSchema,
ListToolsRequestSchema,
ListResourcesRequestSchema,
ReadResourceRequestSchema,
} from "@modelcontextprotocol/sdk/types.js";
import { z } from "zod";
// Define tool input schemas using Zod
const QueryDatabaseSchema = z.object({
query: z.string().describe("SQL query to execute"),
database: z.string().optional().describe("Target database name"),
});
const GetTableSchemaInput = z.object({
tableName: z.string().describe("Name of the table to describe"),
});
// Simulated database (replace with actual database connection)
const mockDatabase = {
users: [
{ id: 1, name: "Alice Johnson", email: "alice@example.com", role: "admin" },
{ id: 2, name: "Bob Smith", email: "bob@example.com", role: "user" },
{ id: 3, name: "Carol White", email: "carol@example.com", role: "user" },
],
products: [
{ id: 1, name: "Laptop", price: 999.99, stock: 50 },
{ id: 2, name: "Mouse", price: 29.99, stock: 200 },
{ id: 3, name: "Keyboard", price: 79.99, stock: 150 },
],
};
// Create the MCP server
const server = new Server(
{
name: "database-mcp-server",
version: "1.0.0",
},
{
capabilities: {
tools: {},
resources: {},
},
}
);
// Handle tool listing requests
server.setRequestHandler(ListToolsRequestSchema, async () => {
return {
tools: [
{
name: "query_database",
description:
"Execute a SQL-like query against the database. " +
"Supports SELECT statements with WHERE clauses.",
inputSchema: {
type: "object",
properties: {
query: {
type: "string",
description: "SQL query to execute (SELECT only)",
},
database: {
type: "string",
description: "Target database name (optional)",
},
},
required: ["query"],
},
},
{
name: "get_table_schema",
description:
"Retrieve the schema information for a specific table, " +
"including column names and data types.",
inputSchema: {
type: "object",
properties: {
tableName: {
type: "string",
description: "Name of the table to describe",
},
},
required: ["tableName"],
},
},
{
name: "list_tables",
description: "List all available tables in the database",
inputSchema: {
type: "object",
properties: {},
required: [],
},
},
],
};
});
// Handle tool execution requests
server.setRequestHandler(CallToolRequestSchema, async (request) => {
const { name, arguments: args } = request.params;
switch (name) {
case "query_database": {
const { query } = QueryDatabaseSchema.parse(args);
// Simple query parser (production should use proper SQL parser)
const tableMatch = query.toLowerCase().match(/from\s+(\w+)/);
if (!tableMatch) {
return {
content: [
{
type: "text",
text: "Error: Could not parse table name from query",
},
],
};
}
const tableName = tableMatch[1] as keyof typeof mockDatabase;
const data = mockDatabase[tableName];
if (!data) {
return {
content: [
{
type: "text",
text: `Error: Table '${tableName}' not found`,
},
],
};
}
return {
content: [
{
type: "text",
text: JSON.stringify(data, null, 2),
},
],
};
}
case "get_table_schema": {
const { tableName } = GetTableSchemaInput.parse(args);
const data = mockDatabase[tableName as keyof typeof mockDatabase];
if (!data || data.length === 0) {
return {
content: [
{
type: "text",
text: `Error: Table '${tableName}' not found or empty`,
},
],
};
}
const schema = Object.keys(data[0]).map((key) => ({
column: key,
type: typeof data[0][key as keyof (typeof data)[0]],
}));
return {
content: [
{
type: "text",
text: JSON.stringify(schema, null, 2),
},
],
};
}
case "list_tables": {
const tables = Object.keys(mockDatabase);
return {
content: [
{
type: "text",
text: JSON.stringify(
{
tables,
count: tables.length,
},
null,
2
),
},
],
};
}
default:
throw new Error(`Unknown tool: ${name}`);
}
});
// Handle resource listing
server.setRequestHandler(ListResourcesRequestSchema, async () => {
return {
resources: [
{
uri: "db://schema/overview",
name: "Database Schema Overview",
description: "Complete overview of all tables and their schemas",
mimeType: "application/json",
},
],
};
});
// Handle resource reading
server.setRequestHandler(ReadResourceRequestSchema, async (request) => {
const { uri } = request.params;
if (uri === "db://schema/overview") {
const overview = Object.entries(mockDatabase).map(([table, data]) => ({
table,
rowCount: data.length,
columns: data.length > 0 ? Object.keys(data[0]) : [],
}));
return {
contents: [
{
uri,
mimeType: "application/json",
text: JSON.stringify(overview, null, 2),
},
],
};
}
throw new Error(`Resource not found: ${uri}`);
});
// Start the server
async function main() {
const transport = new StdioServerTransport();
await server.connect(transport);
console.error("Database MCP Server running on stdio");
}
main().catch(console.error);
Create mcp-config.json for client configuration:
{
"mcpServers": {
"database": {
"command": "node",
"args": ["dist/index.js"],
"cwd": "/path/to/mcp-database-server"
}
}
}
npx tsc
js
// Implement authentication for sensitive operations
const authenticateRequest = async (token: string): Promise<boolean> => {
// Validate JWT or API key
const isValid = await validateToken(token);
