Agentic AI

How to Set Up the ChatGPT-MCP Connector (ChatGPT’s New Feature)

Learn how to set up the ChatGPT-MCP connector from OpenAI. Step-by-step guide for beginners to connect ChatGPT with apps, data, and tools.

By Ananya Rajeev 16 min read
chatgpt mcp connector setup guide

OpenAI ‘s ChatGPT has introduced a new feature called the ChatGPT-MCP connector that lets you link ChatGPT with external apps and data. This feature transforms ChatGPT from a standalone chatbot into a tool that can fetch information from your files, apps, or even custom databases in real time. In this beginner-friendly guide, we’ll explain what the ChatGPT-MCP connector is, why it’s useful, and provide a step-by-step walkthrough on setting it up. By the end, you’ll know how to enable this connector and use it to supercharge ChatGPT with your own sources of information.

What Is the ChatGPT-MCP Connector?

The ChatGPT-MCP connector is essentially a bridge between ChatGPT and other applications or data sources. “MCP” stands for Model Context Protocol, which is a standard way for AI models like ChatGPT to communicate with external tools securely and efficiently. In simple terms, the connector allows ChatGPT to connect to third-party apps or your internal systems and pull in relevant information or perform actions – all within your chat conversation.

Think of it like giving ChatGPT extra eyes and hands. With connectors, ChatGPT can see and retrieve data that you give it access to. For example, with a Google Drive connector, ChatGPT could search your Drive for a document. With a custom MCP connector, ChatGPT could even query a company database or trigger an action in an internal app. This capability is new, and it extends ChatGPT’s functionality beyond just text generation, making it more of an assistant that can interact with other services on your behalf.

Why “MCP”? The Model Context Protocol is the underlying technology that makes custom connectors possible. It standardizes how ChatGPT asks external services for information (like a question or a search query) and how those services answer back. Because of MCP, developers can create connectors that ChatGPT understands without needing special code for each app. The end result: ChatGPT can work with a variety of tools through a common protocol. For everyday users, this just means you can connect ChatGPT to many things, from popular apps to custom systems, as long as there’s an MCP-compatible connector available.

Why Use ChatGPT Connectors?

ChatGPT’s connectors (including the MCP connector) bring several benefits that can make your life easier:

  • Real-Time Information: Connectors let ChatGPT pull up-to-date data instead of relying solely on its training knowledge. For instance, ChatGPT can fetch today’s entries from your calendar or the latest numbers from a spreadsheet if those apps are connected. This keeps answers current and relevant.
  • Personalized Assistance: With connectors, ChatGPT can access your own files and services (with permission). That means responses can be tailored to your context. If you connect your email, you could ask “Summarize the last email from my boss,” and ChatGPT would actually read your inbox (with your authorization) to give an answer.
  • Automation and Productivity: The connector feature turns ChatGPT into an action-oriented assistant. It can not only find information but also help perform tasks. For example, a connector might allow ChatGPT to create a task in your project management app or fetch customer data from a CRM. You don’t have to switch between applications – you can do it through a chat prompt.
  • Custom Integrations: The MCP connector in particular shines for custom needs. If you have an internal database or a less common tool, you (or your development team) can set up an MCP server for it. ChatGPT can then use that connector to interact with your specific data or tools. This is powerful for businesses and power users who want to integrate ChatGPT into their unique workflows.
  • Expanded Capabilities: Overall, connectors expand what ChatGPT is capable of. It’s no longer limited to just answering general knowledge questions or writing text. It can reference your tools and data directly. This makes ChatGPT more of a one-stop assistant that can help with both information and actions across various platforms.

In short, using ChatGPT connectors can save time, reduce the need to manually look up information, and help personalize your AI experience. Now, let’s get into how to set up this new feature step by step.

Prerequisites for Using the MCP Connector

Before you start, there are a couple of things to check to ensure you can use the ChatGPT-MCP connector:

  • Access to Connectors: Currently, the connector feature (especially custom MCP connectors) is available to certain ChatGPT plans. Make sure you have a ChatGPT version that supports connectors. For example, ChatGPT Plus/Pro or Enterprise plans have this feature enabled, whereas the free version might not show the Connectors option. If you’re an individual user on a paid plan, you should see “Connectors” in your settings. In a workplace setting, your organization’s admin may need to enable connectors for your team.
  • An MCP Server or Supported App: Decide what you want to connect ChatGPT to. For popular apps like Google Drive, Gmail, Dropbox, etc., OpenAI provides built-in connectors – you just need an account with those services. But if you want to set up a custom MCP connector, you’ll need an MCP server URL. This could be provided by a third-party service or something you’ve built (if you are a developer). Essentially, an MCP server is a program that exposes certain actions or data for ChatGPT to access via the connector. If you’re not a developer, you might stick to the pre-made connectors for common apps, or use an MCP connector only if someone has set up the service for you.
  • Permissions and Trust: If you are connecting to a company application or any third-party service, ensure you have the permission/credentials for it. When setting up the connector, ChatGPT will ask you to confirm that you trust the application you are connecting. This is a safety step since custom connectors are not vetted by OpenAI. Be sure you trust the source of the connector before proceeding.

