The Model Context Protocol (MCP) is an open standard introduced by Anthropic in November 2024 that solves the “universal language problem” in AI-tool integration. Just as USB-C provides a universal connector for devices, MCP provides a standardized way for any AI assistant to connect to external tools and data sources without requiring custom integration code.

Before MCP, every AI framework (LangChain, OpenAI, Google’s dev kit) had its own proprietary way of connecting to external systems. Building an AI agent that could read emails, check calendars, and update databases meant writing separate integration code for each tool, with different authentication methods, error handling, and data formats.

MCP operates on JSON-RPC 2.0 and provides three core capabilities:

• Tools: Executable functions that perform actions (deploy infrastructure, send emails, create tickets).
• Resources: Access to data and information (read files, query databases, fetch metrics).
• Prompts: Templated workflows that guide AI behavior for specific tasks.

When an AI connects to an MCP server, it automatically discovers available capabilities through standardized methods. This means AI assistants can adapt to new tools without requiring updates or retraining. The protocol supports dynamic tool discovery, real-time notifications, and works across cloud providers, on-premises systems, and hybrid environments.

Why MCP matters for enterprise infrastructure:

• Standardization reduces complexity: What used to require bespoke development becomes configuration.
• Universal Interoperability: Your integrations work with Claude, ChatGPT, or any MCP-compatible AI.
• Real-Time Adaptability: AI agents learn about new capabilities automatically as infrastructure evolves.

Read more about the foundations of MCP →

The Itential Model Context Protocol (MCP) server enables AI assistants like Claude to safely orchestrate network automation and infrastructure management across enterprise environments.

The Itential MCP server exposes enterprise-grade infrastructure management capabilities to AI assistants through standardized tools.

How It Works

The MCP architecture creates a secure bridge between AI and your infrastructure via the Itential Platform:

When you ask Claude “Show me all devices in the Atlanta datacenter,” the AI:

• Understands your natural language request.
• Calls the appropriate Itential MCP tool (get_devices).
• Receives structured data from your Itential Platform.
• Formats and presents the results in a readable way.

This means AI makes decisions, Itential provides the execution and MCP provides tool exposure, ensuring enterprise-grade security and compliance at every step.

Installation

Option 1: Virtual Environment Installation (Recommended)

Prerequisites: Python >=3.10 is the minimal requirement for this installation method.

# Create a virtual environment
 python -m venv <venv-dir>/itential-mcp-venv
 
# Activate the virtual environment
 # On macOS/Linux:
 source <venv-dir>/itential-mcp-venv/bin/activate
 
# On Windows:
 <venv-dir>\itential-mcp-venv\Scripts\activate

# Install from PyPI
 python -m pip install itential-mcp
 
# Verify installation
 itential-mcp version

Note: Remember to activate the virtual environment each time you want to use the itential-mcp command.

source <venv-dir>/itential-mcp-venv/bin/activate

Option 2: Docker Container

For containerized deployments:

# Pull the latest version
docker pull ghcr.io/itential/itential-mcp:v0.x.x

Configuring Claude Desktop

Claude Desktop requires specific configuration to connect to the Itential MCP server using environment variables. Access the configuration through Settings → Developer → Edit Config.

Environment Variable Configuration

For Virtual Environment Installation:

{
  "mcpServers": {
   "itential-mcp": {
    "command": "<venv-dir>/itential-mcp-venv/bin/itential-mcp",
    "args": ["run"],
    "env": {
     "ITENTIAL_MCP_PLATFORM_HOST": "your-platform.example.com",
     "ITENTIAL_MCP_PLATFORM_CLIENT_ID": "your-client-id",
     "ITENTIAL_MCP_PLATFORM_CLIENT_SECRET": "your-client-secret",
     "ITENTIAL_MCP_LOG_LEVEL": "INFO"
    }
   }
  }
 }

Docker Configuration

If using Docker:

{
  "mcpServers": {
   "itential-mcp": {
    "command": "docker",
    "args": [
     "run", "-i", "--rm",
     "-e", "ITENTIAL_MCP_PLATFORM_HOST=your-platform.example.com",
     "-e", "ITENTIAL_MCP_PLATFORM_CLIENT_ID=your-client-id",
     "-e", "ITENTIAL_MCP_PLATFORM_CLIENT_SECRET=your-client-secret",
     "ghcr.io/itential/itential-mcp:v0.x.x"
    ]
   }
  }
 }

Note: Basic authentication is also supported using ITENTIAL_MCP_PLATFORM_USER and ITENTIAL_MCP_PLATFORM_PASSWORD instead of the OAuth variables above.

After saving any configuration, restart Claude Desktop to activate the MCP server connection.

Verifying Your Connection

After configuration, verify the connection in your AI assistant:

In Claude Desktop

1. Open a new conversation
2. Click the “Search and Tools” icon or type “Show MCP tools”
3. Look for “itential-mcp” in the connected servers list
4. Try a simple query: “Check the Itential platform health”

In VS Code with Cline

5. Open the Cline chat (usually in the sidebar)
6. Check the status bar for “MCP: Connected”
7. Type: “@mcp list tools” to see available Itential tools
8. Try: “Using Itential MCP, check the platform health”

More a visual learner? Check out this video from William Collins showing you how it’s done:

Once connected, try these queries in order to build confidence:

1️⃣ Health Check (Read-only, Safe)

“Check if the Itential platform is healthy”

What happens: Uses get_health tool to verify platform status

Expected response: Platform status, API availability, and component health

2️⃣ Device Discovery (Read-only)

“Show me all devices in Itential”

What happens: Queries device inventory with filters

Expected response: Formatted table of devices with management IPs

3️⃣ Workflow Listing (Read-only)

“What automation workflows are available?”

What happens: Searches workflows by keyword

Expected response: List of workflow names with descriptions

4️⃣ Configuration Check (Read-only)

“Get the show version for device switch-01?”

What happens: Retrieves device configuration

Expected response: Show ver with output interpretation

5️⃣ Launch a Gateway Manager Service

Launch service aws-ec2-list for region us-west-1

What happens: Initiates a service

Expected response: Parse script response and gives information about the output

6️⃣ Launch a Workflow

Launch Port Turn Up Service workflow

What happens: Ask for clarification on input parameters and initiates the workflow

Expected response: Job ID and progress updates

You’ve successfully configured the Itential MCP server to work with your AI assistant, creating a powerful bridge between natural language and enterprise infrastructure management. This setup enables:

• Natural language network operations: Ask questions, get answers.
• Safe automation: Start with read-only, progress to changes.
• Team Collaboration: Share configurations and patterns.
• Rapid Troubleshooting: AI-assisted problem resolution.

The combination of Python virtual environments for isolation, flexible configuration options, and progressive learning examples ensures a smooth journey from setup to production use.

Next Steps

• Complete the first five queries
• Create your own query templates
• Share your configuration with your team
• Join the community at github.com/itential/itential-mcp

Welcome to the future of infrastructure automation – where AI and enterprise systems work together seamlessly through the Model Context Protocol.