Getting Started

Follow this guide to set up Kentik’s AI Advisor MCP Server.

🏃‍♀️ FAST TRACK: Already familiar with MCP and have your Kentik API token ready? Go to Configuration Examples and copy the snippet for your preferred agent.

Prerequisites

An active Kentik account with AI Advisor enabled.
Kentik API credentials (email and API token).
An MCP-compatible AI agent or development tool, e.g., Claude Desktop, GitHub Copilot (see Configuration Examples).

Get Your API Token

In the Kentik Portal, navigate to your User Profile (top-right menu).
Select the Authentication tab.
Copy your API Token.

Connect the MCP Server

Because the Model Context Protocol is a universal standard, connecting your preferred client is usually as simple as pasting a short JSON snippet or running a single terminal command.

To connect your AI agent or dev tool with AI Advisor MCP Server:

Update the appropriate Configuration Example with the endpoint and header information that applies to you.
Follow the instructions to apply the configuration in the agent or dev tool.
Start asking questions - remember to mention "Kentik" in your prompts

Note: If you’re using a custom AI agent framework, see Custom Applications and MCP Payload Reference for additional details.

Endpoints

Choose the URL for your region/cluster.

Region	Endpoint URL
US Cluster	`https://api.kentik.com/mcp`
EU Cluster	`https://api.kentik.eu/mcp`

Required Headers

Header Name	Value
`X-CH-Auth-Email`	<your-email@company.com>
`X-CH-Auth-API-Token`	<your-api-token>

Usage

Here are some tips on interacting with the AI Advisor MCP Server via your AI agent.

Effective Prompting

IMPORTANT: Always explicitly mention "Kentik" in your initial prompts.
Do: "Show me vmx devices from Kentik" or "What's the status of my Kentik devices?"
Don't: "Show me all devices."

If you don't include the “Kentik” keyword, the external AI agent may not realize it needs to invoke the Kentik MCP tools and might return a generic or incorrect response.

Real-time Progress

While AI Advisor processes your query, you will see real-time progress updates in supported MCP clients (like Claude Desktop and GitHub Copilot). These notifications show the AI’s reasoning process:

🔍 Created session... - Initial session created
⏱️ [2s] Analyzing device health metrics... - Processing updates
⏱️ [4s] Checking interface utilization... - Continued progress
✅ Completed - Final answer ready

Next, let’s check out the Configuration Examples for the AI Advisor MCP Server.