Once these prerequisites are in place, you’re ready to connect ChatGPT to the outside world!

How to Set Up the ChatGPT-MCP Connector (Step by Step)

Setting up the ChatGPT-MCP connector is straightforward. The interface is user-friendly, so even beginners can get it done by following these steps. Here’s how to add a new connector in ChatGPT:

  1. Open ChatGPT Settings: In the ChatGPT app (web or desktop), click on your profile icon or user menu (usually found in the bottom-left or top-right corner). From the menu, select “Settings.” This will open the settings panel.
  2. Go to Connectors: In the Settings panel, look at the sidebar or tabs and find “Connectors.” Click on it. This is where all connector options are listed. If you’re on a supported plan, you should see a list of available apps (like Google Drive, Gmail, etc.) and an option to add a custom connector.
  3. Add a New Connector: In the Connectors section, look for a button that says “Create” or “Add New Connector.” Click this button to start creating a new connector. A dialog box or form will appear, titled “New Connector” (or similar).
  4. Enter Connector Details: Now you need to fill in the details for your connector:
    • Name: Give your connector a clear name. It can be anything descriptive. For example, “My Company Database” or “Weather API Connector.” Choose a name that will remind you what this connector is for.
    • Description (Optional): You can add a short description of what the connector does or the data it provides. This is helpful if you plan to have multiple connectors. For instance, “Connector to internal sales database” or “Fetches data from XYZ service.”
    • MCP Server URL: This is the most important part. Enter the URL of the MCP server or endpoint you are connecting to. The format will usually look like a web URL (e.g., https://your-server.com/mcp or an address provided by a service). If you are using a connector for a well-known app through an official integration, this might be handled automatically. But for custom connectors, you’ll have to input the server address that hosts the MCP interface. Double-check that the URL is correct.
  5. Trust the Connector: If this is a custom MCP connector, ChatGPT will warn you that it’s not verified by OpenAI. You will see a checkbox or prompt saying something like “I trust this application.” Only check this if you trust the source of the connector. (For example, if it’s your own server or a reputable service.) This step is crucial for security – it ensures you’re aware that you’re linking ChatGPT to an outside system.
  6. Create/Add the Connector: After filling in the name, description, and URL (and confirming trust), click the “Create” or “Add” button in the dialog to finish. ChatGPT will now save this connector. You should see your new connector appear in the list of connectors in your settings. It might show up alongside any other connectors, possibly labeled as a “Custom” connector.
  7. Authenticate if Required: Depending on the connector, the next step might prompt you to connect or log in. For example, if you added a built-in connector (say for Google Drive or HubSpot), after clicking Connect you would be redirected to log into that account and grant ChatGPT permission. For a custom MCP connector, if the server requires an API key or login, you may have to provide that in this step. Follow any on-screen instructions to complete the authentication. If no authentication is needed (some custom connectors might not require login), then you can skip this.
  8. Confirmation: Once added (and authenticated if needed), the connector is set up! In the Connectors list, it should now show as Connected or Available. For example, it might have a toggle or a status indicator. Ensure it’s enabled or active. If it’s a custom connector with no login, it might just be listed as ready to use.

That’s it – you’ve set up the ChatGPT-MCP connector. The process is quite similar to adding an app on your phone: you give ChatGPT the details and permission to use an external service, and ChatGPT installs the “bridge” to that service.

Using the Connector in ChatGPT Chats

Now that your connector is set up, how do you actually use it during a conversation with ChatGPT? It’s easy to use, but it’s not automatic by default (except for certain connectors that ChatGPT can auto-use). Here’s how you can use your new connector in a chat:

  • Start a New Chat: Go back to the main ChatGPT interface and open a new chat (or you can use an existing chat if appropriate). It’s often best to start fresh when trying a new connector.
  • Enable Connectors for the Prompt: Look at the message input area where you type queries. You should see a small icon or label for Tools (sometimes depicted as a plug, search icon, or similar). Click on it, and you’ll see options like “Use connectors” or “Run deep research.” Select “Use connectors.” This tells ChatGPT that your next query will involve external sources.
  • Select Your Connector: After clicking “Use connectors,” you’ll be prompted to choose which source(s) to use. You should see the connector you added listed here, alongside any other apps you’ve connected. Click the checkbox or select your new MCP connector (and any other relevant sources if you want to use multiple). You can usually select more than one source if needed. For instance, you could query both your custom database and the web in one go. But to keep things simple, just pick the one connector for now.
  • Ask Your Question or Give a Task: Now type your query in the chat box. Ask something that requires the connector’s data or action. For example, if your connector links ChatGPT to a company database, you might ask: “What are the latest sales figures for this month from our database?” If it’s connected to a project management tool, you could ask: “Create a new task in [Tool] for designing the homepage.” Essentially, you interact with ChatGPT as usual, but now it has the ability to reach into the connected app to get information or execute commands.
  • Review the Answer: When you send your question, ChatGPT will use the connector behind the scenes to fetch the info it needs. You might notice it thinking a bit longer as it’s retrieving data. Once it compiles an answer, you’ll see the response just like any other ChatGPT answer – but this time, it’s enriched with your connected data. Often, ChatGPT will provide citations or references if it pulled specific data (for example, a citation might indicate it got the info from your connected source). You can click those references (if provided) to see the actual data in the original app, which is a great way to verify and follow up.
  • Continue the Conversation: You can ask follow-up questions referencing the data it just pulled, and ChatGPT will continue to use the connector as needed. For instance, if it gave you a list of recent sales orders from a database, you could then ask, “Which of those orders is the largest, and who is the client?” and it will dig deeper via the connector if possible. Remember to keep the connector active (in some interfaces, you might need to toggle “Use connectors” for each new question in that session, depending on the mode).

