Connect Cloud Storage to ChatGPT Using MCP

Connecting cloud storage with ChatGPT through MCP gives ChatGPT a way to work with your files and folders directly, instead of relying only on files uploaded in each chat.

Once the connection is set up, you can ask ChatGPT to find files, create folders, organize content, or work across documents using simple prompts.

In this hands-on guide, I’ll show the full setup process in ChatGPT Desktop and then test the connection with a real file task.

Step 1: Open Settings

First, open ChatGPT on desktop.

Click on your profile at the bottom-left side and then click Settings.

This is where all the connection and plugin options are available.

Step 2: Open the MCP Section

Inside Settings, go to Plugins.

At the top, you will see the MCPs option. Open it.

Here you can see the MCP servers already connected with ChatGPT.

Step 3: Add a New Server

Now click on Add server.

A new screen will open where you have to enter the MCP details.

This is where we are going to add SAIT.

Step 4: Add the Server Name

In the Name field, I entered:

sait

Then I selected Streamable HTTP as the connection type.

There is nothing much to change here. The name is mainly for identifying the server later.

Step 5: Enter the MCP URL

Next comes the important part.

In the URL field, I added:

https://mcp.sait.ai/mcp/oauth

I kept the remaining fields as they were and clicked Save.

Once saved, SAIT appeared in the MCP server list.

Step 6:Review permissions

Now go back to the MCP list.

You will see sait there with an Authenticate button.

Click on it.

This starts the login and permission process between ChatGPT and SAIT.

Step 7: Check the Permissions

Before the connection is completed, you will be asked what level of access you want to give.

There are mainly two options:

Read Only if you only want ChatGPT to read or search the files.

Read & Write if you also want ChatGPT to create, edit, move or organize files.

For this test, I selected Read & Write and clicked Continue.

Step 8: Approve the Connection

After that, one more confirmation screen appears.

Here you can check which account is connecting and which MCP URL is being used.

In my case, it showed:

Connecting as: Codex
Connecting to: https://mcp.sait.ai/mcp/oauth

Once I checked the details, I clicked Approve.

Step 9: Connection Done

After approval, the authentication is completed.

There is no extra setup required after this.

Now the main thing is to check whether ChatGPT is actually able to communicate with SAIT.

Step 10: Test the Connection

To test it, I opened a new chat and asked:

What can you tell me about my SAIT account?

ChatGPT was able to read the connected account information and show the details available through the MCP connection.

That confirmed that the connection was working.

I Also Tried a File Task

After checking the connection, I wanted to see whether it could actually do something useful with the files.

So I asked:

Can you organize these images into new folders based on subject matter?

The images were divided into different folders based on what was in them.

For example:

• City and Architecture
• Night Sky and Mountains
• Water Landscapes
• Nature Closeups
• Sky and Clouds
• Color and Abstract

The existing Test folder was not changed.

This was a better test than just checking account details because here ChatGPT was actually working with the connected files.

Add screenshot here.

What Else Can You Try?

After the connection is working, you can try normal prompts like:

Show me the files in this folder

Create a new folder

Find files related to marketing

Organize these files by type

Move these images into separate folders

Show me the latest files

Summarize the documents in this folder

You do not need to use any special command. You can simply tell ChatGPT what you want it to do.

My Take

The setup part is quite simple.

The useful part is what happens after the connection is done.

Normally, if I want ChatGPT to work with a file, I have to upload it in the chat. With an MCP connection, ChatGPT can work with the files already available in the connected workspace, depending on the permission you give.

That is what makes this setup interesting.