Skip to main content

Custom AI app

If you're building an AI application with the OpenAI API, you can add Holistics as an MCP tool so your app can query data and run analytics as part of its workflow.

Overview​

Here's what you'll do to connect your custom AI app to Holistics:

  • Turn on MCP Server: A one-time step in your Holistics workspace, done by an admin. If it's already on, skip ahead.
  • Get your connection URL: Choose the environment you want to work with:
    • Reporting: Published Holistics objects (those on the Reporting tab).
    • Development: Unpublished Holistics objects in your current AMQL project branch (those on the Development tab).
  • Get your API key: A backend service can't complete an interactive OAuth login, so this uses API key authentication.
  • Connect in your app: Include the Holistics MCP server in your tool list when calling the OpenAI Responses API.

Step 1: Turn on MCP Server​

This is a one-time setup done by a workspace admin. Go to AI Settings and enable both Holistics AI and the MCP Server.

ai-mcp-setting

Step 2: Get your connection URL​

Your connection URL depends on your data center region and the Holistics environment you want your AI to work with.

Select your data region

Select your working environment

Let your AI work with published Holistics objects — those on the Reporting tab.

Get your connection url

https://mcp-apac.holistics.io/reporting/mcp


URL format
https://mcp-<REGION>.holistics.io/<ENV>/mcp
  • <REGION>: Choose based on your data center: apac, eu, or us
  • <ENV>:
    • reporting (default): Let your AI work with published Holistics objects (i.e. objects on your Holistics Reporting tab)
    • development: Let your AI work with unpublished Holistics objects in your current AMQL project branch (i.e. objects on your Holistics Development tab)

Step 3: Obtain API key​

Find your key under Account settings in Holistics.

Holistics Account Settings page showing the API key location

Step 4: Connect in your app​

Include the Holistics MCP server in your tool list when calling the OpenAI Responses API:

import OpenAI from "openai";
import type { ResponseInput } from 'openai/resources/responses/responses';

const client = new OpenAI({
  apiKey: "<YOUR_OPENAI_KEY>",
});

async function chat(input: ResponseInput) {
  const completion = await client.responses.create({
    model: "gpt-4.1",
    input: input,
    tools: [
      {
        type: "web_search",
      },
      {
        type: "mcp",
        server_label: "holistics_mcp",
        server_url: "<MCP_SERVER_ADDRESS>",
        server_description: "Holistics MCP to access my BI and business data",
        headers: {
          "X-Holistics-Key": "<YOUR_HOLISTICS_API_KEY>",
        },
        require_approval: "never",
      },
    ],
    tool_choice: "auto",
    parallel_tool_calls: true,
  });

  console.log(completion.output_text);
}

Replace <MCP_SERVER_ADDRESS> with your connection URL from Step 2 and <YOUR_HOLISTICS_API_KEY> with your API key from Step 3.

Need more help?​

See the MCP FAQ for questions about permissions, data access, and security.

Open Markdown
Let us know what you think about this document :)