Tip: If you have automatic connectors (for example, ChatGPT can auto-use Gmail or Calendar if you have those connected and the feature is enabled), you might not need to manually select them every time – ChatGPT will use them when it thinks it’s relevant. However, for custom MCP connectors, you’ll likely be selecting them when you want to tap into them, until any auto-use feature is introduced for those.

Using connectors in chat might feel like having a supercharged ChatGPT. Instead of the AI saying “I don’t have access to that information” for personal or private data, it can actually retrieve exactly what you ask for, from the source you’ve connected.

Tips for Getting the Most Out of ChatGPT-MCP Connectors

As you begin to use this new feature, here are a few tips and best practices to ensure a smooth experience:

  • Use Clear Prompts: When dealing with connectors, be clear in your request so ChatGPT knows to use the connector. For example, include the context like “in my Google Drive” or “from my database” in your question. E.g., “Find the ‘Annual Report 2025’ in my Google Drive and summarize it.” This nudges ChatGPT to use the appropriate connector and gives it a hint on what tool or data to access.
  • Verify Responses: Even though ChatGPT is pulling actual data via connectors, it’s always good to double-check important information. If ChatGPT cites a source (like a document or an email it found), click the citation or go to the app itself to make sure the AI interpreted it correctly. This is especially true for critical or sensitive tasks.
  • Mind the Limits: Connectors are powerful, but they have some limitations. Some connectors might not be able to access certain types of data (for instance, a Google Drive connector might not read images or very large spreadsheets). Custom MCP connectors will only be as useful as the tools they expose. If your MCP server doesn’t have a function for a very specific request, ChatGPT might not get what it needs. If you run into a limitation, you may need to adjust your question or check if the connector’s scope can be expanded.
  • Security and Privacy: Always remember that when you connect an app or a service to ChatGPT, you are granting the AI access to that data. Make sure you’re comfortable with that. For personal accounts, maybe avoid connecting highly sensitive information. For work accounts, ensure you follow your company’s data policies. OpenAI has built connectors with security in mind (data is typically accessed only as needed, and you can revoke access by disconnecting anytime), but it’s wise to be cautious with any integration.
  • Stay Updated: The connector feature is still new and evolving. OpenAI might add more connectors for popular services, improve the MCP standard, or change how the interface works as they refine the feature. Keep an eye on release notes or the help center for any updates. New capabilities (like expanded auto-usage or new tool integrations) could roll out, further boosting what you can do with ChatGPT. Staying informed will help you take full advantage of connectors.
  • Disconnect if Not Used: If you no longer need a connector or you just wanted to test it, you can disconnect it for safety. Simply go back to Settings > Connectors, find the app or custom connector, and there will be an option to Disconnect or Remove it. This ensures ChatGPT can’t access that source anymore until you connect it again.
chatgpt mcp connector setup guide
chatgpt-mcp-connector-setup-guide

Conclusion

The ChatGPT-MCP connector is an exciting new feature that unlocks a world of possibilities for ChatGPT users. In this guide, we covered what it is, why it’s useful, and exactly how to set it up and use it. To recap, the MCP connector allows ChatGPT to securely connect with external applications and data using the Model Context Protocol. By setting up connectors, you can ask ChatGPT to tap into your files, emails, databases, or other services and get answers that would otherwise be beyond its reach.

For beginners, it might sound technical at first, but the process of enabling a connector is quite simple – it’s mostly clicking, naming, and pasting a URL. Once set up, using a connector is as easy as asking ChatGPT a question in plain language. You’ll quickly get the hang of it.

This new feature turns ChatGPT into more of a personal assistant that can handle tasks across different platforms. Whether you’re a professional looking to streamline work (like querying a CRM or project management tool) or an individual who wants ChatGPT to help manage personal data (like searching your notes or scheduling events), connectors can save you time and effort.

Give the ChatGPT-MCP connector a try and see how it enhances your workflow. As with any powerful tool, start with simple experiments – maybe connect a calendar or a drive and ask a few questions. Once you see it in action, you’ll likely discover many ways it can help you. Happy connecting, and enjoy the richer, more personalized assistance that ChatGPT can now offer!

Further Reading & Resources

If you enjoyed this guide and want to explore more AI innovations, check out these related posts on the Ossels AI Blog:

And here are a few useful external resources to expand your knowledge:

Ananya Rajeev

Ananya Rajeev is a Kerala-born data scientist and AI enthusiast who simplifies generative and agentic AI for curious minds. B.Tech grad, code lover, and storyteller at heart.