if (!isValid) {
throw new Error("Authentication failed");
}
return true;
};
// Wrap tool handlers with authentication
const withAuth = (handler: Function) => async (request: any) => {
const token = request.params.meta?.authToken;
await authenticateRequest(token);
return handler(request);
};
js
import { RateLimiter } from "limiter";
const limiter = new RateLimiter({
tokensPerInterval: 100,
interval: "minute",
});
const withRateLimit = (handler: Function) => async (request: any) => {
const remainingRequests = await limiter.removeTokens(1);
if (remainingRequests < 0) {
throw new Error("Rate limit exceeded. Please try again later.");
}
return handler(request);
};
python
import NodeCache from "node-cache";
const cache = new NodeCache({ stdTTL: 300 }); // 5 minute cache
const withCache = (cacheKey: string, handler: Function) => async (request: any) => {
const cached = cache.get(cacheKey);
if (cached) {
return cached;
}
const result = await handler(request);
cache.set(cacheKey, result);
return result;
};
python
import winston from "winston";
const logger = winston.createLogger({
level: "info",
format: winston.format.json(),
transports: [
new winston.transports.File({ filename: "error.log", level: "error" }),
new winston.transports.File({ filename: "combined.log" }),
],
});
const withErrorHandling = (handler: Function) => async (request: any) => {
try {
const result = await handler(request);
logger.info("Request processed successfully", {
tool: request.params.name,
});
return result;
} catch (error) {
logger.error("Request failed", {
tool: request.params.name,
error: error.message,
});
throw error;
}
};
Always validate and sanitize inputs using schemas:
import { z } from "zod";
const SafeQuerySchema = z.object({
query: z.string()
.max(1000)
.refine(
(q) => !q.toLowerCase().includes("drop"),
"DROP statements are not allowed"
)
.refine(
(q) => !q.toLowerCase().includes("delete"),
"DELETE statements are not allowed"
),
});
js
const auditLog = async (action: string, user: string, details: object) => {
await db.insert("audit_logs", {
timestamp: new Date().toISOString(),
action,
user,
details: JSON.stringify(details),
});
};
Build an MCP server that connects AI assistants to internal documentation:
search_docs, get_document, list_categories
Create an MCP server for infrastructure management:
deploy_service, scale_pods, get_logs, rollback
Connect AI to CRM and ticketing systems:
create_ticket, update_status, get_customer_history
Build an MCP server for financial reporting:
generate_report, calculate_metrics, forecast
import { Pool } from "pg";
const pool = new Pool({
max: 20,
idleTimeoutMillis: 30000,
connectionTimeoutMillis: 2000,
});
// Reuse connections across requests
const query = async (sql: string, params: any[]) => {
const client = await pool.connect();
try {
return await client.query(sql, params);
} finally {
client.release();
}
};
For large datasets, implement streaming:
const streamResults = async function* (query: string) {
const cursor = db.query(query).cursor(100);
for await (const batch of cursor) {
yield batch;
}
};
js
const parallelTools = async (requests: ToolRequest[]) => {
const results = await Promise.allSettled(
requests.map((req) => executeToolHandler(req))
);
return results;
};
python
import zlib from "zlib";
const compressResponse = (data: string): Buffer => {
return zlib.gzipSync(data);
};
The Model Context Protocol is positioned to become the standard for AI integrations. Upcoming developments include:
Real-time data streaming for live dashboards and monitoring applications.
Support for image, audio, and video processing tools.
Interconnected MCP servers sharing capabilities across organizations.
Native telemetry, tracing, and monitoring features.
Industry-standard authentication and authorization patterns.
The Model Context Protocol represents a paradigm shift in how we build AI integrations. By providing a standardized, secure, and scalable approach to connecting AI models with external systems, MCP enables developers to create powerful, context-aware applications that bridge the gap between AI capabilities and real-world data.
As AI continues to transform software development, MCP servers will become essential components of modern application architectures. By mastering MCP today, you position yourself at the forefront of the AI integration revolution.
Gupta Abhishek Premkumar is a software professional dedicated to advancing AI-powered innovation. With expertise in AI integration, distributed architectures, and enterprise software systems, he builds scalable solutions that bridge the gap between emerging technologies and impactful business outcomes
© 2026 Abhishek Gupta. This article is licensed under Creative Commons Attribution 4.0 International License.
Keywords: MCP, Model Context Protocol, AI Integration, LLM, Large Language Models, TypeScript, Node.js, API Development, AI Tools, Enterprise AI, Developer Tools, Open Source