Step-by-Step: Setting Up the NEW Native MCP Nodes in n8n

Key Concepts

• MCP (Model Context Protocol): A protocol enabling AI models to access and utilize external tools and services.
• MCP Server Node: A native n8n node that acts as a central hub, providing access to various tools (Gmail, Google Calendar, databases, etc.) to MCP clients.
• MCP Client Node: A native n8n node that allows AI agents or other applications to connect to an MCP server and utilize its available tools.
• Host: An application (like Claude desktop or an AI agent) that consumes the tools provided by an MCP server.
• Tool: A specific function or service (e.g., sending an email, accessing a calendar, querying a database) exposed by the MCP server.
• SSE (Server-Sent Events): The protocol used for communication between the MCP server and client.
• Call n8n Workflow Tool: A tool within the MCP server that allows triggering other n8n workflows, extending the server's capabilities.

Setting up the MCP Server Node

1. Add MCP Server Trigger: In a new n8n workflow, add the "MCP Server Trigger" node.
2. Production URL: Switch to the "Production URL" for a stable endpoint. The "Test URL" is not recommended.
3. Path: Define a path for the server (e.g., "MCP"). This will be part of the URL.
4. Authentication: Set authentication to "None" for simplicity in this example.
5. Add Tools: Connect various tool nodes (e.g., Gmail, Google Calendar, Pinecone Vector Database) to the MCP Server Trigger.
   • Gmail Tool: Configure the Gmail tool to send emails, using "Define by Model" for automatic parameter mapping.
   • Google Calendar Tool: Configure the Google Calendar tool to retrieve calendar events, using "Define by Model" for automatic parameter mapping.
   • Pinecone Vector Database Tool: Configure the Pinecone tool to query a vector database (e.g., for crypto knowledge).
   • Call n8n Workflow Tool: Use this tool to trigger other n8n workflows. Specify the workflow to call (e.g., "personal expense agent") and provide a description for the AI agent to understand when to use it.
6. Activate Workflow: Activate the n8n workflow containing the MCP server trigger.
7. Save: Save the workflow after adding or modifying tools.

Connecting Claude Desktop to the MCP Server

1. Install Claude Desktop: Download and install the Claude desktop application from Anthropic's website.
2. Enable Developer Tools: In Claude desktop, go to "Help" and click "Enable Developer Tools".
3. Edit Config: Go to "Settings" -> "Developer" -> "Edit Config". This opens the cloud-desktop-config.json file.
4. Replace Config: Replace the contents of the cloud-desktop-config.json file with the provided JSON configuration (available in the video description). This configuration enables the super gateway and specifies the command to run the MCP server.

   {
   "features": {
   "model_context_protocol": {
   "enabled": true,
   "args": [
   "npx",
   "-y",
   "@superflows/gateway",
   "--command",
   "npx",
   "--args",
   "-y",
   "node",
   "index.js",
   "--supergateway_url",
   "YOUR_N8N_PRODUCTION_URL"
   ]
   }
   }
   }
   

5. Replace URL: Replace "YOUR_N8N_PRODUCTION_URL" in the config file with the production URL of your MCP server trigger in n8n.
6. Save and Restart: Save the cloud-desktop-config.json file and restart Claude desktop.
7. Verify Connection: Claude desktop should now display the available MCP tools. If it shows "MCP n8n server disconnected," ensure the n8n workflow is active and the URL in the config file is correct.

Setting up the MCP Client Node with an AI Agent

1. Add Chat Trigger: In a new n8n workflow, add a "Chat Trigger" node.
2. Add Chat Model: Connect the "Chat Trigger" to a chat model node (e.g., OpenAI's GPT-4).
3. Add MCP Client Tool: Add the "MCP Client Tool" node.
4. Paste URL: Paste the production URL of your MCP server trigger into the "SSE Endpoint" field of the MCP Client Tool node.
5. Tools to Include: Choose whether to include "All" tools from the MCP server or "Select" specific tools.
6. Test: Use the chat interface to interact with the AI agent. It should automatically utilize the appropriate tools from the MCP server based on the user's query.
7. Execution Logs: Check the execution logs to verify which tools the AI agent is using.

Examples and Use Cases

• Calendar Integration: Asking "Do I have any meetings tomorrow?" triggers the Google Calendar tool to retrieve and display upcoming meetings.
• Knowledge Base Query: Asking "What did the price of Bitcoin spike to at the end of 2013?" triggers the Pinecone Vector Database tool to retrieve the answer from a crypto knowledge base.
• Expense Tracking: Asking "How much did I spend on marketing in the last quarter of 2024?" triggers a separate n8n workflow (via the "Call n8n Workflow Tool") to retrieve expense data from a vector database.

Key Arguments and Perspectives

• Centralized Tool Management: The MCP server provides a centralized way to manage and expose tools to multiple AI agents or applications.
• Simplified AI Agent Configuration: Instead of configuring each AI agent with individual tools, you can connect them to a single MCP server.
• Dynamic Tool Selection: AI agents can automatically select the appropriate tool from the MCP server based on the user's query.
• Extensibility: The "Call n8n Workflow Tool" allows extending the MCP server's capabilities by integrating with other n8n workflows.

Notable Quotes

• "This reduces the amount of work that we have to do significantly right and this can give your AI agents really a lot of power."
• "This is what makes this MCP server extremely powerful because now you can attach any workflow that you've built inside your n8n through this call n8n workflow tool inside your MCP server trigger."

Technical Terms

• Vector Database: A database that stores data as high-dimensional vectors, enabling efficient similarity searches.
• Index: A data structure used to optimize query performance in a database.

Logical Connections

The video demonstrates how to create an MCP server, connect it to both a desktop application (Claude) and an AI agent (n8n workflow), and use it to access various tools and services. The MCP server acts as a central hub, simplifying tool management and enabling dynamic tool selection by AI agents.

Synthesis/Conclusion

The MCP server and client nodes in n8n provide a powerful way to manage and expose tools to AI agents and other applications. By centralizing tool management and enabling dynamic tool selection, the MCP simplifies AI agent configuration and enhances their capabilities. The "Call n8n Workflow Tool" further extends the MCP server's functionality by allowing integration with other n8n workflows. While the technology is still new, it has the potential to significantly improve the way we interact with AI agents and external services.