Case Study RAG Workflow Automation With N8n And Gdrive MCP Server

🚀 Interact with External Tools via Model Context Protocol (MCP) Using n8n and AI Agents

This project demonstrates how to leverage n8n's AI agent capabilities to discover and execute external tools through the Model Context Protocol (MCP).

🚀 Quick Start

This documentation details how to utilize n8n's AI agent feature to discover and execute external tools via the Model Context Protocol (MCP). Using Brave Search as an example, it shows how to set up a workflow for dynamic tool invocation.

📦 Installation

Start the Docker Container

docker run -p 1234:1234 -e N8N_COMMUNITY_PACKAGES_ALLOW_TOOL_USAGE=true -v n8n_data:/home/node/.n8n node8n GmbH/n8n

This command starts a Docker container with the following configurations:

• -p 1234:1234: Maps port 1234 of the container to port 1234 of the host.
• -e N8N_COMMUNITY_PACKAGES_ALLOW_TOOL_USAGE=true: Sets an environment variable to allow community packages to be used as tools.
• -v n8n_data:/home/node/.n8n: Mounts a volume to persist data.

Install Dependencies

Ensure the following tools are installed:

• Docker
• Node.js (Recommended version 14 or higher)
• npm (Recommended version 6 or higher)

💻 Usage Examples

Tool Discovery and Execution Nodes

MCP Client (List Tools)

{
  "type": "MCPClient",
  "operator": "listTools",
  "credentials": {
    "type": "MCPBraveSearch"
  }
}

This node is used to obtain a list of available tools, including tool names, descriptions, and parameter explanations.

MCP Client (Execute Tool)

{
  "type": "MCPClient",
  "operator": "executeTool",
  "credentials": {
    "type": "MCPBraveSearch"
  },
  "toolName": "{{ $fromAI('tool', 'selected tool to execute') }}",
  "toolParameters": {
    "mode": "definedAutomaticallyByModel"
  }
}

This node executes specific tool operations based on the decisions of the AI agent.

Workflow Steps

1. Receive Chat Messages: Receive user input through n8n's chat interface.
2. AI Agent:
   • Process requests using a preset conversation model (e.g., Groq or OpenAI).
   • Configure a memory module to maintain conversation history.
   • Set system prompts to guide the AI on how to discover and use tools.
3. List Available Tools: The AI agent obtains information about executable tools through the MCP client.
4. Execute Selected Tool: Dynamically select and invoke the corresponding tool (e.g., Brave web search) based on user needs.

🧪 Experiment and Testing

1. After enabling the workflow, enter the following in n8n's chat interface:
   • What tools do you have? or list available tools
     • The AI agent will return a list of available tools and their function descriptions.
   • Tell me about the latest developments in the Model Context Protocol.
     • The AI agent will invoke the Brave search tool to obtain and return relevant information.

🛠️ Troubleshooting

Incorrect Environment Variable Configuration

• Symptom: Unable to execute tools in community packages.
• Cause: The N8N_COMMUNITY_PACKAGES_ALLOW_TOOL_USAGE environment variable is not set or is set incorrectly.
• Solution: Ensure this environment variable is correctly passed when starting the container.

Tool Discovery Failure

• Symptom: Unable to obtain the tool list.
• Cause: The MCP client node is misconfigured or the credentials attribute is not set correctly.
• Solution: Check the configuration of the MCP client node and ensure it uses the correct credential type (e.g., MCPBraveSearch).

Tool Execution Failure

• Symptom: An error occurs when invoking a tool.
• Cause: The toolName parameter is not passed correctly or the toolParameters mode is set incorrectly.
• Solution:
  • Ensure the toolName dynamic expression is accurate.
  • Set the toolParameters mode as needed (e.g., definedAutomaticallyByModel or manually specify parameters).

📚 Summary

Through this documentation, you have successfully set up an n8n-based AI toolchain capable of dynamically discovering and invoking external tools. This approach not only enhances the system's flexibility but also provides strong support for building automated workflows.