Connecting ChatGPT to One Model via MCP

Overview

This guide explains how to connect ChatGPT to One Model using the Model Context Protocol (MCP).

Once connected, ChatGPT can securely query governed workforce metrics from One Model and incorporate those results directly into conversational workflows such as analysis, document drafting, reporting, and presentations.

By connecting ChatGPT to One Model via MCP, you can:

Reduce context switching by querying workforce data directly from ChatGPT
Stay within your existing workflow while drafting content, writing code, or building presentations
Combine AI reasoning with trusted One Model metrics to answer complex workforce questions

Please Note: One Model's MCP integration is currently in an early-access phase. Because AI platforms update their interfaces frequently, you may find that some buttons, menu names, or configuration steps have shifted slightly since this guide was published. System Variations: Steps may vary depending on your specific tenant setup, regional settings, or administrative permissions. Help Us Improve: If a step looks different in your system or you encounter an unexpected result, please let us know! Use the Feedback Form to help us refine these instructions for the final release.

Terminology & Notes

In ChatGPT, MCP integrations are referred to as 'Apps.' These apps act as connectors to external systems such as One Model.
Throughout this guide, the terms App and Connector are used interchangeably, with App reflecting ChatGPT's interface terminology.
This guide uses ChatGPT as the reference AI platform. Other MCP-compatible tools may have similar steps but different terminology.

Audience & Prerequisites

This guide is intended for:

One Model Administrators responsible for setup, configuration and publishing
Business Users and Analysts with approved access who will connect to and use the app

Part 1: One Model Administrators

Setup, Configuration, and Publishing

To authorize the connection (the 'handshake') between ChatGPT and your One Model production instance, an Administrator must have the appropriate access levels to both platforms.

Required Permissions in One Model

Ensure your user account is assigned an Application Access Role (AAR) with the following permissions enabled:

CanAccessOneModelMCP - Allows use of the MCP to communicate with ChatGPT or other external AI tools
CanRegisterMCPClients - Allows you to register and manage the MCP client credentials used by ChatGPT
CanConfigureCompany - Grants access to high-level company settings

These permissions are configured via: Admin > Application Access Role > [Select Role] > Permissions.

Register a Dedicated MCP Client

You register a dedicated MCP client in One Model, which generates an OAuth Client ID and Client Secret that you paste into ChatGPT's app configuration. You stay in control of these credentials and can revoke them independently.

Note on order of steps: ChatGPT generates its OAuth Callback URL inside the app creation screen, so you'll begin creating the app in ChatGPT first, register the client in One Model, and then return to ChatGPT to finish.

Step 1 - Begin creating the App in ChatGPT

Log into ChatGPT
Open Settings (via your profile icon)
Navigate to Security and login
Turn Developer Mode on then close this settings panel

Open Workspace Settings (via your profile icon)
Click Apps
Click the Create button
Configure the One Model connection:
- Name: e.g., One Model
- Connection: the endpoint for your One Model MCP server
  - Be sure to use the correct URL for your Production environment
  - Your MCP Server URL is based on your One Model site URL - add api. before your company ID and /mcp at the end
  - One Model URLs use one of these domains: onemodel.us, onemodel.com.au, onemodel.ca, or onemodel.eu
  - Example: https://api.[companyid].onemodel.[us]/mcp
- Authentication Method: OAuth
- Open the Advanced OAuth Settings and set the Registration Method to User-Defined OAuth Client
- Locate the Callback URL field and click the Copy icon

Keep this ChatGPT window open; you will return here.

Step 2 - Register the client in One Model

Log into your One Model production instance
Navigate to Admin > Integration
Open the MCP Clients tab and click Register New Client

Complete the registration form:
- Client Name: A descriptive label that identifies this client in audit logs and the MCP Clients list, e.g. ChatGPT Production
- Authentication Method: How your client sends its credentials when requesting a token. Set this to POST Authentication.
- Redirect URI: Paste the Callback URI you copied from ChatGPT in Step 1
Click Register

Note: For different AI platforms you may need to experiment between Basic and POST Authentication if it isn’t clear from their documentation.

Capture your credentials

Once the client is registered, One Model will display the OAuth Client ID, OAuth Client Secret, and the Authentication Method you selected.

Important: Copy the Client ID and Client Secret to a secure location before navigating away from the page. If you lose the secret, you'll need to register a new client.

Review your registered clients

Registered clients are listed on the MCP Clients tab, showing the client name, redirect URL, and authentication method. From here you can Modify a client's details or Revoke it.

Step 3 - Finish the App in ChatGPT

Return to your ChatGPT 'New App' window (Registration Method should still be set to User-Defined OAuth Client)
Paste the OAuth Client ID and OAuth Client Secret you captured in Step 2 into the corresponding fields. The rest of the fields will be automatically filled in for you.
On the main panel, click the checkbox “I understand and want to continue”.
Click Create

Publishing the Connector

The connector must be published before it can be used by end users.

Publish the App

You should automatically see the new connector and an option to Publish. This might take a few seconds to show up.
If you don’t see it, search for the newly created One Model app
Acknowledge any required trust or security confirmations
Click Publish

Important Notes on Publishing & Discoverability

Publishing an app makes it available for use, but visibility may not be immediate. Please note:

Publishing does not guarantee immediate discoverability
Developer Mode affects visibility - Apps created or modified while Developer Mode is enabled may only be visible to the creator
Republishing may reset trust or visibility - Users may need to re-acknowledge trust or refresh their session
User sessions are cached - End users may need to refresh the page, switch chats, or log out and back in

Verify the connection

Once the app is created, you'll need to authorize it for your own session before you can test it. Follow the steps in Part 2: Business Users and Analysts to connect, accept permissions, and verify the integration is working.

Part 2: Business Users and Analysts

Connecting to and Using the One Model App

Required Permissions

To connect ChatGPT to One Model, your user account must have an Application Access Role (AAR) with:

CanAccessOneModelMCP enabled

This is managed by your One Model Administrator via: Admin > Application Access Role > [Select Role] > Permissions

Log in to One Model

Connect the One Model App in ChatGPT

In ChatGPT, select Plugins from the left panel
Search for your newly created app, e.g. “One Model”
Select the app and click the + (plus) button to connect
In the Connect One Model modal, click Continue to One Model
A confirmation screen will display the connector's capabilities
1. These permissions are defined by One Model
2. They are read-only during setup

Review the consent page and click Accept to finalize the connection.

Click Start Chat

Once connected, the One Model app appears below the Ask anything prompt.

Note: After initial setup, the app can also be accessed via the + button in the chat interface ro calling it with the @ symbol, e.g. @One Model.

Test the App

Enter a query that requires One Model data, for example:

What is my external hire rate trend?
What is my current headcount?

ChatGPT will route the request through One Model's MCP layer and return results based on governed workforce data.

Tip: If you encounter errors or no data, confirm that the app is visible, connected, and that you have the required permissions.

Feedback Form

We encourage beta testers to complete the Feedback Form to share what's working well, report issues, and suggest enhancements.

If you experienced app visibility or discoverability issues, please note:

Whether the app was newly published or republished
Whether Developer Mode was enabled at the time