Connect the nodegoat MCP to your AI Assistant
The nodegoat Model Context Protocol (MCP) server allows you to connect your environment to AI assistants that are powered by Large Language Models (LLMs). This Guide focuses on two assistants: Claude.ai and ChatGPT.com. These two assistants are widely used and provide straightforward configuration options for configuring a connection to an MCP.
The MCP exposes the model and data as well as stored Filters, Scopes, and Conditions. This means that the more stored settings you have in your Project, the more information your assistant of choice has to work with.
To make this work, first configure the API of your research environment. You should have at least one project enabled in the API settings, and you should have configured at least one active client.
If you do not have access to the API module that can be accessed via 'Management', please contact the administrator of your nodegoat environment or send a message to support@nodegoat.net.
The MCP will have the same clearance as the authorised user. If you want the MCP to be in a read-only mode, use a user with the clearance level 'Demo'. If you want the MCP to be able to add, edit, and delete Data, but not the Data Model, use a user with the clearance level 'User'. If you want the MCP to be able to add, edit, and delete Data and be able to add, edit, and delete the Data Model, use a user with the clearance level 'Administrator'.
In order to create an MCP for a project in your environment, you need to specify the identifier of your Project. To find this identifier, go to ‘Management’ and ‘Projects’. Here you can view or edit your Project of choice. Under the name of the Project, the identifier is shown. Use this identifier in the following URL:
https://api.nodegoat.net/project/[Your Project ID]/.mcpClaude.ai
After logging in to your Claude account, use the menu on the left to click on ‘Customize’. At the top of this page click ‘Connectors’ and then click the ‘+ Add’ button on the right to select ‘Add custom connector’. Fill in the name of the Connector (e.g. ‘nodegoat Correspondence Data’) and enter the URL of the MCP.
In the next dialog, keep the ‘Authentication’ setting to ‘Sign in now’. In the ‘OAuth client’ settings select ‘Use your own OAuth client’ and fill in the ‘OAuth client ID’ and the ‘OAuth client secret (optional)’. Both of these values can be found when you open the API client in your nodegoat environment. The ‘Identifier’ value is used for the ‘OAuth client ID’ and the ‘Passkey (secret)’ is used for the ‘OAuth client secret (optional)’. Click ‘Add’.
In the next step you click the ‘Connect’ button to connect to the MCP. This will send you to a page where you can authorise the currently logged in user. If no user is logged in, you can login to a nodegoat account and then authorise this user. After completing this step, all available options of the MCP are listed in the connector.
ChatGPT.com
After logging to your ChatGPT account, use the menu on the left to click on ‘Plugins’. Click the ‘Add’ button on the right and select ‘Add custom MCP server’. Enter a name (e.g. ‘nodegoat Correspondence Data’) and enter the URL of the MCP in the ‘Connection’ input field.
Keep the ‘Authentication’ settings to ‘OAuth’ and click ‘Advanced OAuth settings’. Enter the ‘OAuth client ID’ and ‘OAuth client secret’ values. Both of these values can be found when you open the API client in your nodegoat environment. The ‘Identifier’ value is used for the ‘OAuth client ID’ and the ‘Passkey (secret)’ is used for the ‘OAuth client secret’. Check the checkbox next to ‘I understand and want to continue’ and click ‘Create as a plugin’. In the next dialog click ‘Connect [Name of your Connector]’.
In the next step you click the ‘Connect’ button to connect to the MCP. This will send you to a page where you can authorise the currently logged in user. If no user is logged in, you can login to a nodegoat account and then authorise this user. After completing this step, all available options of the MCP are listed in the connector.
Using the MCP
In both AI assistants you can now write a question that references the connector or be very specific and state ‘use the connector [name of connector] to generate a pie chart based on the amounts of letters sent by people in this dataset’.
If you have connected a user with the clearance level ‘Administrator’ you can also use the MCP to adjust the data model. Use this with caution to avoid unintentionally losing data.