Using ChatGPT Developer Mode with Gram-hosted MCP servers

The new Developer Mode in ChatGPT  lets you wire external capabilities into ChatGPT via custom connectors.

:::note[Developer Mode] Developer Mode is currently only available on Plus and Pro plans. It is not available on Free or Business plans. You can read more about Developer Mode in the OpenAI documentation . :::

When combined with Model Context Protocol (MCP)  servers hosted on Gram, you can give ChatGPT access to your APIs and tools in just a few minutes.

This guide walks you through connecting ChatGPT to a Gram-hosted MCP server  using the example Acme To-Do API. You’ll learn to set up the connection, configure the custom connector, and test it with natural language prompts for managing your to-dos.

We’ll use the Acme To-Do API as our example. It’s a simple Python Flask REST API for managing to-do list items. You can find the complete OpenAPI document here: acmetodo.yaml.

Prerequisites

To follow this tutorial, you need:

• A Gram account 
• A ChatGPT account with access to Developer Mode: Use the Plus or Pro plan.
• A deployed version of the Acme To-Do API: Fork our Acme To-Do API repository  and deploy it to your own infrastructure.

Creating a Gram MCP server

If you already have a Gram-hosted MCP server configured, skip to the section on connecting ChatGPT to your Gram-hosted MCP server. For an in-depth guide to how Gram works and more details on creating a Gram-hosted MCP server, check out the Gram concepts guide .

Setting up a Gram project

In the Gram dashboard , click New Project to start the guided setup flow for creating a toolset and MCP server.

Enter a project name and click Submit.

Click Get Started on the project homepage. Gram will now guide you through the following steps:

Upload the OpenAPI document

Upload the Acme To-Do OpenAPI document, acmetodo.yaml, enter the name of your API, and click Continue.

Create a toolset

Give your toolset a name (for example, Acmetodo) and click Continue.

Notice that the names of the tools generated from your OpenAPI document are displayed in this dialog.

Configure MCP

Enter a URL slug for the MCP server (for example, acme-todo) and click Continue.

Gram creates the toolset from the OpenAPI document.

Click Toolsets in the sidebar to view the Acmetodo toolset.

You can edit individual tools in the toolset by clicking on the name or description of a tool.

Configuring environment variables

:::note[Server URL] By default, the server URL is set by the OpenAPI document. The acmetodo.yaml OpenAPI document uses a base URL of https://api.ritzademo.com, which points to the live demo Acme To-Do API.

If you use your own API, set the server URL to the URL of your deployment. You can either edit the OpenAPI document and skip the step below, or set the server URL in the environment variables. :::

Environments store API keys and configurations separate from your toolset logic.

In the Environments tab, click the Default environment. Click the three-dot menu in the top-right corner of the environment card and then click Fill for Toolset. Select the Acmetodo toolset and click Fill Variables to automatically populate the required variables.

Set the base URL variable for your API. For example, if your Acmetodo API is deployed at https://acme-todo.example.com, set ACMETODO_SERVER_URL to that value. Click Save.

Publishing an MCP server

Let’s make the toolset available as an MCP server.

Go to the MCP tab, find the Acmetodo toolset, and click on it to open the MCP details page.

Enable the MCP server by clicking the Enable button and then Enable Server.

Scroll down to the Visibility section and ensure the visibility of the server is set to Public.

Scroll down to the Hosted URL section and copy the URL.

Connecting ChatGPT to your Gram-hosted MCP server

ChatGPT supports the use of MCP servers as connectors in Developer Mode. Here’s how to add an MCP server connector to ChatGPT:

Add a Remote MCP server connector in ChatGPT

In Gram, go to the MCP tab, open your toolset, and copy the Hosted URL.

In the ChatGPT web UI, open Settings, go to the Connectors tab, and scroll down to Advanced settings. Click the Enable Developer Mode switch.

Return to the Connectors tab and click the Create button.

In the New Connector dialog, fill in the details as follows:

• Name: acme-todo

• Description: Briefly describe the connector.

• Server URL: Paste your Gram Hosted URL.

• Authentication: Here, you can choose to use OAuth for authentication. The Acme To-Do MCP server doesn’t require authentication, so we’ll choose No authentication.

  

• I trust this application: Click the checkbox to confirm that you trust the app.

Lastly, click Create.

Once saved, the MCP server appears as a connector in the Connectors section.

ChatGPT lists your tools as Actions. You can enable or disable individual tools.

Use the connector in a chat

Start a new chat in ChatGPT, click the + icon to open the dropdown, and select Developer Mode under More.

Enable the acme-todo connector.

Ask ChatGPT to perform an action that requires the Acme To-Do API. For example:

Troubleshooting

• Connector not connecting: Test the MCP server in the Gram Playground. Then, verify that the MCP server URL in the tool matches the MCP server URL in Gram.

  

• Authentication errors: For private servers, confirm your Gram API key in Settings -> API Keys and ensure the Authorization: Bearer ... header is configured in ChatGPT.

• Tools not appearing: Confirm that your toolset includes the expected tools and that the environment variables have been set. Finally, republish the server if you updated the toolset.

What’s next

You now have ChatGPT connected to a Gram-hosted MCP server built from the Acme To-Do API. Ready to build your own? Try Gram today  and turn any OpenAPI document into agent-ready tools your models can use safely.