What is MCP?

The Model Context Protocol (MCP) is an open standard for connecting AI systems to external tools, services, and data sources.

Think of MCP as:

USB-C for AI applications.

Just as USB-C allows laptops, phones, and accessories to communicate using a common interface, MCP allows AI models and applications to communicate with external capabilities using a common protocol.

Continue Exploring MCP

Karyam uses MCP as a first-class platform capability, but this guide only covers the concepts most relevant to building AI systems within Karyam.

To learn more about the protocol itself, including specifications, SDKs, transports, and implementation details, visit the official Model Context Protocol website:

🌐 https://modelcontextprotocol.io/

Why MCP Exists

Modern AI systems need access to tools.

They need to:

Create tickets
Query databases
Access APIs
Manage repositories
Read files
Trigger workflows
Interact with enterprise systems

Historically, every one of these integrations required custom development.

Example:

AI Agent
    ↓
Custom GitHub Integration

AI Agent
    ↓
Custom Jira Integration

AI Agent
    ↓
Custom Slack Integration

AI Agent
    ↓
Custom Database Integration

This quickly becomes difficult to maintain and scale.

The MCP Approach

MCP introduces a standardized interface between AI systems and external capabilities.

Instead of building custom integrations:

AI Agent
    ↓
MCP Client
    ↓
MCP Server
    ↓
External Tool

Any AI application that understands MCP can communicate with any MCP-compatible server.

MCP Architecture

MCP follows a client-server architecture.

AI Application
        ↓
MCP Client
        ↓
MCP Server
        ↓
External System

Examples:

Karyam Agent
        ↓
GitHub MCP Server
        ↓
GitHub API

Karyam Agent
        ↓
Jira MCP Server
        ↓
Jira Cloud

Karyam Agent
        ↓
PostgreSQL MCP Server
        ↓
Database

MCP Components

MCP defines several core concepts.

Tools

Tools are executable capabilities exposed by an MCP Server.

Examples:

Create Issue
Execute Query
Send Message
Create Pull Request

Example:

GitHub MCP Server
├── Create Issue
├── Create Pull Request
├── List Repositories
└── Search Code

Resources

Resources provide access to external information.

Examples:

Files
Documents
Database records
Configuration data

Resources allow AI systems to consume information without directly modifying it.

Prompts

Prompts are reusable instructions provided by MCP Servers.

They help standardize interactions and workflows across tools.

Dynamic Discovery

One of MCP’s most powerful features is automatic discovery.

When an AI application connects to an MCP Server:

Connect Server
        ↓
Discover Tools
        ↓
Discover Resources
        ↓
Discover Prompts

No manual tool definitions are required.

Example Discovery Process

GitHub MCP Server Connected
        ↓
Discovered Tools:
- Create Issue
- Create Pull Request
- List Repositories
- Search Code
- Get Workflow Runs

The AI system can immediately begin using these capabilities.

Transport Mechanisms

MCP supports multiple communication methods.

Common transports include:

STDIO

Used for local MCP Servers running on the same machine.

Example:

AI Client
    ↓
STDIO
    ↓
Local MCP Server

HTTP

Used for remote MCP Servers hosted over a network.

Example:

AI Client
    ↓
HTTP
    ↓
Remote MCP Server

Why MCP Matters

MCP changes how AI ecosystems evolve.

Without MCP:

Every AI Platform
        ↓
Builds Every Integration

With MCP:

Integration Built Once
        ↓
Used Everywhere

This dramatically reduces integration complexity.

The Growing MCP Ecosystem

The MCP ecosystem is growing rapidly.

Examples include:

Development

GitHub
GitLab
Jira

Communication

Slack
Teams
Google Chat

Productivity

Notion
Confluence
Google Drive

Infrastructure

Kubernetes
AWS
Docker

Databases

PostgreSQL
MongoDB
Redis

Enterprise Systems

ERP
CRM
HRMS

MCP and Enterprise AI

Enterprise AI systems require:

Secure integrations
Standardized interfaces
Governance
Observability
Vendor independence

MCP provides the foundation for this future.

MCP in Karyam

Karyam uses MCP as a first-class platform capability.

This allows organizations to:

Connect external systems
Reuse existing MCP servers
Extend AI capabilities without custom development
Build interoperable AI systems

Why Karyam Chose MCP

Karyam believes enterprise AI should be:

Open
Extensible
Interoperable
Vendor-neutral

MCP aligns closely with these principles.

Rather than creating proprietary integration ecosystems, Karyam embraces open standards that allow organizations to build AI systems that evolve with their business.

Next Steps

Now that you understand MCP, continue with:

Connecting Servers
Authentication
Server Discovery
Examples