Connect Tabidoo to ChatGPT using MCP

The following procedure is intended for the ChatGPT desktop application.

Before starting, you need a Tabidoo API token. We recommend using a Shared API Token with only the required application roles.

For more information, see Tabidoo MCP Server.

Step 1 – Add the Tabidoo MCP server

Open the ChatGPT desktop application and go to the application settings.

Navigate to:

Integrations → Plugins → MCP

Select:

Add server

Configure the server as follows.

Name

Tabidoo

Type

Streamable HTTP

URL

https://app.tabidoo.cloud/mcp

Then add an authorization header.

Header

Authorization

Value

Bearer <YOUR_TABIDOO_API_TOKEN>

Replace <YOUR_TABIDOO_API_TOKEN> with the API token generated in Tabidoo.

Your configuration should therefore contain:

Name: Tabidoo
Type: Streamable HTTP
URL: https://app.tabidoo.cloud/mcp

Authorization: Bearer <YOUR_TABIDOO_API_TOKEN>

Step 2 – Restart ChatGPT

Save the MCP server configuration.

Then completely quit and restart ChatGPT.

Simply closing the application window may not restart the whole process.

On Windows, fully exit the application.

On macOS, use:

⌘Q

and start ChatGPT again.

Step 3 – Test the connection

Open a new conversation and ask:

What tables do I have in Tabidoo?

This verifies access to the Tabidoo structure.

Then try:

Show me the last 5 records from the Customers table.

This verifies reading data.

If your API token also has write permissions, try:

Create a new record in the Customers table with ...

If ChatGPT returns the requested Tabidoo data, the connection is working.

The MCP connection can then be used in regular conversations in the desktop application.

Example – Analyze invoices

Suppose you have an Invoices table and want to find your most important customer.

You can ask:

Go through the Invoices table in Tabidoo.

For last year, find the customer with the highest total invoiced
amount based on the "Total amount incl. VAT" field.

For that customer, show:
- total invoiced amount,
- number of invoices,
- average invoice value.

Compare the result with the second and third largest customers.

Then analyze:
- their share of total annual revenue,
- their payment behavior based on invoice statuses,
- and suggest 2–3 recommendations for working with this key customer.

Summarize the result in a short overview.

ChatGPT can retrieve the records through MCP, calculate the requested values and add an analysis of the results.

You can then continue in the same conversation:

Break this customer's revenue down by month.

Show me their overdue invoices.

Create a table with the top 5 customers.

Example – Overdue invoices and reminders

You can also ask ChatGPT to identify overdue invoices and prepare payment reminders.

For example:

Find all invoices in the Invoices table with the status "Overdue".

Sort them from the highest amount to the lowest and calculate
the total outstanding amount.

For the three largest overdue invoices, prepare a polite but clear
payment reminder email.

Include:
- customer name,
- invoice number,
- amount,
- due date,
- request for payment.

ChatGPT can filter the invoices, calculate the outstanding amount and prepare the reminder messages.

You can continue with requests such as:

Make the wording more direct. This is the second reminder.

Add late-payment interest of 0.05% per day.

Create a table of debtors including the number of days overdue.

Troubleshooting

If you save the MCP configuration but experience a problem, describe the error to ChatGPT.

The supplied setup guide notes that ChatGPT can often help identify the problem and suggest a solution.

Also check that:

• the MCP URL is correct,
• the API token is valid,
• Bearer is included before the token,
• the token has access to the required Tabidoo applications.

Security

Keep your API token private. If the token is exposed, replace it in Tabidoo.