Goal: Simplify AI Agent's interaction with various tools by combining them through a single MCP server that will perform tasks contextually.
Scope of application: Automate tasks using AI and integrating with tools like Airtable to manage data more effectively.
Tools:
- n8n cloud account updated to the latest version (https://docs.n8n.io/manage-cloud/update-cloud-version/)
- OpenAI profile
- Airtable account with the spreadsheet created (as an example)
How does this work?
The cloud version of n8n now allows you to create your own MCP server and use it in various other automations.
If you used to have an AI Agent in a workflow that connected many different tools and had to write clear instructions for each of them, now your Agent will have only one tool that will access your pre-configured server and contextually perform your task on its own.
First, create a new workflow in n8n
Here we will create and configure our MCP server
Create the first block trigger called MCP Server Trigger.
So far, let's not touch the node's settings
Click on + under Tools to add the tools you need
In this guide, I'll show you the process of creating and configuring Airtable as an example.
Go to the settings of the first Airtable tool
- Plug in your credentials
- Operation: Create
- Choose yours for your previously prepared Base and Table
To do this, I created a table with tasks like this and called it “To do”
- Next to Values to send In the Task column where the task itself will be recorded, click on the star icon on the right. This feature will allow the AI model to decide on its own and enter a name there based on your dialogue with it.
- Status - Leave it on because it should not be checked “done” when creating a task
Now we are creating another such Airtable tool
- Connect credentials
- Choose your spreadsheet
- This time the surgery will be Search
- IN Filter By Formula we don't write anything, we leave it as it is
- Return All - must be included
- In Add options, add Output Fields
- IN Output Fields select both columns from the table
Now we repeat the procedure again, but this time Operation will be Update
But in Values to Update select the Let the model define this parameter function in all three fields so that AI decides for them and substitutes information based on the context of your conversation.
Also under the field id I advise you to click on Add a description and add some guide for AI so that it doesn't get confused, for example: Airtable record id, usually in format “GTysbas42Hythids73"
You have created your own MCP server with the most basic Airtable features, and this server can be supplemented with the tools you need without any restrictions.
Save workflow and activate
Now we are creating a new workflow for testing.
Creating the first trigger node - the most classic On chat message
Then go to Advanced AI
And choose AI Agent
Now click on + below Chat Model and choose a language model from OpenAI
Add the most common memory inside n8n − Simple Memory
And now in Tools find and add MCP Client Tool
Here, you'll need to add a link to SSE Endpoint from your workflow with the MCP server.
Let's go back to workflow, where you created the MCP server.
Click on the block MCP Server Trigger
Please select Production URL and copy the url link below
Paste this link in your second automation in SSE Endpoint
Now save your workflow.
It's ready.
You can test and adapt this new feature to your personal needs.